Optimization Tips

random I personaly don't use attach/remove mc (I think I've done that to death on this thread), but as soon as you attach a mc set a flag / increment a counter / shove a value into an array / etc.

If you need to remove it check the flag first. Although I can't think of an example where you wouldn't know what you hadn't attached in the first place ?

Checking a var using an if statement is quicker than testing an mc properties.

As to the thread being too complicated, anyone can ask about things that have been posted for clarification. I guess my tut on FLASM was so good that everyone got it and didn't need to ask anything ( Or no one was that interested in it...hmmmm ) ;)

Squize.

Does anyone have any insight into which one of the following would be faster (appologies if this has already been mentioned):

Code:

---

onClipEvent(load){
  h = _height;
  w = _width;
}
onClipEvent(enterFrame){
  if(h == w){
      //stuff
  }
}

---

Code:

---

onClipEvent(enterFrame){
  if(_height == _width){
      //stuff
  }
}

---

Basically, does it take time to find out attributes? and if so, would it be quicker to assign them to a variable?

Cheers,

The first example ( Storing the properties in a var ) is quicker.

The second one will have to get the properties from the movie clip every time.

Hope that's some help.

Squize.

Thanks squize - its what i thought, but didnt know.

Cheers,

Hey, thought we might as well add filesize optimization. So:

jpegs are generally smaller because they dont support an alpha channel, but if you really want to optimize, you can create the graphics dynamically using code (check out strille's game).

NEVER "break apart" big jpegs/bitmaps that you've imported.

While you're testing out your movie (Ctrl-Enter) check out the top menu: it has a bandwidth profiler and also lets you test how fast your movie will load at diff. conection speeds.

:)

PNGs are good too. Plus they support 24bit alpha channels. And they are small.

Ok i got two more :D

Would it be faster to break the 'for' loop like so, or will it break itself if i tell it to return a value..?

code:

---

MovieClip.prototype.checkActiveTile = function (obj){ 

	for(t = 0; t < activeArray.length; t++){

		if(activeArray[t] == obj){

			return true;

			break;

		}

	}

	return false;

}

---

Aaaand, secondly. Do empty Movieclips contribute to CPU load, i.e. would 10 visible MCs on screen be faster than 10 visible MCs on screen with one empty MC inside each visible one..?

Thanks guys, my games keep getting faster and faster...

Running a while loop in place would be quicker. Also return exits a function straight away so in this case you don't need the break ( The theory is really sound, just needs tweaking a little ).

PHP Code:

---

MovieClip.prototype.checkActiveTile = function (obj){
        var t=0;
        var a=activeArray.length;
        while(t++<a){
                if(activeArray[t] == obj){
                        return true;
                }
        }
        return false;
} 


---

Written cold but it should work ;)

As to your second point, mc's do contribute to CPU load in some way no matter what.
Do you mean having 10 mcs running but pointing to a blank frame being faster than 10 "normal" mcs ? Yeah 'cause there would be no data to plot ( Sorry if I've misunderstood the question ;) ).

I imagine the way the player works is ( This is just a guess, if someone wants to put me right ):
When plotting an mc it checks the visible property. If it's false it'll just move onto the next mc to plot. If the visible property is true it'll run through a linked list containing all the line data ( Length, start / end points, colour, alpha etc ) and plot all that out. So for a blank mc the linked list will just be a terminator.
Therefore quicker than plotting a usual mc but slower than using the visible property.

BTW How do you put the "actionscript" tag in your post ?

Squize.

Cheers squize, as usual youve been a big help...

I know a while loop is faster, but its it really fast enough for me to bother replacing all my 'for' loops?

P.s. tag is: [as ] [/ as]

:D

Thanks for the tag info.

I'm a great believer in only optimising when it's needed. If your game runs nice and smooth on what you consider to be the target spec machine then why worry ?
If it works and more importantly you like playing it, then it's job done.

If you really need that extra mph in your game then just tackle the time critical sections, the parts that have to run every frame. If you've got for loops in there then yeah replace them with while. It's not a huge amount of effort, just boring :)

I mean your map plotting routine doesn't need optimising, it's only called at the start of a level, whereas your collision routines should be as quick as possible.

It's not just the actual code you want to look at but how you do things ( If that makes sense ).
For example, I'm writing an asteriod game at the moment. For the collision detection I just split the screen up into quarters. Every time a rock moves it calculates which quarter it's in and the mc gets stored in the relevant array. Then when it comes to collision checking I just work out which quarter the player is in and then use a similar approach to the one pred posted ( Cycling through the array ) but just on the specific quarter.
The same thing could be easily done with a platform game.

