I have a bit of a memory leak issue in my Flex application, and the short version of my question is: is there any way (in AcitonScript 3) to find all live references to a given object?
What I have is a number of views with presentation models behind each of them (using Swiz). The views of interest are children of a TabNavigator, so when I close the tab, the view is removed from the stage. When the view is removed from the stage, Swiz sets the model reference in the view to null, as it should. I also removeAllChildren() from the view.
However when profiling the application, when I do this and run a GC, neither the view nor the presentation model are freed (though both set their references to each other to null). One model object used by the view (not a presenter, though) IS freed, so it's not completely broken.
I've only just started profiling today (firmly believing in not optimising too early), so I imagine there's some kind of reference floating around somewhere, but I can't see where, and what would be super helpful would be the ability to debug and see a list of objects that reference the target object. Is this at all possible, and if not natively, is there some light-weight way to code this into future apps for debugging purposes?
Cheers.
Assuming you are using Flex Builder, you could try the Profiler. In my experience, it's not so good for profiling performance, but it's been of great help for finding memory leaks.
It's not the most intuitive tool and it takes a while to get used to it (I mean, to the point where it actually becomes helpful). But, in my opinion, investing some time to at least learn the basics pays off. There's an enormous difference between just seeing how much memory the player is using globally (what System.totalMemory gives you, a very rough, imprecise and often misleading indicator) and actually track how many instances of each object have been created, how many are still alive and where were they allocated (so you can find the potential leak in the code and actually fix it instead of relying in black magic).
I don't know of any good tutorials for the FB profiler, but maybe this'll help to get you started.
First, launch the profiler. Uncheck performance profiling and check everything else (Enable memory profiling, watch live memory data and generate object allocation stack traces).
When the profiler starts, you'll see stats about the app objects, grouped by class. At this point, you might want to tweak filters. You'll see a lot of data and it's very easy to be overwhelmed. For now, ignore everything native to flash and flex stuff, if possible, and concentrate on some object that you think it should be collected.
The most important figures are "cumulative instances" and "instances". The first is the total number of instances created so far; the second, the number of said instances that are still alive. So, a good starting point is get your app to the state where the view you suspect that leaks gets created. You should see 1 for "cumulative instances" and "instances".
Now, do whatever you need to do to get to the point where this view should be cleaned up (navigate to other part of the app, etc) and run a GC (there's a button for that in the profiler UI). A crucial point is that you will be checking the app behaviour against your expectations -if that makes sense-. Finding leaks automatically in a garbarge collected environment is close to impossible by definition; otherwise, there would be no leaks. So, keep that in mind: you test against your expectations; you are the one who knows the life cycle of your objects and can say, "at this point this object should have been collected; if it's not, there's something wrong".
Now, if the "instances" count for you view goes down to 0, there's no leak there. If you think the app leaks, try to find other objects that might not have been disposed properly. If the count remains at 1, it means your view is leaked. Now, you'll have to find why and where.
At this point, you should take a "memory snapshot" (the button next to the Force GC button). Open the snapshot, find the object in the grid and double click on it. This will give you a list of all the objects that have a reference to this object. It's actually a tree, and probably each item will contain in turn a number of backreferences and so on. These are the objects that are preventing your view from being collected. In the right panel, also, you will an allocation trace. This will show how the selected object was created (pretty much like a stack trace).
You'll probably see a hugh number of objects there. But your best bet is to concentrate in those that have a longer life cycle than the object you're examining (your view). What I mean is, look for stage, a parent view, etc; objects on which your view depends on rather than objets that depend on your view, if that makes sense. If your view has a button and you added a listener to it, your button will have a ref to your view. In most cases, this is not a problem, since the button depends on the view and once the view is collect, so is the button. So, the idea is that since there are a lot of objects, you should try to stay focused or you will get nowhere. This method is rather heuristic, but in my experience, it works.
Once you find the source of a leak, go back to the source, change the code accordingly (maybe this requires not just changing code but refactoring a bit). Then repeat the process and check whether your change has caused the desired effect. It might take a while, depending on how big or complex is your app and how much you know about it. But if you go step by step, finding and fixing one problem at the time, you'll eventually get rid of the leaks. Or at least the worst and more evident ones. So, while a bit tedious, it pays off (and as a nice aside, you'll eventually understand what a waste of time is in most cases to use weak refs for every single event handler on the face of this earth, nulling out every single variable, etc, etc; it's an enlightening experience ;).
