Performance tests

[Alec McEachran]

I've put together some rudimentary tests for comparing the relative
speeds of events and signals. I would welcome all of your feedback:

http://alecmce.com/as3/events-and-signals-performance-tests

Alec

[Ondina]

Hey Alec,
Thanks for the cool tests:)
Here my results for a Flex App and an AIR App(both release build) ,
Windows:

Flex App in Browser:
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
performancetests.CapturedEventsTest (5
iterations)
Player version: WIN 10,1,51,66 (debug)<-----DEBUG version
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
method.................................................ttl ms...avg ms
dataEvent .............................................801 160.20
dataEventOptimised ..............................814 162.80
dataSignal ..........................................1446 289.20
simpleEvent..........................................787 157.40
simpleEventOptimised ........................ 775 155.00
simpleSignal .......................................1290 258.00
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
performancetests.EventsTest (5
iterations)
Player version: WIN 10,1,51,66 (debug)
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
method...................................................ttl ms...avg
ms
dataEvent .............................................462 92.40
dataEventOptimised ............................. 460 92.00
dataSignal ...........................................419 83.80
simpleEvent ....................................... 448 89.60
simpleEventOptimised ......................... 448 89.60
simpleSignal .......................................289 57.80
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

AIR App:
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
performancetests.CapturedEventsTest (5
iterations)
Player version: WIN 10,1,50,334 (regular)
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
method..............................................ttl ms...avg ms
dataEvent.......................................... 552 110.40
dataEventOptimised........................... 551 110.20
dataSignal..........................................861 172.20
simpleEvent....................................... 529 105.80
simpleEventOptimised ....................... 526 105.20
simpleSignal...................................... 788 157.60
––––––––––
performancetests.EventsTest (5
iterations)
Player version: WIN 10,1,50,334 (regular)
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
method...................................................ttl ms...avg
ms
dataEvent...............................................274 54.80
dataEventOptimised .............................. 266 53.20
dataSignal ............................................190 38.00
simpleEvent...........................................247 49.40
simpleEventOptimised ..........................244 48.80
simpleSignal..........................................116 23.20
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

[alec]

Hmm, that's extremely interesting. It implies that FP 10.1 may have
optimised event handling. However, may I point out that the player
really should be a release player, not the debug build (unless you
release debug builds?!)

Thanks for the feedback, this demands closer attention.

[alec]

I have just downloaded and installed the latest Flash Player
10.1.51.66 and ran the tests through Safari on my browser. I get very
different results:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
performancetests.EventsTest (5
iterations)
Player version: MAC 10,1,51,66 (regular)
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
method...................................................ttl ms...avg
ms
dataEvent 2344
468.80
dataEventOptimised 2270 454.00
dataSignal 313
62.60
simpleEvent 2755
551.00
simpleEventOptimised 2285 457.00
simpleSignal 174
34.80
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
performancetests.CapturedEventsTest (5
iterations)
Player version: MAC 10,1,51,66 (regular)
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
method...................................................ttl ms...avg
ms
dataEvent 2708
541.60
dataEventOptimised 2703 540.60
dataSignal 1182
236.40
simpleEvent 2724 544.80
simpleEventOptimised 2717 543.40
simpleSignal 1042 208.40
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

[Ondina]

Hi Alec,
As you can see the performance is better in the AIR application. I
don’t know which are the differences between the Flash players
(Browser<-->AIR). Anyway it seems that for AIR there is a regular
player running ..
I’ll run the tests on my MAC OS too, where as far as I remember I
have a regular Flash player installed, not the debug version for the
browser. But then WIN and MAC are also 2 different worlds…..
I’ll post the results later today or over the weekend.
If Signals are faster than Events it’s a good thing, but to me what
really matters is that Signals aren’t worse than Events :) I’m
trying to use as3-signals inside my robotlegs project.
Ondina

[Ondina]

Hehe, I don’t release debug builds, but the player is a Debugger
Version, that I need during development. I don’t know whether that
matters or not when it comes to speed, memory, etc?
Ondina

[Joel Hooks]

On Jan 27, 2010, at 2:09 AM, Ondina wrote:

> If Signals are faster than Events it’s a good thing, but to me what
> really matters is that Signals aren’t worse than Events :) I’m
> trying to use as3-signals inside my robotlegs project.

Yes and yes. As long as Signals aren't orders of magnitude SLOWER than I am happy.

[alec]

I agree, but speed is a necessary component of the evaluation process
for libraries, so I'd like to get numbers I can rely on!

Also, if there is a major disparity between Mac and Windows here, then
that opens up a whole can of worms that needs to be looked at very
carefully. If it looks like Adobe might make me write the following,
I'm going to give up Flash:

if (isIE) ... (or isWindows in this case!)

Ideally Ondina, I'd suggest we want to test in a release player inside
a browser, since that's where the majority of applications are
targeted (I stand to be corrected on this, and of course, you test
where it's most relevant to you!)