There are so many approaches to doing things out there that it's well worth hunting around. No matter how good you are someone else has written a routine better than you ever could.
Once you've got that routine then you can blitz it with code optimisation.

Sorry about the rambling, some times the muse just takes hold ;)

Squize.

Right, well this is a function which needs fast optimization because it can be called more than once a frame :rolleyes: ...

I made this function for removing an item from an array:

code:

---

MovieClip.prototype.removeObject = function (obj){ 

	var t = 0

	var a = objectArray.length

	while(t<a){

		var tmp = objectArray.shift();

			if(tmp != obj){

				objectArray.push(tmp);

			}

		t++;

	}

}

---

But have recently found the below, and was wondering which is likly to be faster? Its not for arrays longer than 10 items...

code:

---

Array.prototype.findAndRemove = function(toRemove)

{

	this = this.toString().split(ToRemove + ",").join("").split(",");

};

---

This should be the random10122 & Squize thread ;)

In your first example why are you incrementing the counter var ( t ) outside of the while statement ? while (t++<a) would be quicker ( Or even better using the -(-) thing, to be found somewhere in this thread ).

Have you set up a test with getTimer and a huge array using the two different prototypes ?

I would say the first would be quicker, the second looks like an awful lot of data manipulation ( Although nicer code ). The second one is converting the whole array to a string, and then getting Flash to look for a specific ( Or group of ) character and then removing it.

Is this idea any use to you: Instead of removing values from the array possibly multiple times in a frame, could you just set the object = null ? Then on any routines which need this data ( Collision etc. ) just check if the obj != null
Then when you have a quiet moment in your game run the prototype as a bit of housekeeping. By quiet moment I mean baddies not shooting, tiles not being updated etc. as opposed to changing levels, player killed etc.

Anyway just a thought.

Squize.

Ahh thanks again squize.. If i used while (t++<a)

Would it not miss out '0' ? Also, with the getTimer thing im not really sure how to do that. It would be handy if we could have an FLA for this thread though where you download it - put your code in a space and run it. Then it tells you how fast it is, so you can compare techniques quickly and easily - then post the results here rather than asking (like im doing currently...)

Cheers,

Sorry my bad, you'd have to set the t=-1;

code:

---

var t=-1;

var a=100;

while (t++<a){

	trace(t);

}

---

Here's RipX's timing routines ( From way back on page 4 of this thread ).

code:

---

_global.MAX_VALUE = 39999;

function loop() {

	var z;

	var ms = getTimer();

	trace("start test");

	for (z=0; z<MAX_VALUE; z += 10) {

		z += 1;

		z += 1;

		z += 1;

		z += 1;

		z += 1;

		z += 1;

		z += 1;

		z += 1;

		z += 1;

		z += 1;

	}

	trace("result : "+(getTimer()-ms)+" ms");

}

loop();

---

It's handy if everyone uses the same timer so there won't be any big disagreement where figures don't match.

Squize.

http://webreference.com/programming/...hap10/3/2.html


Just been looking through some of this stuff, most of it should apply to Flash as Javascript and flash are based on exactly the same principles and building blocks - so im lead to believe.

The last loop on that page is (in theory) four times faster than a 'for' loop... :O

Shocking.

Tested that out with some of my engine code and it wasn't any faster than the for loops I'm using! I cannot see a speed advantage in using for loops over while loop - even though everyone say while loops are faster - I just can't get a result to prove that?

RipX

Im not sure if this has already been mentioned. I hope not... because i went away and actually timed stuff myself instead of asking :eek: :)

code:

---

// loop 2 appears to be upto 33% faster //



loops = 20000;



ms = getTimer()



i=loops;

while(--i>-1){

	c++

}



trace("loop 1:"+(getTimer()-ms)+" - "+c)



ms = getTimer()



i=loops;

while(i-->0){

	b++

}



trace("loop 2:"+(getTimer()-ms)+" - "+b)

---

NEVER use loops, unless for initialisation. ALWAYS use ASBroadcaster to keep fps alive

What if it's just used in a function and called as needed?

- Explain AS broardcaster if your saying it's good!

RipX

For a fantastic setInterval replacement using ASBroadcaster check out the always fantastic http://www.andre-michelle.com/ ( Direct link: http://www.andre-michelle.com/files/...sedInterval.as but do yourself a favour and check his site out ).

Now he is the daddy :)

Squize.