Hope this helps.
Flash GC uses a mix of ref counting and mark and sweep, so it does detect circular references. It seems rather you're having another reference in you object graph. The most common reason is, that the objects you want disposed still are having event handlers registered on objects that are not disposed. You could try to ensure that handlers are always registered with weak reference. You could also override addEventListener and removeEventListener in all (base) classes, if possible, to look which listeners are registered and whether there are chances for some not to be removed.
Also, you can write destructors for your objects, that for ui components clear graphics and remove all children, and for all objects, removes references to all properties. That way, only your object is kept in RAM, which shouldn't require much memory (a small footprint of 20 B or so, plus 4 B per variable (8 for a Number)).
greetz
back2dos
also a useful heuristics for finding memory leaks: http://www.tikalk.com/flex/solving-memory-leaks-using-flash-builder-4-profiler
Related
I have an issue with a Flex application, one that I didn't build, so I can provide all my findings but sorry for any lack of clarity.
There is a Flex app with 7 main views. And there is a memory issue when navigating between views.
All these views were in a ViewStack, but due to some involving 3D objects I assumed it was too much to have it all in the display list. I'm now clearing all children from the stack and adding/removing them when needed. This gave a small performance increase, but still becomes unresponsive with use. The strange thing is, with this and the original method, the CPU climbs with use but eventually levels out somewhere. Now I'm creating new instances of each screen when they are navigated to and setting the previous variable to null. Now it looks like CPU is spiking when the view is created, but leveling out to something much much lower than it was. This felt like progress, but now the available memory keeps climbing where it wasn't before....
My understanding was calling remove child or remove all children would mark the object for deletion when the garbage collector next ran. I can't see any other references to the instance. My code is along the lines of
this.parentApplication.viewstack.removeAllChildren();
this.parentApplication.viewstack.addChild(new HomeScreen);
I have a function for each button to add a new instance like the above.
The only thing I can see and feel silly asking but need confirmation, is each view extends a class called "Screen", this class contains a singleton reference to some core components
this.model = PancakeApplication.instance.model;
this.meaModel = MeaApplication.instance.meaModel;
this.meaModel.addEventListener(ScreenChangeEvent.SCREEN_CHANGE, electedScreenChangeHandler);
Would this trick the garbage collector into thinking it was still needed?
General advice on clearing Objects from the memory pool would awesome!!! I've never needed to analyze the Flash Player in such depth.
SOLVED: I think it's an error with sound drivers, removing all sound and shes purring like a kitten. Works on my machine fine with windows XP, but not on the touch pad the application is crashing on with windows 7 (unsure of the drivers looking into them now).
UPDATE: Now I'm thinking its not the drivers, tried 3 different versions, all with no improvement. I did discover the sound was fading in and out with the TweenLite lib. Doesn't look like there are any memory leaks in TweenLite as it works fine on other machines. Just the use of volumeEasingFunction seems to consume increasing amounts of CPU until it freaks out. It is crappy hardware running windows 7, which doesn't help.
The first thing that comes to mind is you should be setting the use weak reference parameter to true in your event listener. It is the fifth parameter, so in your example:
this.meaModel.addEventListener(ScreenChangeEvent.SCREEN_CHANGE, electedScreenChangeHandler, false, 0, true);
Grant Skinner has a great 3 part series on AS3 Resource Management that would probably help you get a better idea of what to look for. You can find the details about weakly references listeners in part 3 or in a standalone article written before part 3 was posted.
I'm in the process of trying to 'learn more of' and 'learn lessons from' functional programming and the idea of immutability being good for concurrency, etc.
As a thought exercise I imagined a simple game where Mario-esq type character can run and jump around with enemies that shoot at him...
Then I tried to imagine this being written functionally using immutable objects.
This raised some questions that puzzled me (being an Imperative OO programmer).
1) If my little guy at position x10,y100 moves right 1 unit do I just re-instantiate him using his old values with a +1 to his x position (e.g x11,y100)?
