X64: Use roundsd for DoMathFloor.

X64: Use roundsd for DoMathFloor.


TEST=mjsunit/math-floor

Committed: http://code.google.com/p/v8/source/detail?r=7613

[Vitaly Repeshko, 2011-04-13 21:11:50]

Drive-by question.

http://codereview.chromium.org/6835021/diff/1/test/mjsunit/math-floor.js
File test/mjsunit/math-floor.js (right):

http://codereview.chromium.org/6835021/diff/1/test/mjsunit/math-floor.js#newc...
test/mjsunit/math-floor.js:34: assertEquals(expect, test(input));
Don't we need (at least) two calls to test to ensure the IC is initialized?

[William Hesse, 2011-04-14 08:13:03]

LGTM, unless you agree that there is a flaw in DeepEquals for arrays containing
named properties.

http://codereview.chromium.org/6835021/diff/1/test/mjsunit/math-floor.js
File test/mjsunit/math-floor.js (right):

http://codereview.chromium.org/6835021/diff/1/test/mjsunit/math-floor.js#newc...
test/mjsunit/math-floor.js:36: assertEquals(expect, test(input));
What about the fact that -0 and +0 compare equal?  Don't you need to compare 1/x
and 1/y?

http://codereview.chromium.org/6835021/diff/1/test/mjsunit/mjsunit.js
File test/mjsunit/mjsunit.js (right):

http://codereview.chromium.org/6835021/diff/1/test/mjsunit/mjsunit.js#newcode66
test/mjsunit/mjsunit.js:66: var clazz = classOf(value);
object_class?  Better than a misspelling.

http://codereview.chromium.org/6835021/diff/1/test/mjsunit/mjsunit.js#newcode152
test/mjsunit/mjsunit.js:152: return false;
What about if a is an Object and b is an Array?
{0:"foo", 1:"bar} and ["foo", "bar"] - OK, I see length would have to be a
non-enumerable property on the object for it to happen.  Could this happen?

What about arrays with named properties - the named properties are not checked.