Uniformly handle 'undefined' store to Float64Array and Float32Array.

Uniformly handle 'undefined' store to Float64Array and Float32Array.

Previous behavior diverged in ICs and Crankshaft. When storing to a Float32Array or Float64Array, the ICs treated undefined as zero while Crankshaft treated it as NaN. Now both ICs and Crankshaft treat it as NaN, which is consistent with the WebGL & ECMAScript spec.

R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=10714

[Michael Starzinger, 2012-02-15 13:39:28]

LGTM (with a few comments). Could you also add a short description to the CL
that we had diverging behavior between non-Crankshaft and Crankshaft code, so
that it is clear that this CL just unifies our internal behavior?

https://chromiumcodereview.appspot.com/9402008/diff/1/src/objects.cc
File src/objects.cc (right):

https://chromiumcodereview.appspot.com/9402008/diff/1/src/objects.cc#newcode1...
src/objects.cc:11770: // Clamp undefined to zero (default). All other types have
been
Comment no longer applies, it should say "clamp undefined to nan".

https://chromiumcodereview.appspot.com/9402008/diff/1/src/objects.cc#newcode1...
src/objects.cc:11790: // Clamp undefined to zero (default). All other types have
been
Likewise.

https://chromiumcodereview.appspot.com/9402008/diff/1/test/mjsunit/external-a...
File test/mjsunit/external-array.js (right):

https://chromiumcodereview.appspot.com/9402008/diff/1/test/mjsunit/external-a...
test/mjsunit/external-array.js:324: for (var i = 0; i < 8; i++) {
Is there a particular reason the assignment has to be done in a loop?

https://chromiumcodereview.appspot.com/9402008/diff/1/test/mjsunit/external-a...
test/mjsunit/external-array.js:339: store_float32_undefined(float32_array);
Is there a particular reason the store function is called twice here?

https://chromiumcodereview.appspot.com/9402008/diff/1/test/mjsunit/external-a...
test/mjsunit/external-array.js:343: for (var i = 0; i < 8; i++) {
Likewise.

https://chromiumcodereview.appspot.com/9402008/diff/1/test/mjsunit/external-a...
test/mjsunit/external-array.js:358: store_float64_undefined(float64_array);
Likewise.

[danno, 2012-02-16 07:55:40]

Feedback addressed, landing.

http://codereview.chromium.org/9402008/diff/1/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/9402008/diff/1/src/objects.cc#newcode11770
src/objects.cc:11770: // Clamp undefined to zero (default). All other types have
been
On 2012/02/15 13:39:29, Michael Starzinger wrote:
> Comment no longer applies, it should say "clamp undefined to nan".

Done.

http://codereview.chromium.org/9402008/diff/1/src/objects.cc#newcode11790
src/objects.cc:11790: // Clamp undefined to zero (default). All other types have
been
On 2012/02/15 13:39:29, Michael Starzinger wrote:
> Likewise.

Done.

http://codereview.chromium.org/9402008/diff/1/test/mjsunit/external-array.js
File test/mjsunit/external-array.js (right):

http://codereview.chromium.org/9402008/diff/1/test/mjsunit/external-array.js#...
test/mjsunit/external-array.js:324: for (var i = 0; i < 8; i++) {
On 2012/02/15 13:39:29, Michael Starzinger wrote:
> Is there a particular reason the assignment has to be done in a loop?

Done.

http://codereview.chromium.org/9402008/diff/1/test/mjsunit/external-array.js#...
test/mjsunit/external-array.js:339: store_float32_undefined(float32_array);
On 2012/02/15 13:39:29, Michael Starzinger wrote:
> Is there a particular reason the store function is called twice here?

Done.

http://codereview.chromium.org/9402008/diff/1/test/mjsunit/external-array.js#...
test/mjsunit/external-array.js:343: for (var i = 0; i < 8; i++) {
On 2012/02/15 13:39:29, Michael Starzinger wrote:
> Likewise.

Done.

http://codereview.chromium.org/9402008/diff/1/test/mjsunit/external-array.js#...
test/mjsunit/external-array.js:358: store_float64_undefined(float64_array);
On 2012/02/15 13:39:29, Michael Starzinger wrote:
> Likewise.

Done.