2) (If my first assumption is correct)
If my input thread moves little guy right 1 unit and my enemy AI thread shoots little guy and enemy-ai-thread resolves before input-thread then my guy will loose health, then upon input thread resolving, gain it back and move right ...
Does this mean I can't fire-&-forget my threads even with immutability?
Do I need to send my threads off to do their thing then new()up little guy synchronously when I have the results of both threaded operations? or is there a simple 'functional' solution?
This is a slightly different threading problem than I face on a day to day basis.
Usually I have to decide if I care about what order threads resolve in or not. Where as in the above case I technically don't care if he takes damage or moves first. But I do care if race conditions during instantiation cause one threads data to be totally lost.
3) (Again if my first assumption is correct) Does constantly instantiating new instances of an object (e.g Mario guy) have a horrible overhead that makes it a very serious/important design decision ?
EDIT
Sorry for this additional edit, I wasn't what good practice is on here about follow up questions...
4) If immutability is something I should strive for and even jump though hoops of instantiating new versions of objects that have changed...And If I instantiate my guy every time he moves (only with a different position) don't I have exactly the same problems as I would if he was mutable? in as much that something that referenced him at one point in time is actually looking at old values?.. The more I dig into this the more my head's spinning as generating new versions of the same thing with differing values just seems like mutability, via hack. :¬?
I guess my question is: How should this work? and how is it beneficial over just mutating his position?
for(ever)//simplified game-loop update or "tick" method
{
if(Keyboard.IsDown(Key.Right)
guy = new Guy(guy){location = new Point(guy.Location.x +1, guy.Location.y)};
}
Also confusing is: The above code means that guy is mutable!(even if his properties are not)
4.5) Is that at all possible with a totally immutable guy?
Thanks,
J.
A couple comments on your points:
1) Yes, maybe. To reduce overhead, a practical design will probably end up sharing a lot of state between these instances. For example, perhaps your little guy has an "Equipment" structure which is also immutable. The new copy and the old copy can reference the same "equipment" structure safely, since it's immutable; so you only have to copy a reference, not the whole thing. This is an common advantage you only get thanks to immutability -- if "equipment" was mutable, you couldn't share the reference, since if it changed, your "old" version would change too.
2) In a game, the most practical solution to this issue would probably be to have a global "clock" and have this sort of processing happen once, at a clock tick. Note that your exact scenario would still be a problem if you didn't write it in a functional style: Suppose H0 is the health at time T. If you passed H0 to a function which made a decision about health at time T, you took damage at time T+1, and then the function returned at time T+5, it might have made the wrong decision based on your current health.
3) In a language that encourages functional programming, object instantiation is often made as cheap as possible. I know that on the JVM, creating small objects on the heap is so fast that it's rarely a performance consideration in any practical situation at all, and in C# I've never encountered a situation where it was a concern either.
If my little guy at position
x10,y100 moves right 1 unit do I just
re-instantiate him using his old
values with a +1 to his x position
(e.g x11,y100)?
Well, not necessarily. You could instantiate the guy once, and change its position during play. You may model this with agents. The guy is an agent, so is the AI, so is the render thread, so is the user.
When the AI shoots the guy, it sends it a message, when the user presses an arrow key that sends another message and so on.
let guyAgent (guy, position, health) =
let messages = receiveMessages()
let (newPosition, newHealth) = process(messages)
sendMessage(renderer, (guy, newPosition, newHealth))
guyAgent (guy, newPosition, newHealth)
"Everything" is immutable now (actually, under the hood the agent's dipatch queue does have some mutable state probably).
If immutability is something I
should strive for and even jump though
hoops of instantiating new versions of
objects that have changed...And If I
instantiate my guy every time he moves
(only with a different position) don't
I have exactly the same problems as I
would if he was mutable?
Well, yes. Looping with mutable values and recurring with immutable ones is equivalent.
Edit:
For agents, the wiki is always helpful.
Luca Bolognese has an F# implementation of agents.
This book (called by some The Intelligent Agent Book), though targeting the AI applications (instead of having a SW engineering point of view) is excellent.
If everything in the global system state, outside the current stack frame, is immutable, unless gives another thread a reference to something on the stack (VERY DANGEROUS) there won't be any way for a threads to do anything to affect each other. You could fire and forget, or simply not bother firing in the first place, and the effect would be the same.
