I am using opengles to display a LOT of triangles.
When I run the app on my phone (a Samsung mini) if I display more than about 5,000 triangles the app exits as soon as it attempts to draw. There is no error message, it just exits. Anything between 4 and 5k triangles runs on the phone but the screen flashes white occasionally as I move the camera.
I assume this is a limitation of the phones abilities. It works fine on my Asus tablet.
My question is simply do you agree that this a limitation of the phone?
I have not included any code here because as I say it works fine on the tablet.
If you agree I will see what I can do to optimize the code.
Many thanks for your time folks.
Related
I have a mobile ThreeJS scene based on device orientation that goes quite good on Android devices (55 fps). After some time performance starts to drop badly, but this is not reflected in the FPS, only in the smoothness of the device orientation detection. And only here, because if I leave the phone on a desk (so device orientation doesn´t play any role) the rest of the code works fine and smooth as expected.
I noticed that closing the browser or even restarting the device doesn´t fix the lack of smoothness. In a moment of dispair, I tried the same script on Firefox mobile, and it was smooth. Going back to Chrome, the problem seemed to be fixed (!).
So every time Chrome gets this lag, I open Firefox to solve it, the weirdest thing. What is happening here? How can I prevent it? Also how can be possible that the fps continue being so high but the device orientation goes so laggy?
If I remember well, this started to happen after using layers on my code (combined with the StereoEffect I was already using). Does this have any sense for anybody? I´m sooo stuck on this.. Any clue will be appreciated.
I've seen claims on the net that the newly released iOS 7.1's iBeacon support.
Specifically:
The system is supposed to still notify your app about
didEnterRegion/didExitRegion events, even if the user explicitly
kills your app.
didEnterRegion/didExitRegion notifications are
supposed to be faster from the background and/or with the device
locked.
I have not been able to confirm either of these claims with my own testing. In fact, I seem to be less likely to get didEnterRegion/didExitRegion notifications from a locked device. (more accurately I seem to get didEnterRegion notices, but not didExitRegion notices). That could be because Apple made me remove my BLE background mode entries in my info.plist - I'm not completely sure. I'm still trying to sort this out.
I had trouble setting up my tests at first, but I have witnessed background region entry callbacks after killing an app in iOS 7.1 on both iPhone 4s and iPhone 5s models. See comments below for testing details and instructions to reproduce.
I have also done tests on background detection times on an iPhone 4S, and I still see delays of 15 minutes on iOS 7.1. My full test results and methodology are described here.
Finally, I have also done some tests on the fluctuations on the "accuracy" (distance in meters) measurement on the same device before and after the upgrade to iOS 7.1. I do not see an obvious difference in the noise on the estimate. The graphs below show results before and after the upgrade, with an iBeacon 0.5 meters away for 60 seconds then moved to 3 meters away for 60 seconds. In both cases, the transmitter was a properly calibrated iPhone 4S w/ iOS 7.1 and the receiver was an iPhone 5S.
iOS 7.0.6 Estimated distance
iOS 7.1 Estimated distance
As has been mentioned in several articles circulating around the internet, beacon sensing is available even when you swipe your app away from the multi-tasking view. However in my experiments, a region enter/exit event doesn't call the didDetermineState: directly (Probably because I hadn't been using the AppDelegate to initiate any beacon sensing but instead triggering monitoring based on UI events). Instead if you have registered for Background Location Updates, your AppDelegate's didFinishLaunchingWithOptions: method would get called with the value for key UIApplicationLaunchOptionsLocationKey in the parameter launchOptions set.
You can do a simple check like this to test if this is indeed a location update that has bought your app into the background to perform some task.
if ([launchOptions objectForKey:UIApplicationLaunchOptionsLocationKey])
You can then either register your monitored regions again or start ranging immediately.
P.S. CLLocationManager retains your previously monitoredRegions on app restore but without starting monitoring again using the same UUID and identity, you would not get the enter/exit region event in CLLocationManagerDelegate (which had brought your back up to life)
David has done some wonderful work on this, so I'm writing this cautiously... but I'm seeing something quite different from him in my tests.
