I am working on an OpenGL application for the iPhone...
My app has only 2 views:
An OpenGL view and, as a subview for the OpenGL view, a view with the sole purpose of catching touch events...
The problem is that after about 10-15 minutes of keeping the app running on the device, I get a big (0.5s-1s) delay between every touchesMoved:withEvent: call
The animation runs smooth, and CPU usage is also not the problem (10% at most)
I have no idea what might be causing this
That is weird, eh.
This happens ON THE DEVICE right? When you are not running tethered from XCode?
I would guess you are using up a lot of memory, either a leak or just in some way using up more and more memory as time goes on.
Are you familiar with the various memory tools to watch what is going on?
Also, what about this: launch a few other large apps that remain in the background. Run your app until the problem exhibits. Then, kill the other apps. Does the problem suddenly go away? If so that would suggest you're low on memory.
Would be interested to hear.
Related
I'm writing a finger sketch type app and had the app working great on iOS 3.0. However, running the app under iOS 4.0 or greater is causing problems. Specifically, I am receiving touch events to my app approximately 5x slower on the new OS than on the old. This obviously causes my app to draw poorly because I'm capturing 5x less data points to draw between.
Any ideas on how to speed up touch event frequency on iOS4?
Certain graphics operations take a lot longer in iOS4.x than in 3.x (somewhere between 2X to 10X slower). The longer graphics execution times could be blocking the main UI thread, and not leaving enough time for the main thread run loop to handle user (touch) events.
Profile your drawRect code and see if it's now taking longer that one refresh rate tick time. If so, try speeding up or breaking up your graphics renders, or try a lower frame rate, and see if any of that helps.
You should profile your code with Shark to see where the bottleneck is. What is probably happening is some of your code that executes on the main thread (most likely something that runs in response to a touch) is taking longer than expected which is preventing your app from receiving touch events
Block the main thread less.
I have a game that runs fine as is (around 30fps), but fps went down the drain when I tried to implement ads. I tried Greystripe and iAds but with same result (iAds were maybe bit worse). Average fps is almost same, but there are huge spikes all the times (1-2 spikes per second) and game is unplayable.
I guess it is because ad is in another view. I read somewhere that opengl apps on iphone don't like having another views with them, but there is plenty of games with ads on app store. How do they do it?
My implementation should be ok. I did everything as documentation and samples told me. I have my opengl view and ad view as subviews in app window, adview being in front of opengl view and thus covering part of it. Could this be the problem? Is it better to make opengl view smaller to left space for ad so they don't overlap? Do you have any other ideas what could be wrong?
Lope, I've created a gist at this link with a singleton "AdManager" class I wrote to handle iAds using cocos2d. Cocos2d sits on top of OpenGL, of course, and I've found that this code doesn't affect FPS even for relatively complicated games.
You'll have to modify this a bit to work with your application, changing out the cocos2d calls, etc, but this will give you asynchronous loading of iAds, which should help the FPS issue.
To use this class, include its header and call
[[AdManager sharedManager] attachAdToView:self.view];
wherever you need iAds. The ads will remain hidden until an ad loads, at which time they'll pop up at the top of screen. (The class works for iOS 4.0, 4.1 and 4.2).
Also, I should add that I have cocos2d running inside of an overall UIViewController that I call "Cocos2DController". When I attach the ads to a cocos2d view, I'm using
[[AdManager sharedManager] attachAdToView:[[CCDirector sharedDirector] openGLView]];
Best of luck!
We can hit and miss with apple's choices, but go for the sure thing and implement the ads in other parts to be appealing and not intrusive. It will be better for the framerate, and for you.
Try downloading the ads in a seperate, low priority, thread. You can, thus, nsure that the ads loading does not take too much CPU time. With a bit of CPU synchronisation you can make sure you don't try to display the new ad until it is completely ready to display. Sure it will suck some CPU time away from what you are trying to do but set your priorities right and it should only suck time when you are busy doing nothing.
Please excuse the thread necro'ing here, but I've used Stack Overflow a lot to help me through the problems I've had during coding, and thought my experience might be useful to someone in the future.