Assuming there are some parts of the global state that are mutable, one useful pattern is:
Do
Latch a mutable reference to an immutable object
Generate a new object based upon the latched reference
Loop While CompareExchange fails.
The compare exchange should update the mutable reference to the new one if it still points to old one. This avoids the overhead of locking if there is no concurrent access, but may perform worse than locking if many threads are try to update the same object and generating a new instance from the latched one is slow. One advantage of this approach is that there is no danger of deadlock, though in some situations liveLock could occur.
Another functional approach to this sort of problem is to take a step back and separate out the idea of state from the idea of your little guy.
Your state will include your little guy's position, as well as the position of your baddy and it's shot, and then you have some functions that take some or all of the state and do things like generating the next state and drawing the screen.
The timing issues you're talking about when things you want to parallelize depend on each other are real problems that won't magically go away, although the solutions may be more or less convenient in different languages.
Several suggestions have already been made, and there are a variety of concurrency solutions. The central clock and agents would work, as would Software Transactional Memory, Mutexes or CSP (go style channels), and probably others. The best approach is going to depend on the specifics of the problem, and to a certain extent on personal taste.
As for the head-spinning, try not to get too caught up in whether a thing is changing or not. The point of immutability is not that things don't change, it's that you can create pure functions so that your program is easier to reason about.
For example, an OO program might have a drawing function that iterates over all the objects in a scene, and asks them all to draw themselves, where a functional program might have a function that takes a state and draws a frame.
The end result would be the same scene, but the way the logic and the state is organised is very different.
I, for one, find that it's much easier to work on when you have all the data over here, in one big input lump, and all the drawing logic there, encapsulated in some functions. There are some pretty clear architectural wins too - serialization, testing, and swapping out front ends gets a lot easier with this sort of structure.
Not everything in your program should be immutable. A player's position is something you would expect to be mutable. His name, maybe not.
Immutability is good, but you should perhaps rethink your approach to use more concurrent solutions than simple "immutable"ize everything. Consider this
Thread AI gets copy of your position
You move three units to the left.
AI shoots you based on your old position, and hits... shouldn't happen!
Also, most gaming is done in "game ticks" - there's not much multithreading going on!
I'm currently working on a very large Flash platformer game (hundreds of classes) and dealing with an issue where the game slowly grinds to a halt if you leave it on for long enough. I didn't write the game and so I'm only vaguely familiar with its internals. Some of the mysterious symptoms include,
The game will run fine for a determinate amount of time (on a given level) when all of a sudden it will exponentially start leaking memory.
The time it takes for the game to hit the point where it leaks exponentially is shorter when there's more sprites on the screen.
Even when nothing is being visibly rendered to the screen, the game slows down.
The game slows down faster with more frequent sprite collisions.
Completely disabling the collision code does slow down the degradation but doesn't prevent the game from eventually dropping frames.
Looking at the source and using Flex profiler, my prime suspects are,
There are many loitering objects, especially WeakMethodClosure, taking up large amounts of memory.
The program makes extremely extensive use of weak event listeners (dozens dispatched per frame).
BitmapData is being copied every time a new sprite is created. These are 50x50 pixel sprites that spawn at about 8 sprites per second.
I know it's next to impossible to tell me the problem without seeing the source, so I'm just looking for tidbits that might help me narrow this down. Has anybody experienced this evasive performance degradation in their own projects? What was the cause in your case?
I recently completed a optimization of an large project.
And I can give you some architectural advices:
Main principle – try do as little as possible function / events calls
Get rid off all but one
onEnterFrame / onInterval / onTimer cycles. Do everything you
need in one general event call.
You probably will need a lot of static arrays for storing processed object
references.
Do your graphics / render stuff also in one main loop
Try use big (probably prerendered) canvases instead of small sprites / bitmaps.
Usually it usable for backgrounds. But also working good for a smaller objects
(like trees, platforms, etc)
Rid off small bitmap resources, assemble it to one tile-sheet and draw your stuff
directly from it, through source–rect property
I hope its helps you!
Memory leaks – such headash.
P.S. Test your game in different browsers, IE – most leaking of all, sometimes it don`t clear memory every after refresh.
