Bug 173087 – Web Inspector: Instrument active pixel memory used by canvases

• Component: Web Inspector
• Status: RESOLVED FIXED

[Matt Baker]

Summary:
Instrument active pixel memory used by canvases.

1) Add property `Canvas.memoryCost` to the Canvas protocol. Get canvas pixel memory using:

size_t HTMLCanvasElement::memoryCost() const;
size_t HTMLCanvasElement::externalMemoryCost() const;

2) Add event `canvasMemoryChanged` to the Canvas protocol. Instrumentation points:

void HTMLCanvasElement::createImageBuffer() const;
void HTMLCanvasElement::setImageBuffer(std::unique_ptr<ImageBuffer>) const;

3) It might be useful to display the maximum pixel memory available to all canvases. This info can be retrieved by hooking into `maxActivePixelMemory()` (see HTMLCanvasElement.cpp:190).

[Radar WebKit Bug Importer]

<rdar://problem/32719261>

[Devin Rousso]

Created attachment 312930 [details]
[Patch] WIP

[Devin Rousso]

Created attachment 312931 [details]
[Image] After Patch is applied

[Nikita Vasilyev]

(In reply to Devin Rousso from comment #3)
> Created attachment 312931 [details]
> [Image] After Patch is applied

    Memory: 4.47 MB / 4 GB

This wasn't clear to me that the second number is "maximum pixel memory available to all canvases".

What happens if canvas pixel memory is greater than this number?

[Devin Rousso]

Created attachment 313466 [details]
Patch

[Build Bot]

Comment on attachment 313466 [details]
Patch

Attachment 313466 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.webkit.org/results/3969320

New failing tests:
inspector/canvas/memory.html

[Build Bot]

Created attachment 313476 [details]
Archive of layout-test-results from ews107 for mac-elcapitan-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: ews107  Port: mac-elcapitan-wk2  Platform: Mac OS X 10.11.6

[Build Bot]

Comment on attachment 313466 [details]
Patch

Attachment 313466 [details] did not pass mac-debug-ews (mac):
Output: http://webkit-queues.webkit.org/results/3969301

Number of test failures exceeded the failure limit.

[Build Bot]

Created attachment 313478 [details]
Archive of layout-test-results from ews117 for mac-elcapitan

The attached test failures were seen while running run-webkit-tests on the mac-debug-ews.
Bot: ews117  Port: mac-elcapitan  Platform: Mac OS X 10.11.6

[Build Bot]

Comment on attachment 313466 [details]
Patch

Attachment 313466 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.webkit.org/results/3969595

New failing tests:
inspector/canvas/memory.html

[Build Bot]

Created attachment 313482 [details]
Archive of layout-test-results from ews100 for mac-elcapitan

The attached test failures were seen while running run-webkit-tests on the mac-ews.
Bot: ews100  Port: mac-elcapitan  Platform: Mac OS X 10.11.6

[Devin Rousso]

Created attachment 313486 [details]
Patch

[Devin Rousso]

Created attachment 313488 [details]
Patch

[Joseph Pecoraro]

(In reply to Nikita Vasilyev from comment #4)
> (In reply to Devin Rousso from comment #3)
> > Created attachment 312931 [details]
> > [Image] After Patch is applied
> 
>     Memory: 4.47 MB / 4 GB
> 
> This wasn't clear to me that the second number is "maximum pixel memory
> available to all canvases".

I agree. My initial thought was that the canvas was currently 4.5MB and can grow to 4GB which was not correct. How can we make this clearer? Maybe:

    Memory: 4.47MB (4 GB Limit)

> What happens if canvas pixel memory is greater than this number?

That shouldn't be allowed to happen.

[Joseph Pecoraro]

Comment on attachment 313488 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=313488&action=review

r- to see another patch after dropping the "max" thing. The rest looks good.

> Source/JavaScriptCore/inspector/protocol/Canvas.json:42
> +                { "name": "memoryCost", "type": "number", "optional": true, "description": "Memory usage of the canvas." }

This is great!

