Use corerct conversions for DataView accessors.

Use corerct conversions for DataView accessors.

We now use DoubleTo(U)Int32 that follows ES specification.

R=titzer@chromium.org,rossberg@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=15659

[Dmitry Lomov (no reviews), 2013-07-12 12:50:05]

PTAL

[titzer, 2013-07-12 16:37:01]

https://codereview.chromium.org/18703007/diff/1/src/runtime.cc
File src/runtime.cc (right):

https://codereview.chromium.org/18703007/diff/1/src/runtime.cc#newcode1221
src/runtime.cc:1221: static T DataViewValueConverter(double value);
Can we name this function verb-ly, such as "ConvertDoubleToDataView"?

https://codereview.chromium.org/18703007/diff/1/test/mjsunit/harmony/dataview...
File test/mjsunit/harmony/dataview-accessors.js (right):

https://codereview.chromium.org/18703007/diff/1/test/mjsunit/harmony/dataview...
test/mjsunit/harmony/dataview-accessors.js:302: function
CheckOutOfRange(typename, value, expected) {
Mmmm...jsiness. Could we have a version of this test that doesn't use computed
properties? I know its more verbose, but can result in different hydrogen code.

[Dmitry Lomov (no reviews), 2013-07-12 18:22:45]

PTAL

https://codereview.chromium.org/18703007/diff/1/src/runtime.cc
File src/runtime.cc (right):

https://codereview.chromium.org/18703007/diff/1/src/runtime.cc#newcode1221
src/runtime.cc:1221: static T DataViewValueConverter(double value);
Done.
ConvertValueToDataView is not a good name - we convert value to T, not to
DataView.
I called it DataViewConvertValue (similar to DataViewGetValue and
DataViewSetValue)
On 2013/07/12 16:37:02, titzer wrote:
> Can we name this function verb-ly, such as "ConvertDoubleToDataView"?

https://codereview.chromium.org/18703007/diff/1/test/mjsunit/harmony/dataview...
File test/mjsunit/harmony/dataview-accessors.js (right):

https://codereview.chromium.org/18703007/diff/1/test/mjsunit/harmony/dataview...
test/mjsunit/harmony/dataview-accessors.js:302: function
CheckOutOfRange(typename, value, expected) {
On 2013/07/12 16:37:02, titzer wrote:
> Mmmm...jsiness. Could we have a version of this test that doesn't use computed
> properties? I know its more verbose, but can result in different hydrogen
code.

Done.

[titzer, 2013-07-15 09:09:18]

lgtm

[Dmitry Lomov (no reviews), 2013-07-15 09:32:22]

Committed patchset #2 manually as r15659 (presubmit successful).