Convert some for-in loops for performance

Convert some for-in loops for performance

R=scheglov@google.com

Committed: https://github.com/dart-lang/sdk/commit/61818c95a133bf6cf3a58095d8a78b6fca64e370

[Brian Wilkerson, 2016-05-09 17:48:10]

brianwilkerson@google.com changed reviewers:
+ scheglov@google.com

[Brian Wilkerson, 2016-05-09 17:48:10]

For-in loops cause a lot of the megamorphic dispatch, and at least one of these
methods was showing up as significant in observatory.

[scheglov, 2016-05-09 17:59:29]

LGTM

[Brian Wilkerson, 2016-05-09 18:01:19]

Description was changed from

==========
Convert some for-in loops for performance
==========

to

==========
Convert some for-in loops for performance

R=scheglov@google.com

Committed:
https://github.com/dart-lang/sdk/commit/61818c95a133bf6cf3a58095d8a78b6fca64e370
==========

[Brian Wilkerson, 2016-05-09 18:01:20]

Committed patchset #1 (id:1) manually as
61818c95a133bf6cf3a58095d8a78b6fca64e370 (presubmit successful).

[Bob Nystrom, 2016-05-09 18:15:23]

rnystrom@google.com changed reviewers:
+ rnystrom@google.com

[Bob Nystrom, 2016-05-09 18:15:23]

DBC: When doing micro-optimization like this that makes code less idiomatic, I
think it's really helpful to leave a comment saying:

// Using a manual for loop instead of a for-in loop here because it makes
benchmark ___ X% faster.

Otherwise, someone will come along later and say, "Oh, I can clean this up to be
more idiomatic." and not realize or be able to measure that they've regressed
perf.

(Of course, there is also the larger question of why the most natural,
beautiful, idiomatic way to implement something is slower, but that's more a
question for our implementations.)