[turbofan] Add new Float32Abs and Float64Abs operators.

[turbofan] Add new Float32Abs and Float64Abs operators.

These operators compute the absolute floating point value of some
arbitrary input, and are implemented without any branches (i.e. using
vabs on arm, and andps/andpd on x86).

R=svenpanne@chromium.org

Committed: https://crrev.com/9af9f1d0262e2dab66c62331a03f6a76504f3866
Cr-Commit-Position: refs/heads/master@{#27662}

[Benedikt Meurer, 2015-04-08 09:13:23]

bmeurer@chromium.org changed reviewers:
+ mstarzinger@chromium.org

[Benedikt Meurer, 2015-04-08 09:13:24]

Hey Michi,

This seems to give a nice boost on Embenchen Bullet, which uses Math.abs quite
heavily. It's big, but mostly straightforward. Please take a look.

-- Benedikt

[Benedikt Meurer, 2015-04-08 10:32:05]

bmeurer@chromium.org changed reviewers:
+ svenpanne@chromium.org
- mstarzinger@chromium.org

[Sven Panne, 2015-04-08 11:52:03]

LGTM with a nit

https://codereview.chromium.org/1066393002/diff/60001/src/compiler/machine-op...
File src/compiler/machine-operator.h (right):

https://codereview.chromium.org/1066393002/diff/60001/src/compiler/machine-op...
src/compiler/machine-operator.h:202: // Floating point abs complying to IEEE 753
(single-precision).
+1 ^^

[Benedikt Meurer, 2015-04-08 11:54:42]

https://codereview.chromium.org/1066393002/diff/60001/src/compiler/machine-op...
File src/compiler/machine-operator.h (right):

https://codereview.chromium.org/1066393002/diff/60001/src/compiler/machine-op...
src/compiler/machine-operator.h:202: // Floating point abs complying to IEEE 753
(single-precision).
On 2015/04/08 11:52:03, Sven Panne wrote:
> +1 ^^

Done.

[Benedikt Meurer, 2015-04-08 11:55:16]

Committed patchset #5 (id:80001) manually as
9af9f1d0262e2dab66c62331a03f6a76504f3866 (presubmit successful).

[commit-bot: I haz the power, 2015-04-08 11:55:17]

Patchset 5 (id:??) landed as
https://crrev.com/9af9f1d0262e2dab66c62331a03f6a76504f3866
Cr-Commit-Position: refs/heads/master@{#27662}