Bug 129851 – Add fast path for id/name/style/class in getAttribute/setAttribute bindings.

• Component: WebCore JavaScript
• Status: RESOLVED INVALID

[Andreas Kling]

Add fast path for id/name/style/class in getAttribute/setAttribute bindings.

[Andreas Kling]

Created attachment 226068 [details]
Patch

[Ryosuke Niwa]

Can we make sure we haven't regressed other attar accessed significantly with this?

[Dongseong Hwang]

Hi, I'm cherry-picking this to Blink, https://codereview.chromium.org/189483004/
In the bug, I measured performance. Both were run on my Ivy Bridge laptop. 

PerformanceTests/Bindings/get-attribute.html -> 16% improvement 
Before : Avg get-attribute: 273.565932runs/s 
After : Avg get-attribute: 317.818405runs/s 

PerformanceTests/Bindings/set-attribute.html -> 77% improvement 
Before : Avg set-attribute: 476.101349runs/s 
After : Avg set-attribute: 847.042662runs/s

[Dongseong Hwang]

Comment on attachment 226068 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=226068&action=review

> Source/WebCore/dom/Element.cpp:954
> +void Element::bindingsSetAttribute(const String& localName, const String& newValue, ExceptionCode& ec)

Could I ask why you used String& instead of AtomicString&?

[Dongseong Hwang]

Blink gave up cherry-picking this patch because it causes regression of other attributes significantly (about 50%).

getAttribute("id") -> 11% improvement 
Before : Avg get-attribute: 329.751776runs/s 
After : Avg get-attribute: 367.252965runs/s 

getAttribute("group") -> 51% regression 
Before : Avg get-attribute-href: 329.339992runs/s 
After : Avg get-attribute-href: 159.540438runs/s 

setAttribute("id")  -> 19% improvement 
Before : Avg set-attribute: 809.784217runs/s 
After : Avg set-attribute: 965.404519runs/s 

setAttribute("group") -> 45% regression 
Avg set-attribute-href: 756.714409runs/s 
Avg set-attribute-href: 415.186785runs/s

[Chris Dumez]

(In reply to comment #5)
> Blink gave up cherry-picking this patch because it causes regression of other attributes significantly (about 50%).
> 
> getAttribute("id") -> 11% improvement 
> Before : Avg get-attribute: 329.751776runs/s 
> After : Avg get-attribute: 367.252965runs/s 
> 
> getAttribute("group") -> 51% regression 
> Before : Avg get-attribute-href: 329.339992runs/s 
> After : Avg get-attribute-href: 159.540438runs/s 
> 
> setAttribute("id")  -> 19% improvement 
> Before : Avg set-attribute: 809.784217runs/s 
> After : Avg set-attribute: 965.404519runs/s 
> 
> setAttribute("group") -> 45% regression 
> Avg set-attribute-href: 756.714409runs/s 
> Avg set-attribute-href: 415.186785runs/s

Dongseong, are those the results with Blink or the results with WebKit? 
Results may be different.

[Ryosuke Niwa]

Note that we didn't land this patch.

[Andreas Kling]

Oh this. We don't need this now that we construct perfect strings in JSC.