Avoid anonymous methods - change them into class level methods.
Use weak reference in addEventListener and/or make sure you remove all the listeners of an object before removing it with removeChild
Make sure you removeChild all the sprites instead of just letting them fly off the screen. Also, if possible, reuse the sprites instead of creating new ones.
You should consider object pooling if you have a lot of creation/destruction going on, especially with heavy objects like bitmapdata.
see Object Pool class
Sounds like you need to profile your application to see what is happening.
This thread had a couple of suggestions, but, ultimately, you will need to just put in code to help determine what is going on.
Profiling ActionScript-3 Code
You may want to see if you can just run some smaller parts over a period of time and see if you see a slowdown.
You may want to unit test your application, so you can run various parts rapidly, looking for memory leaks.
One framework is: http://asunit.org/, and another is: http://opensource.adobe.com/wiki/display/flexunit/
Unit testing is something I use a great deal for profiling, so you can test at the top level of your game, run it thousands of times, looking for problems, then run each part and see which has problems, and just work your way down. This is a manual process, but if the two ideas in the SO thread listed in the beginning don't help, this may be your best approach.
Are you using too much memory? Or is your cpu usage too high?
First determine if its a memory or processor limit you are hitting. It sounds like the later, seems like there are many objects doing stuff around ... likely those extra sprites aren't being freed well. Look for dependencies among objects / variables / anything in those events, make sure the sprites are removed, pay attention to any handlers of EnterFrame or recurring events.
It sounds far more likely that you are capping out on processor speed than memory. You have to try extra hard to memory cap a Flash application.
Fortunately there are a lot of easy things you can do to keep CPU down...
1) Stringently manage event listeners for everything, especially mouse listeners. Do you have $texas event listeners on all your sprite objects? That could be a problem.
2) Access arrays using int or uint instead of Number. This is huge, and this is one of those backwater Adobe tricks. Accessing array object with int and uint is way faster than Number and if you do a lot of iteration (and it sounds like you do) this could shave precious milliseconds off your frame execution.
3) In the same vein as #2, monitor your math operations and what types you are using for certain operations. The slowest thing you can do in math operations for AS3 is repetitive casting (feeding ints to a function that returns a Number), or performing basic operations like add and subtract on Number instead of int.
The great thing about having a wtfhuge program like this in Flash is that even a minor optimization change could have a major impact on performance. I once toyed with a raytrace engine in AS3 where I declared one extra variable and it killed my FPS from 30 to 23.
Flash has a rather infamous issue (many consider it a bug) that causes event listeners for timers and the ENTER_FRAME event to not get garbage collected, even if they were weakly referenced. So, even though it's good practice to use weakly referenced events, you should still remove all your event listeners when they are no longer needed.
I have been working on an Adobe Flex application for some months now, and the application is meant to run 24/7 for days (weeks!) continuously. However, I'm now seeing that after a few days of running nonstop the computer it runs on tells me that the system is low on virtual memory and gives me an error about Page File usage. Once I close the Flex app, the Page File usage goes down from 1.9 GB to 100 MB (or less). It seems that its using up all this memory and not freeing it although I have been very careful in my app to not keep huge arrays.
The app does some graphing and draws a lot of shapes (to greate a 'gauge') and then gets rid of them by re-declaring that object as another 'gauge'.
Any idea why my page file usage is climbing so high?!
You most probably have eventListeners that are not being removed. They keep references to objects and prevent them from being garbage collected.
You can use the profiler in Flex Builder professional to see where your memory usage is going. Like another poster mentioned, event listeners are alot of times the culprits in cases like this, but more generally, just because you think you are getting rid (destroying or deleting) a variable, doesn't mean that it is really getting taken care of by the garbage collector. If any reference (like an event listener) still exists to that variable (or object) it will not be collected. The profiler will point out these things.
I've heard rumors that putting anything on the Stage will create memory leaks. In other words, you can be as careful as possible with your code, but you'll still leak memory. This has not been validated by Adobe, as far as I know. A good test might be to instantiate a Shape and a Sprite and a MovieClip, add them to the display list, and then let the app run overnight. Would love to hear the results if you do end up testing this.
Is it possible to programmatically force a full garbage collection run in ActionScript 3.0?