> Source/JavaScriptCore/inspector/protocol/Canvas.json:80
> +        {
> +            "name": "maxPixelMemory",
> +            "description": "Gets the maximum pixel memory available for all canvases.",
> +            "returns": [
> +                { "name": "maxPixelMemory", "type": "number", "description": "Maximum pixel memory available for all canvases." }
> +            ]

Lets drop this. I don't think this max is useful. If/when the page crosses this max pixel memory a console message will be logged that lists what the maximum size is.

> Source/JavaScriptCore/inspector/protocol/Canvas.json:99
> +                { "name": "canvasId", "$ref": "CanvasId", "description": "Identifier of canvas who's memory cost changed." },

Simplify the description.

> Source/WebCore/inspector/InspectorInstrumentation.h:1200
> +    if (InstrumentingAgents* instrumentingAgents = instrumentingAgentsForDocument(&canvasElement.document()))

This should `FAST_RETURN_IF_NO_FRONTENDS(void());`

[Devin Rousso]

Created attachment 313996 [details]
Patch

[Joseph Pecoraro]

Comment on attachment 313996 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=313996&action=review

r=me, nice tests

> LayoutTests/inspector/canvas/memory.html:55
> +            InspectorTest.log(`Change size of canvas to ${width}x${height}.`);

We typically use this "\xd7", not that its important here though.

> Source/JavaScriptCore/inspector/protocol/Canvas.json:42
> +                { "name": "memoryCost", "type": "number", "optional": true, "description": "Memory usage of the canvas." }

Lets specify "in bytes" somewhere.

> Source/WebInspectorUI/.eslintrc:37
> +        "CanvasAgent": true,

Pro.

> Source/WebInspectorUI/UserInterface/Controllers/CanvasManager.js:110
> +    CanvasMemoryChanged: "canvas-manager-canvas-memory-changed",

Its weird to me that this event is on the CanvasManager and not the Canvas instance that is changing.

If it was on the Canvas object then listeners could either listen on a particular instance:

    instance.addEventListener(WebInspector.Canvas.Event.MemoryChanged, () => {});

Or for all canvases:

    WebInspector.Canvas.addEventListener(WebInspector.Canvas.Event.MemoryChanged, () => {});

> Source/WebInspectorUI/UserInterface/Models/Canvas.js:96
> +    set memoryCost(memoryCost) { this._memoryCost = memoryCost; }

This is where I would have expected the event to be dispatched.

[Devin Rousso]

Created attachment 314063 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 314063 [details]
Patch

Clearing flags on attachment: 314063

Committed r218908: <http://trac.webkit.org/changeset/218908>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Michael Catanzaro]

Hey Devin, this new test is failing on GTK:

--- /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/inspector/canvas/memory-expected.txt
+++ /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/inspector/canvas/memory-actual.txt
@@ -7,5 +7,5 @@
 
 -- Running test case: Canvas.memory.canvasMemoryChanged
 Change size of canvas to 200x200.
-Memory cost of canvas updated to 166400.
+Memory cost of canvas updated to 160000.

Is that harmless, or should it exactly equal 160000 on our port?

[Devin Rousso]

(In reply to Michael Catanzaro from comment #21)
> Hey Devin, this new test is failing on GTK:
> 
> ---
> /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/
> inspector/canvas/memory-expected.txt
> +++
> /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/
> inspector/canvas/memory-actual.txt
> @@ -7,5 +7,5 @@
>  
>  -- Running test case: Canvas.memory.canvasMemoryChanged
>  Change size of canvas to 200x200.
> -Memory cost of canvas updated to 166400.
> +Memory cost of canvas updated to 160000.
> 
> Is that harmless, or should it exactly equal 160000 on our port?

As far as I understand, the memory cost is just (width * height * 4).  I'm not actually sure why it shows as 166400 :|  I think 160000 is fine.