Ensure newly allocated empty Arrays are transitioned to FAST_ELEMENT

Ensure newly allocated empty Arrays are transitioned to FAST_ELEMENT

BUG=
TEST=

[danno, 2011-12-29 15:08:01]

PTAL.

[Jakob Kummerow, 2012-01-04 20:55:16]

LGTM with comments.

Dynamic decision of the default elements kind is nifty, I like it :-)

Now that it's January, please change the year to 2012 in all files in this CL.

http://codereview.chromium.org/9050001/diff/6002/src/bootstrapper.cc
File src/bootstrapper.cc (right):

http://codereview.chromium.org/9050001/diff/6002/src/bootstrapper.cc#newcode2327
src/bootstrapper.cc:2327: context->set_fast_array_element_bias(0);
I have a weak preference to use "Smi::FromInt(0)" here, just to make it clear
what the "0" is.

http://codereview.chromium.org/9050001/diff/6002/src/flag-definitions.h
File src/flag-definitions.h (right):

http://codereview.chromium.org/9050001/diff/6002/src/flag-definitions.h#newco...
src/flag-definitions.h:123: DEFINE_bool(smi_only_arrays, true, "tracks arrays
with only smi values")
Is this change intentional?

http://codereview.chromium.org/9050001/diff/6002/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/9050001/diff/6002/src/objects.cc#newcode1224
src/objects.cc:1224: if (global_context->untransitioned_js_array_map() != NULL
&&
Don't you mean s/NULL/GetHeap()->undefined_value()/? (cf. changes to heap.cc)

http://codereview.chromium.org/9050001/diff/6002/src/objects.cc#newcode1245
src/objects.cc:1245: &dummy);
nit: indentation

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind-f...
File test/mjsunit/elements-kind-fast-element-bias.js (right):

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind-f...
test/mjsunit/elements-kind-fast-element-bias.js:38: // Ensure that there is a
global bais for not aggressively transitioning Arrays
s/bais/bias/

Even better, re-word the comment. Maybe "Test that repeated SMI->FAST
transitions are avoided by initializing Arrays with FAST_ELEMENTS right away"?
And then I'd add the inverse of the check at the end before messing with the
bias:

if (support_smi_only_arrays) {
  assertTrue(%HasFastSmiOnlyElements(new Array()));
}

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind-s...
File test/mjsunit/elements-kind-smi-only-bias.js (right):

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind-s...
test/mjsunit/elements-kind-smi-only-bias.js:38: // Ensure that there is a global
bais for not aggressively transitioning Arrays
s/bais/bias/

Or re-word: "Test that Arrays are initialized with SMI_ONLY elements when the
code benefits from transitioning to FAST_DOUBLE elements".

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind.js
File test/mjsunit/elements-kind.js (left):

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind.j...
test/mjsunit/elements-kind.js:42: print("Tests do NOT include smi-only
arrays.");
intentional removal?

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind.js
File test/mjsunit/elements-kind.js (right):

http://codereview.chromium.org/9050001/diff/6002/test/mjsunit/elements-kind.j...
test/mjsunit/elements-kind.js:45: // Ensure that there is a global bais for not
aggressively transitioning Arrays
suggestion: "Ensure that Arrays default to SMI_ONLY elements."