Let's say I've created a bunch of Display objects with eventListeners and some of the DO's have been removed, some of the eventListeners have been triggered and removed etc... Is there a way to force garbage collection to run and collect everything that is available to be collected?
Yes, it's possible, but it is generally a bad idea. The GC should have a better idea of when is a good time to run than you should, and except for a very specific case, like you just used 500MB of memory and you need to get it back ASAP, you shouldn't call the GC yourself.
In Flash 10, there is a System.gc() method you can call (but please don't, see above) - keep in mind System.gc() only works in the debugging version of Flash player 10+.
In Flash 9, there is an unsupported way to force it via an odd LocalConnection command, but it may not work in all versions. See this post by Grant Skinner.
There is a new API for telling the GC that it might be a "relatively good moment" to collect.
See the Adobe API docs for
System.pauseForGCIfCollectionImminent
And also this Adobe blog post from shortly after the method was introduced in Player version 11
The method takes an "imminence" argument; basically, you feed in a low number (near 0.0) if you really want the collector to run, even if there has not been much activity (currently measured by bytes-allocated) since the last collection, and you feed in a large number (near 1.0) if you only want the collection pause to happen if we were already near the point where a collection would happen anyway.
The motivation here is for situations in e.g. games where you want to shift the point where GC's happen by a small amount, e.g. do the GC during a change of level in the game, rather than two seconds after the player started exploring the level.
One very important detail: This new API is supported by both the Release and the Debugger Flash Runtimes. This makes it superior to calling System.gc().
For all currently released versions, System.gc() only works in the debug version of the Flash player and ADL (the debug environment for AIR apps). Flash player 10 beta currently does work in all flavors.
I agree with Davr, it's a bad idea to do. The runtime will usually have a better idea than you do.
Plus, the specifics of how the garbage collector works is an implementation detail subject to change between flash player versions. So what works well today has no guarantee to work well in the future.
As others said: do not try to GC manually, there are hacks but it's not safe.
You should try recycling objects when you can - you'll save a lot of memory.
This can be applied for instance to BitmapDatas (clear and reuse), particles (remove from display and reuse).
I have a comment on those saying you should never do GC manually. I'm used to manual memory management in C++ and I prefer sharedptr a lot over GC, but anyway.
There is a specific case where I can't find another solution than do a GC. Please consider: I have a DataCache class, the way it work is it keeps result objects for certain method calls that send out updated events when refreshing/receiving data. The way the cache is refreshed is I just clean all results from it and send the event which causes any remaining listeners to re-request their data and listeners that went out of scope should not rerequest which cleans out not needed results. But apparently, if I can't force all listeners that still dangle waiting for GC to be cleaned up immediatly before sending out the "ask you data again" event, those dangling listeners will request data again unnecessarily. So since I can't removeEventListener because AS3 doesn't have destructors I can't see another easy solution than forcing a GC to make sure there's no dangling listeners anymore.
(Edit) On top of that I cannot use removeEventListener anyway for binding which were setup in mxml, for example (using my custom DataCacher class which handles remoteobj)
<mx:DataGrid id="mygrid" dataProvider="{DataCacher.instance().result('method').data}" ... />
When the popup window containing this datagrid is closed, you would expect the bindings to be destroyed. Apparently they live on and on. Hmm, shouldn't flex destroy all bindings (meaning eventlisteners) from an object when it's being marked for GC because the last reference is deleted. That would kinda solve the problem for me.
At least that's why I think, I'm still a beginner in Flex so any thoughts would be appreciated.
try {
new LocalConnection().connect('foo');
new LocalConnection().connect('foo');
} catch (e:*){
trace("Forcing Garbage Collection :"+e.toString());
}
If you have to, calling the gargabe collector could be useful... so, you have to be carefull how and when you do it, but there is no doubt that there are times when is neccesary.
for example, if you have an app that is modular, when you change from one view to the other, all the deleted objects could represent a large amount of memory that should be available as faster as possible, you just need to have control of the variables and references you are disposing.
recycling doesn't really help. I used one loader that repeatedly loaded the same jpg every 500ms. task manager still reported a non stop increase in memory.
tried and proven solution here.
http://simplistika.com/as3-garbage-collection/