Migrate to new indexed getter marking scheme.

Migrate to new indexed getter marking scheme.

TBR=blois@google.com, vsm@google.com

Committed: https://code.google.com/p/dart/source/detail?r=22995

[Anton Muhin, 2013-05-22 08:53:03]

Committed patchset #1 manually as r22995 (presubmit successful).

[blois, 2013-05-22 18:50:11]

This causes a lot of our array accessors to change from foo[index] to
foo.item(index)- are there better metrics we can use to keep generating
foo[index]?

(see
https://codereview.chromium.org/15697002/diff/1/sdk/lib/html/dart2js/html_dar...
for the generated code diff)

I just tried a quick jsperf test, and item(index) is about 20% slower on Chrome
and 67% slower on FF.

[blois, 2013-05-22 18:50:50]

On 2013/05/22 18:50:11, blois wrote:
> This causes a lot of our array accessors to change from foo[index] to
> foo.item(index)- are there better metrics we can use to keep generating
> foo[index]?
> 
> (see
>
https://codereview.chromium.org/15697002/diff/1/sdk/lib/html/dart2js/html_dar...
> for the generated code diff)
> 
> I just tried a quick jsperf test, and item(index) is about 20% slower on
Chrome
> and 67% slower on FF.

jsperf test:
http://jsperf.com/index-vs-item

[Anton Muhin, 2013-05-22 19:00:53]

That's actually vice versa unless I miss something.

Due to changes in IDL attribues, lig regen
(https://codereview.chromium.org/15697002/) switched a lot of places from
indexed access to .item, but even performance issues aside that just didn't
work.

This CL undos harm previously done.

[blois, 2013-05-22 19:26:58]

lgtm

Ah, now I see. Duh!

[vsm, 2013-05-22 23:23:11]

lgtm, but should we generate item at all (with apparently different bounds
checks behavior)?