• Open Menu Close Menu
  • Apple
  • Shopping Bag
  • Apple
  • Mac
  • iPad
  • iPhone
  • Watch
  • TV
  • Music
  • Support
  • Search apple.com
  • Shopping Bag

Lists

Open Menu Close Menu
  • Terms and Conditions
  • Lists hosted on this site
  • Email the Postmaster
  • Tips for posting to public mailing lists
Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode?
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode?


  • Subject: Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode?
  • From: Troy Stephens <email@hidden>
  • Date: Thu, 24 Jan 2008 13:10:01 -0800

On Jan 24, 2008, at 11:14 AM, Colin Cornaby wrote:
Thanks,

This appeared to be the issue. Removing my bitmap image rep's release seems to have fixed both the stepper issue and the state restore issue.

Would the best time to release my image rep be after I have changed the contents of the layer then? Is there a chance that CoreAnimation could still try to read my old CGImage as part of an animation after I change the contents of my layer?

Yes. Even after you've removed the reference to the CGImage from your layer tree, it may still be referenced by the render tree until the render tree catches up to the change. This makes assessing the necessary lifetime of the NSBitmapImageRep more difficult.


For a truly robust solution when images are coming and going like this, you want the CGImage you hand off to Core Animation to own (retain) its own source pixel data instead of depending on the existence of an NSBitmapImageRep to keep that data around.

You should be able to do this, while keeping performance impact to a minimum, by using CGBitmapContextCreate() with a NULL "data" pointer to create a bitmapped CGContext that manages its own memory, drawing the CIImage into that context (evaluation of the filter chain to produce the output pixels happens here), and then using CGBitmapContextCreateImage() to get a CGImage. The resultant CGImage will then own the pixel data that it references for its full lifetime, and you won't have to worry about keeping a supporting NSBitmapImageRep around. It's a bit more code to write, but should leave you in a much safer place given the image swapping you're doing.

There -might- also be faster ways of rendering a continually updated CIImage in a layer tree. I'm not an expert on CI+CA performance issues, but using a CIContext to render the CIImage directly into a CAOpenGLLayer comes to mind as one possibility that might be worth investigating. I'll see if I can find out more about that.

Troy

On Jan 24, 2008, at 10:52 AM, Troy Stephens wrote:

Hi Colin,

I suspect the problem you're encountering is that the CGImage yielded by a given cycle is sometimes outliving the NSBitmapImageRep it was created from, and whose pixel data the CGImage still references. When CoreGraphics goes to render the CGImage to a destination, it ends up walking off into memory that's been potentially reused for some other purpose.

When using -[NSBitmapImageRep CGImage], it's vitally important to keep the NSBItmapImageRep around for as long as the CGImage is in use. If you fix that problem, your crashes should disappear.

Troy

On Jan 24, 2008, at 9:52 AM, Colin Cornaby wrote:

I have a image I am doing some processing on using CoreImage, which is then set as the contents of a CoreAnimation layer. I know that CALayer has a filters property, which I could use to apply a CoreImage filter in realtime, but for various reasons the image has to be preprocessed. If anyone is interested in why I would be willing to discuss off list.

I have a stepper that adjusts the attributes of the filter. When the stepper is hit, the filter's attributes are changed, my image is re-rendered, and then set again as the contents of my CALayer. I started to notice an interesting crash the other day. Holding the stepper down and rapidly adjusting the image was causing my app to crash. Clicking the stepper rapidly would also produce the same results. But if I hit the stepper more slowly everything seemed to work fine. Manually adjusting the number that the stepper controlled also worked just fine.

I went through my code to try and determine if it was an issue with my rendering code. I went through many steps, including logging images out to files to see if perhaps my rendering code was somehow screwing up and generating bad images. But my software seemed to be generating images just fine.

The issue seems to be that if I try to set the contents property of a CALayer in quick succession it crashes. It's still possible that something in my code is wrong, but I spent all last night debugging it and signs seem to point to an issue within CoreAnimation.

Here is my post image rendering code that pushes the image into CoreAnimation:
CIImage *finalImage = [filter valueForKey:@"outputImage"];
NSBitmapImageRep *finalImageRep = [[NSBitmapImageRep alloc] initWithCIImage:finalImage];
CGImageRef finalCgImage = [finalImageRep CGImage];
[CATransaction begin];
[CATransaction setValue:(id)kCFBooleanTrue
forKey:kCATransactionDisableActions];
[self.contentLayer setContents: (id) finalCgImage];
[CATransaction commit];


And here is the backtrace I get when I set the image too rapidly:
#0 0x96492f10 in sseCGSConvertXXXX8888TransposePermute
#1 0x9640f7f6 in argb32_image
#2 0x932e9822 in ripl_Mark
#3 0x932ebcd1 in ripl_BltImage
#4 0x932d5c25 in ripc_RenderImage
#5 0x932e631c in ripc_DrawImage
#6 0x9640343d in CGContextDrawImage
#7 0x90cb544f in create_image_by_rendering
#8 0x90cb4f48 in CARenderImageNewWithCGImage
#9 0x90cdcc78 in CARenderPrepareCGImage
#10 0x90cdcbd3 in CALayerPrepareCommit
#11 0x90cdcbd3 in CALayerPrepareCommit
#12 0x90cdcbd3 in CALayerPrepareCommit
#13 0x90cdcbd3 in CALayerPrepareCommit
#14 0x90cb112f in CAContextCommitTransaction
#15 0x90cb0e27 in CATransactionCommit
#16 0x9017a10f in -[NSCell trackMouse:inRect:ofView:untilMouseUp:]
#17 0x9060f4f7 in -[NSStepperCell trackMouse:inRect:ofView:untilMouseUp:]
#18 0x901794cc in -[NSControl mouseDown:]
#19 0x90177d9b in -[NSWindow sendEvent:]
#20 0x90144a2c in -[NSApplication sendEvent:]
#21 0x900a2705 in -[NSApplication run]
#22 0x9006f9ba in NSApplicationMain
#23 0x00007136 in main at main.m:13


If this looks like I'm passing a bad image to CA I'll keep plugging away at my image code and try to log what's going on deeper in the process.

I also have this same crash occurring in another situation that is slightly more complicated to explain, but involves when I am loading back up a document and restoring the state. Again, debugging shows that the image I'm passing CA is good, but CA will either crash, or if I'm lucky, display garbage. Given that the stepper seems to shine more light on this issue, I'm currently trying to fix this bug in relation to the stepper instead of looking at my state restorer right now.

If anyone has seen this sort of crash before, could they give me some sort of clue as to what it means?

Thanks,
Colin
_______________________________________________

Cocoa-dev mailing list (email@hidden)

Please do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com

Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden



_______________________________________________

Cocoa-dev mailing list (email@hidden)

Please do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com

Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden


  • Follow-Ups:
    • Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode?
      • From: Colin Cornaby <email@hidden>
References: 
 >Changing CALayer's contents too rapidly causes CoreAnimation to implode? (From: Colin Cornaby <email@hidden>)
 >Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode? (From: Troy Stephens <email@hidden>)
 >Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode? (From: Colin Cornaby <email@hidden>)

  • Prev by Date: Re: NSTimer preventing sleep?
  • Next by Date: Re: String encoding problem (WSMethodInvocation)
  • Previous by thread: Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode?
  • Next by thread: Re: Changing CALayer's contents too rapidly causes CoreAnimation to implode?
  • Index(es):
    • Date
    • Thread