Comments on: Events vs. Callbacks

By: Hays Clark

Hays Clark — Mon, 05 Sep 2011 19:02:00 +0000

Interesting, this makes as lot of sense, but I think that there’s another detail which is masking the truth. ( I might be wrong, but bare with me for a sec. )

Your ‘Callback’ system is faster then the native Event system in tests, but this doesn’t make sense because the Event system already has the upper hand. The Event system is native ‘Machine Code’ already because it’s compiled into the Player and should run as fast as C code, vs your callback system is ‘Interpreted Code’ and can’t run as fast as the built in player functionality. So something else is making the Event code run SLOWER then your ‘Interpreted Code’. I believe the Flash Players JIT is causing this slowdown. Here is a little article about the JIT in Flash. The key part is that Constructors remain in ‘ActionScript Byte Code’ while everything else is converted to ‘Machine Code’. It’s my guess that because the Event systems heavy uses Constructors, that is the reason why the Event system is slow. By elimination the cloning it’s optimizing the slow construction of objects in Flash, however, you might also be able to speed up the Event system without having to kill the clone functionality. A good test might be to optimize the constructors by only using and init() method and see if your performance speeds up.

http://blog.classsoftware.com/index.cfm/2010/11/4/Heavy-constructor-JIT-optimisation-in-ActionScript#c33099700-D56E-59C1-07EACFF9ED89BF3A

-Hays

By: L_B

L_B — Tue, 30 Aug 2011 09:28:00 +0000

My mistake, i was too fast…sorry for the flood (you can delete our conversation below)

By: L_B

L_B — Tue, 30 Aug 2011 09:03:00 +0000

The choice of your event “change” (Event.CHANGE) impact your first performance test.
Replace “change” by “Change” for example and you will see the difference.

By: Troy Gilbert

Troy Gilbert — Mon, 29 Aug 2011 16:42:00 +0000

The choice of event name has zero impact on the discussion above.

By: L_B

L_B — Mon, 29 Aug 2011 12:48:00 +0000

The problem come of the name you use for your event : “change” (Event.CHANGE).
If you use Event.CLEAR for exemple or a custom string you will not have this problem.
If you use “change”, you’ll have a lot of dispatch of your event and therefore a lot of duplication i think.

By: Csj4032

Csj4032 — Thu, 21 Jul 2011 01:02:00 +0000

Thanks

By: Anonymous

Anonymous — Sun, 22 Aug 2010 20:15:00 +0000

You’re right, see my follow-up post: http://troygilbert.com/2009/09/events-vs-callbacks-revisited/

By: Amy

Amy — Sun, 22 Aug 2010 20:00:00 +0000

Did you verify that when you override the clone() method to return this that the “listener” function actually gets called? My experiments say it doesn’t (http://insideria.com/2009/12/can-event-performance-be-impro.html), which would certainly explain your increased performance when you do that :).

By: rinn's me2DAY

rinn's me2DAY — Sun, 02 May 2010 09:41:33 +0000

퍼플린의 생각…

Event vs CallBacks – Troy Gilbert 관련 포스트를 블로그에 하나 쓸까 했으나.. 아직 언제 시간이 될지 모르것군 -ㅅ-;;;;…

By: troygilbert

troygilbert — Fri, 19 Mar 2010 20:45:59 +0000

Go ahead, feel free to translate and post. Please just include a link back to this article (or to my homepage, http://troygilbert.com/). Thanks!