584293 - Speed up .style on HTML and XUL elements

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect, P1)

Description

[Boris Zbarsky [:bzbarsky]]

As the summary says: make .style faster.  Measurements suggest the upcoming patch makes it 3.5x faster.

The change to nsXULElement::GetStyle overlaps with a similar change in bug 569719; hence the dependency.  Zack should land his stuff first.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: Fix

Two notes:

1)  This doesn't handle SVG elements.  That's a different interface, but would be easy to do if we have a lot of free bits in the typemapping.  Do we?

2)  This is still slower than webkit.  Remaining time spent is 55% under xpc_qsXPCOMObjectToJsval (self time, time in NativeInterface2JSObject, time in FindInJSObjectScope, QI on the decl).  22% is spent in the GetStyle quickstub (a lot of this setting up the ToJsval function call with its many arguments).  5% is spent calling GetStyle on the element.  9% is spent in trace-generated code, and 6% in GetPropertyWithNativeGetter.

Comment 2

[Peter Van der Beken [:peterv]]

Comment on attachment 462694 [details] [diff] [review]
Fix

>+nsIDOMCSSStyleDeclaration*
>+nsStyledElement::GetStyle(nsresult* retval)
> {
>+  nsXULElement* xulElement = nsXULElement::FromContent(this);
>+  if (xulElement) {
>+    nsresult rv = xulElement->EnsureLocalStyle();
>+    if (NS_FAILED(rv)) {

It would be nice if EnsureLocalStyle returned void (all it throws is OUT_OF_MEMORY?). Oh well.

>   nsGenericElement::nsDOMSlots *slots = GetDOMSlots();
>-  NS_ENSURE_TRUE(slots, NS_ERROR_OUT_OF_MEMORY);
>+  if (!slots) {

I don't think this can happen anymore?

>diff --git a/content/html/content/src/nsGenericHTMLElement.cpp b/content/html/content/src/nsGenericHTMLElement.cpp

>+  NS_IMETHOD GetStyle(nsIDOMCSSStyleDeclaration** aStyle) {

Brace on its own line.

>diff --git a/content/xul/content/src/nsXULElement.cpp b/content/xul/content/src/nsXULElement.cpp

>+  NS_IMETHOD GetStyle(nsIDOMCSSStyleDeclaration** aStyle) {

Same here.

The bitmap has 32 bits. Adding nsSVGStylableElement looks fine to me.

Comment 3

[Boris Zbarsky [:bzbarsky]]

> all it throws is OUT_OF_MEMORY?

I think so, but I don't think it's following infallible-only codepaths yet.  Once it is, I'd hope static analysis would win for us.

> I don't think this can happen anymore?

Indeed.  Took that null-check out.

> Brace on its own line.

Done, both places.

I added nsSVGStylableElement.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Attachment: Updated to comments

Comment 5

[Boris Zbarsky [:bzbarsky]]

Pushed http://hg.mozilla.org/mozilla-central/rev/e7b3d37b2bbd

Comment 6

[Boris Zbarsky [:bzbarsky]]

Some combination of this and bug 584287 gave us a 5% win on Dromaeo (DOM) on tinderbox on Linux and Mac.