My simple cocos2d game ran with decent FPS (rarely changed the FPS display at all) until I implemented AdWhirl (integrating AdMob + iAd only). It would then run OK for the first few iterations, but upon upon the 9th or 10th scene refresh (single screen game, time in each scene < 5 seconds on average) the FPS would dive to ~20FPS, and drop again each time the scene refreshed.
Turns out, in my n00biness (this may be particular to me :) ), I was calling the scene from within itself. That is, once the actions had finished, the last action was to call the main scene again (a lazy way of rebuilding the scene for the user to have another go). This init'd the views and view controllers I had inserted to handle the AdWhirl ads all over again, and not only did I have a memory leak, I had 10+ view controllers all trying to request and service ads from AdWhirl. Once I got a clue and took that self-referring loop out, all was good.
Occasionally in low-memory conditions the UIImagePickerController I use gets 'stuck' with the shutter closed. Generally when this the WatchDog is allready jettisoning other background apps so within a second or two the low memory condition is over, yet the UIImagePicker controller is still stuck.
Has anyone else encountered this and implemented any sort of workaround or found a way to detect when the picker is in this state, so that it can be removed and a new one put in its place.
Thoughts / Ideas / Solutions??
Have you tried allocating (and then freeing) a megabyte or two of memory using malloc() before starting the UIImagePicker? Or claim it slightly earlier in your app, and then release it when you want to show the picker.
I guess you should be able to see roughly the amount of memory UIImagePicker needs to run successfully by running your app inside of instruments.
Whilst not ideal, it should cause other apps to be given memory warnings earlier on and hence the UIImagePicker should then have enough memory to run.
That's a known behavior of the camera application too. I guess you can't do anything to it except freeing as much memory as you can before starting it. You could raise a memoryWarning yourself so every application will receive it and start freeing memory but I have no idea how I can do that. I started investigate sending the notification myself (UIApplicationDidReceiveMemoryWarningNotification) without success. I guess we need to send an object along the notification to define the warning level but I'm really not sure.
Is there a way to increase the application priority for the CPU on iPhone? I notice that SpringBoard takes up too much CPU at times causing some fluctuation in performance on the device.
What you're seeing is probably an artifact related to Core Animation. The Core Animation server on the iPhone is owned by Springboard, so many Core-Animation-related actions (calculations related to animation, etc.) appear to be performed by Springboard even though they were initiated by your application. This is not a case of Springboard itself burning CPU time in the background, it's just how some of this processing looks when running Instruments or similar tools against the device.
In this case, I'd look at how you were animating things around the screen (even implicitly using the UIKit interface elements), or how those views or layers were being composited, in order to smooth out your performance issues.
I've observed similar behavior in an Apple example app and a game I am working on. In the game, the behavior is eventually causing the app to crash due to running out of memory. The example app is Touches.
At any point when touches are being tracked, which is when you're moving one of the objects around in Touches, and pretty much any time a touch is down in my game, memory usage goes up steadily, for as long as you continue moving the touch around. Once the touch sequence completes, the memory usage does not go back down. I've gotten Touches, which starts off using less than half a meg, up to about 4MB net allocations with a few minutes playing around. That memory is never deallocated.
So my question is: why does this memory never get deallocated? Am I fundamentally misunderstanding something? Is this a framework flaw? I've read some about issues with the accelerometer and touches leaking, but I'm not using the accelerometer at all in this game.
I apologize if this seems TOO elementary, but...
There's no chance you have NSZombiesEnabled set to YES, do you?
That would prevent the deallocation of any objects whatsoever, and if you did allocate new ones, then EVENTUALLY the app will run out of memory.
Use instruments to find what instances are not being deallocated. I had a similar problem, and all it came down to was that I was calling retain, but not release (due to a logic error).
Im in a very similar situation. In fact, if i just let me app run without touching anything (all it does is run an NSTimer to update the display). And i notice the memory slowly but steadily increasing in usage.
If there are memory leaks its not obvious. Because ive seen memory leaks popup whenever it detects one.
So does that mean there are undetectable memory leaks?