I'm using two phones: an iPhone 4S running iOS 7.1 (11D167) and an iPhone 5S running iOS 7.0.6 (11B651). My iBeacons are manufactured and sold by Bluecats (www.bluecats.com), although I'm not yet using their SDK (ie. I'm just using CoreLocation) and I don't think the manufacturer makes much difference.
I'm getting response times of around 1-2 seconds on both devices when the app is running in the foreground and also when running in the background. The only difference is when I remove the app from the app switcher: iOS 7.0.6 never responds (or perhaps will do in 15 minutes), but iOS 7.1 responds in roughly the same time. When I say "respond", I mean that the CLLocationManager's locationManager:didDetermineState:forRegion: delegate is called by iOS.
I'm testing by actually wandering around my office with phones in hand, so I'm physically moving in and out of range. Strangely (?), in my early testing, where I was sitting at my desk and simulating moving in and out of range by removing and reinserting batteries, I was seeing much slower response times. Perhaps this is part of the difference?
In my testing I have seen the presence of a beacon go un-noticed by an app for up to 15 minutes, but I found something that's interesting. I'm using RedBearLabs mini BTLE sensors as ibeacons and their app to program the beacons, http://redbearlab.com/ibeacon/ (http://redbearlab.com/s/MiniBeacon_v1.zip), seems to have an something in it that immediately starts a scan / update of beacons. If I start a beacon up, and in my app it goes unnoticed, by starting then quitting the MiniBeacon app my app immediately notifies me that there are new beacons. This is the same result when entering or exiting. Their app uses CBCentralManager, which my app doesn't, so maybe a mixture between CBCentralManager and CLBeaconRegion is the way to go? I imagine CLBeaconRegion starts / restarts the bluetooth radio, so maybe that is the reason for this. Just taking a stab at it in hopes that someone with a more complete understanding can help resolve this.
Thanks
My testing also reproduces 15 mins to start scanning when my app is in background mode on iOS7.1.1. Just a bit curious, I have seen quite many youtube videos from different companies showing the app has been waken from background mode as soon as they approach their beacons. Is it sales trick?
I am in the process of porting a Flash Player-based game over to the Desktop (OSX and Windows) via Adobe AIR (3.1). The porting to AIR itself has gone rather smoothly. The one wrinkle I've been dealt is that the game will be distributed over the Steam network. In order to interact with the Steam Client, I've had to write a native extension to expose the Steam SDK APIs to AS3. The native extension support has been implemented for both platforms, and I have the application launching and communicating with Steam as desired.
The area I've run into trouble is dealing with Steam's Overlay, which renders overtop of games when it is activated. Essentially, when a game is launched, the Steam Client suspends the process in order to hook its Overlay library up to either D3D or OpenGL. Initially, the Overlay failed to appear at all as the AIR application descriptor had the default rendermode set to "auto." However, once I switched the rendermode to "gpu" the Overlay would appear as desired.
On the OSX side of things, everything works as expected. I can toggle in and out of the Overlay just fine. On the windows end of the spectrum, I've hit a bit of a problem when I activate the Overlay. Specifically, when the Overlay is enabled (it's rendering overtop of the game) and I either move the mouse or generate keyboard input, both the Overlay and the game both "freeze" (rendering stops) for 2-3 seconds. Additionally, I have noticed that when I open the Task Manager with the game running, the cpu usage is roughly 75-80%. The cpu usage remains the same when I first active the Overlay (which is desired). However, when I move the mouse cursor or press a key on the keyboard, the cpu usage drops to roughly 1%. This problem has occurred on 4 of 5 windows machines (2 XP, 3 Win 7) we've tested on. Naturally, I first contacted Valve about the issue since this only occurs when the Overlay is enabled. I've uploaded both the OSX and Windows builds for their devs to debug; however, my contact suggested I find out more about AIRs rendering/input as well.
Here is a snippet of a post with a Steam Dev detailing how the overlay works:
"The requirements for the overlay on Windows are as follows:
Game must use D3D7, D3D8, D3D9, D3D10, D3D11, or OpenGL
Game must call D3D Present() or OpenGL SwapBuffers() on a fast regular basis (these calls are hooked by the overlay and give it opportunity to do work). For instance 2D games that only call these functions when mouse movement occurs or graphics on screen actually change rather than every frame will not function well.
