Faster moving FixedArray elements around.

Faster moving FixedArray elements around.

Committed: http://code.google.com/p/v8/source/detail?r=3987

[antonm, 2010-03-01 07:49:38]

Mads, Søren,

may you have a look?

It's actually rework of http://codereview.chromium.org/650043/show and mostly
differs by applying the same trick to other Array's methods.

[Søren Thygesen Gjesse, 2010-03-01 10:07:38]

LGTM

http://codereview.chromium.org/660245/diff/15/17
File src/builtins.cc (right):

http://codereview.chromium.org/660245/diff/15/17#newcode276
src/builtins.cc:276: Heap::RecordWrites(dst->address(),
How about adding a static function to FixedArray calculating

  FixedArray::kHeaderSize + dst_index * kPointerSize

as it is used a couple of times here (like the data_start() you already added).
It is actually what FixedArray::SizeFor() does, but I there should be one with a
different name.

http://codereview.chromium.org/660245/diff/15/19
File src/heap.h (right):

http://codereview.chromium.org/660245/diff/15/19#newcode773
src/heap.h:773: // Write barrier support for address[start : start + len] = o.
Is start + len included? ([start : start + len] -> [start : start + len[)

http://codereview.chromium.org/660245/diff/15/29
File test/mjsunit/fuzz-natives.js (right):

http://codereview.chromium.org/660245/diff/15/29#newcode152
test/mjsunit/fuzz-natives.js:152: // That can only be invoked on
Array.prototype.
That -> This

Why can't this survive fuzzing?

[antonm, 2010-03-01 14:49:28]

Søren,

thanks a lot for review and sorry for late response.  I'm going to submit after
rerunning the test (which would take some time).

If you have any concerns (esp. for fuzz-natives.js), chances are I'll address
them before submitting.  Otherwise that would require another CL.

http://codereview.chromium.org/660245/diff/15/17
File src/builtins.cc (right):

http://codereview.chromium.org/660245/diff/15/17#newcode276
src/builtins.cc:276: Heap::RecordWrites(dst->address(),
On 2010/03/01 10:07:38, Søren Gjesse wrote:
> How about adding a static function to FixedArray calculating
> 
>   FixedArray::kHeaderSize + dst_index * kPointerSize
> 
> as it is used a couple of times here (like the data_start() you already
added).
> It is actually what FixedArray::SizeFor() does, but I there should be one with
a
> different name.

Sure.  And I am lucky today: there is already FixedArray::OffsetOfElementAt :)

http://codereview.chromium.org/660245/diff/15/19
File src/heap.h (right):

http://codereview.chromium.org/660245/diff/15/19#newcode773
src/heap.h:773: // Write barrier support for address[start : start + len] = o.
On 2010/03/01 10:07:38, Søren Gjesse wrote:
> Is start + len included? ([start : start + len] -> [start : start + len[)

I used pythonic notation, but switched to [start : start + len[.

http://codereview.chromium.org/660245/diff/15/29
File test/mjsunit/fuzz-natives.js (right):

http://codereview.chromium.org/660245/diff/15/29#newcode152
test/mjsunit/fuzz-natives.js:152: // That can only be invoked on
Array.prototype.
On 2010/03/01 10:07:38, Søren Gjesse wrote:
> That -> This
> 
> Why can't this survive fuzzing?

If I understand the idea of the test correctly, it attempts to call various
native functions passing in different types of parameters. 
FinishArrayPrototypeSetup expects to get only Array.prototype.  I could easily
extend it to be more acceptable if it's preferred approach.