Decrease the binary size by 50 KB by outlining V8XXX::wrap()

Decrease the binary size by 50 KB by outlining V8XXX::wrap()

wrap() is called only when we fail to get a wrapper from DOMDataStore. In other words, wrap() is called only when a wrapper is created. This is a slow path, so we don't need to inline the function.

This reduces the binary size of Linux Release from 135279744 byte to 135230832 byte (saving 50 KB).

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=176098

[haraken, 2014-06-13 06:21:47]

PTAL

[dcarney, 2014-06-13 06:25:24]

lgtm

[haraken, 2014-06-13 06:28:44]

The CQ bit was checked by haraken@chromium.org

[commit-bot: I haz the power, 2014-06-13 06:29:18]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/haraken@chromium.org/331593006/1

[Nils Barth (inactive), 2014-06-13 06:36:20]

This is great size-wise, but what's the performance impact?
(I trust your intuition, but want a few more details.)

I.e., even though wrap is only called 1 time (it's not hot),
maybe this has a serious impact on startup/time to first paint (b/c it's on the
critical path)?

What if you open up a web page with lots of different DOM types?
(I'd assume there'd be regressions from the function call overhead, but
presumably v. small.)

[haraken, 2014-06-13 06:44:03]

On 2014/06/13 06:36:20, Nils Barth wrote:
> This is great size-wise, but what's the performance impact?
> (I trust your intuition, but want a few more details.)
> 
> I.e., even though wrap is only called 1 time (it's not hot),
> maybe this has a serious impact on startup/time to first paint (b/c it's on
the
> critical path)?
> 
> What if you open up a web page with lots of different DOM types?
> (I'd assume there'd be regressions from the function call overhead, but
> presumably v. small.)

I don't see any regression in Dromaeo/dom-modify, which creates a lot of
wrappers. The point is that wrap() calls createWrapper(), which does very heavy
operations. Thus the function-call overhead of wrap() won't matter.

However, abarth@ pointed out that the size measurement is not correct (I forget
to strip symbols). I'm remeasuring...

[haraken, 2014-06-13 06:44:10]

The CQ bit was unchecked by haraken@chromium.org

[Nils Barth (inactive), 2014-06-13 06:53:50]

On 2014/06/13 06:44:03, haraken wrote:
> I don't see any regression in Dromaeo/dom-modify, which creates a lot of
> wrappers.

Ok, that sounds good!

> The point is that wrap() calls createWrapper(), which does very heavy
> operations. Thus the function-call overhead of wrap() won't matter.

Got it: the function call overhead is negligible, since it calls an expensive
function.

> However, abarth@ pointed out that the size measurement is not correct (I
forget
> to strip symbols). I'm remeasuring...

Measure twice, cut once ;)

[haraken, 2014-06-13 06:56:59]

On 2014/06/13 06:53:50, Nils Barth wrote:
> On 2014/06/13 06:44:03, haraken wrote:
> > I don't see any regression in Dromaeo/dom-modify, which creates a lot of
> > wrappers.
> 
> Ok, that sounds good!
> 
> > The point is that wrap() calls createWrapper(), which does very heavy
> > operations. Thus the function-call overhead of wrap() won't matter.
> 
> Got it: the function call overhead is negligible, since it calls an expensive
> function.
> 
> > However, abarth@ pointed out that the size measurement is not correct (I
> forget
> > to strip symbols). I'm remeasuring...
> 
> Measure twice, cut once ;)

Sorry, I found that a correct size gain is 50 KB :) Updated the issue title.

[haraken, 2014-06-13 06:57:03]

The CQ bit was checked by haraken@chromium.org

[commit-bot: I haz the power, 2014-06-13 06:57:23]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/haraken@chromium.org/331593006/1

[commit-bot: I haz the power, 2014-06-13 08:40:12]

Change committed as 176098