Game should use standard Win32 input messages, raw Win32 input messages, or DirectInput for input and the overlay will then detect hotkeys and hide/block input events from the game when active.
It sounds like your game may violate #2 and stops calling Present/SwapBuffers sometimes when the overlay is active. This may happen if you call these functions in response to user input which is now blocked due to the overlay being activated. You should guarantee you keep pumping frames and swapping at a regular interval even if input events aren't occurring."
After a little more prodding, the Valve devs profiled my application to determine if there was any specific problem occuring with the Game Overlay. Unfortunately, they were unable to find anything going on in the Overlay itself. This pretty much means that AIR on Windows doesn't like that the Overlay is blocking Win32 input messages. Here is the Valve dev's response:
"I got your depots and did some testing. Nothing unusual happens in the overlay. Profiling your app with xpref while the issue occurs and taking some minidumps to check callstacks it looks like the app just blocks up completely and uses zero CPU during the time it is blocked, when it happens it calls Present() only at roughly 1 second intervals until it recovers (maybe there is a 1 second timeout somewhere in the AIR code). It's hard to get much detail since I don't have any symbols for the AIR runtime libraries.
It does however look like this is somehow related to input state and AIR being unhappy with win32 input messages stopping. If I change our overlay to not block any input at all once activated (which obviously has some pretty big problems for usability, but just for testing purposes.) then the issue does not occur. It's possible that the AIR code has some weird logic where if it's seen some specific WM_WHATVER message it's expecting another right after and blocks on it waiting somehow.
Hopefully you can work out on your side or with Adobe as to why the application behaves badly in these situations and starts blocking and not presenting at regular intervals."
I've posted on the Adobe forums, but haven't had any such luck over there. Mainly, I'm hoping that someone has either dealt with this before or has an idea about how I could possibly get around the issue. Any suggestions, comments or thoughts would be greatly appreciated!
As it turns out, there is an bug deep in AIR core framework that is the root cause of this issue. Adobe has confirmed the bug, and they are working on a fix for the Cyril (AIR 3.3) release. The status of the bug (#3089755) can be viewed in the Adobe AIR bug list.
In the short-term, I was forced to detect Windows messages that were being consumed by the SteamOverlay, and pass on fake messages to prevent AIR from locking up. I accomplished this by using the Windows API SetWindowsHookEx along with the WH_DEBUG and WH_GETMESSAGE hooks. This is definitely not a desirable approach, but was needed in the short-term until Adobe releases a fix.
Quick question -- if I code something to respond to "MouseLeftButtonDown" such as the pushing of an image, if I leave the code the same way when I ship the app, will this directly translate to the user pushing their finger down on the same spot, and thus fire the code?
Do I have to change the MouseLeftButtonDown to the gesture listener for this to translate, such that MouseLeftButtonDown is only used in place of non-touch monitors when coding to test things?
THanks!
As corrected by Matt in the comments, the MouseLeftButtonDown event is not the same as an image tap. However, the result would be the same in that if your code works on the emulator, regarding the tapping, then it should work on the device.
You should try and get your app running on a device though as there can be things easily overlooked in the emulator. For example, performance can decrease on your phone since it's likely to be quite a lot less powerful than your PC. Therefore, if your app is performing fine on your PC (emulator), it doesn't necessarily mean you'll get the same speeds on the device.
I'm doing a phone app with some animations and they look really clunky on the emulator. I don't have a phone yet so this is the only way I can test my app.
Sometimes the animations start late (up to a second after user input) and the they are almost always very jagged. Far from the smooth fades and transitions that I've seen on the interwebs. I'm not using anything hairy - just basic rotations and opacity fades on one or two elements.
Does anyone else see this in the emulator? If not, i guess I have a bug somewhere. If so, is there a work around? Should I bump the priority of the xde.exe in process explorer? Other?
Thanks!
This may be a consequence of gpu detection no working on your system.
You can verify this by checking if you can see the frame rate counters.
Jeff Wilcox – Frame rate counters in Windows Phone
Note the emulator system requirements here also.
Setup and System Requirements for Windows Phone Emulator