Harmony: implement Math.clz32

Harmony: implement Math.clz32

R=dslomov@chromium.org, svenpanne@chromium.org
BUG=v8:2938
LOG=N

Committed: https://code.google.com/p/v8/source/detail?r=19435

[Dmitry Lomov (no reviews), 2014-02-17 16:01:40]

lgtm

[Sven Panne, 2014-02-18 07:21:24]

LGTM as a first step, too, but I think this should have Crankshaft support.

https://codereview.chromium.org/169783002/diff/130001/src/harmony-math.js
File src/harmony-math.js (right):

https://codereview.chromium.org/169783002/diff/130001/src/harmony-math.js#new...
src/harmony-math.js:157: function MathClz32(x) {
What TC39 seems to intend is basically exposing some form of intrinsics for
processor intstructions. AFAICT "count leading zeroes" has HW support for all
our platforms and boils down to a single CPU instructions, and I think this is
the expectation of people using such intrinsics. Therefore we should add
Crankshaft support, which shouldn't be too hard.

[Yang, 2014-02-18 07:27:21]

On 2014/02/18 07:21:24, Sven Panne wrote:
> LGTM as a first step, too, but I think this should have Crankshaft support.
> 
> https://codereview.chromium.org/169783002/diff/130001/src/harmony-math.js
> File src/harmony-math.js (right):
> 
>
https://codereview.chromium.org/169783002/diff/130001/src/harmony-math.js#new...
> src/harmony-math.js:157: function MathClz32(x) {
> What TC39 seems to intend is basically exposing some form of intrinsics for
> processor intstructions. AFAICT "count leading zeroes" has HW support for all
> our platforms and boils down to a single CPU instructions, and I think this is
> the expectation of people using such intrinsics. Therefore we should add
> Crankshaft support, which shouldn't be too hard.

talked to dmitry about this offline. tahts essentially the plan.

[Sven Panne, 2014-02-18 07:28:20]

Math.imul, Math.clz32, ... What will be next? :)

[Yang, 2014-02-18 10:49:41]

Committed patchset #4 manually as r19435 (presubmit successful).