Comments on: Practical Performance Tweaks

By: jonbcampos

jonbcampos — Mon, 07 Feb 2011 15:13:38 +0000

@justin @steven @simon I forget why but there was a reason that I had to build this (at the time) in a non-release version. Either way all the source code is there and available for you to run the tests yourself and see what happens. I will be creating an update soon to this post focusing on newer features.

By: FlashApplications » Practical Performance Tweaks

FlashApplications » Practical Performance Tweaks — Mon, 07 Feb 2011 12:44:57 +0000

[...] Read the good work of him : http://www.unitedmindset.com/jonbcampos/2009/11/30/practical-performance-tweaks/ [...]

By: Simon Richardson

Simon Richardson — Mon, 07 Feb 2011 10:37:40 +0000

Great to see more attention to optimised code, but you should republish your results on your blog in the release player, although the results seem fairly similar compared to the release player the ms difference is a lot. Debug flash player shouldn’t be used for benchmarking as the end user should be in the release player.

By: Steven Sacks

Steven Sacks — Mon, 07 Feb 2011 10:31:07 +0000

I’ve done my own testing and found that the rule of new doesn’t apply when it comes to Vectors. It is faster to use new than it is to create a new vector from an array.

var v:Vector. = new Vector.();
is faster than
var v:Vector. = Vector.([]);

By: Steven Sacks

Steven Sacks — Mon, 07 Feb 2011 10:27:47 +0000

Thanks for this! It’s very comprehensive.

You should absolutely redo all your tests in the release player and post those results. The debug player has dramatically different performance than the release player. Some things are much slower, some things are a little slower, almost nothing is faster.

For the Math.max() test, you should include this and see how it compares:

i = a > b ? a : b;

Also, check out Michael’s blog for other optimizations:
http://lab.polygonal.de/

By: Justin Mclean

Justin Mclean — Mon, 07 Feb 2011 10:14:44 +0000

Great list of tests and thanks for sharing. You may want to rerun the tests with a release swf in the non debug player you may find that the results will vary quite a bit.

By: Jonathan Campos

Jonathan Campos — Fri, 05 Nov 2010 15:44:03 +0000

@Graham For these tests I didn’t turn off update though I should have made both forms of tests. Good call.

By: Graham Phillips

Graham Phillips — Fri, 05 Nov 2010 15:42:53 +0000

Regarding the performance of ArrayCollection, did you do your test using disableAutoUpdate() ? Generally, you disable auto updates while the collection is being processed, then turn it on so that the events that update the view happen once.

By: Optimization Techniques for Air for Android apps | The World In A State of Flex

Wed, 08 Sep 2010 14:54:02 +0000

[...] with my previous posts on performance optimization I like to focus on the big gains and these techniques that are practical. I say practical because [...]

By: Jonathan Campos

Jonathan Campos — Fri, 12 Feb 2010 16:52:19 +0000

I agree that you receive Null. I did make sure to include that warning when using “as”. But I left it in there so I didn’t get people saying I forgot NumberAsString as one of the permutations. One of those situations where you’re damned if you do, damned if you don’t.