[x86] Use AVX in Crankshaft when available.

[x86] Use AVX in Crankshaft when available.

R=verwaest@chromium.org

Committed: https://crrev.com/622be8f71e70b6ece4ea6a89bcfa1bc4be5e70c1
Cr-Commit-Position: refs/heads/master@{#26159}

Committed: https://crrev.com/03fbb053b39b4765a5b848e4b3dddbccb8d19889
Cr-Commit-Position: refs/heads/master@{#26235}

[Benedikt Meurer, 2015-01-20 13:52:14]

PTAL

[Toon Verwaest, 2015-01-20 13:54:00]

lgtm

[Benedikt Meurer, 2015-01-20 13:54:12]

The CQ bit was checked by bmeurer@chromium.org

[Toon Verwaest, 2015-01-20 13:54:24]

The CQ bit was checked by verwaest@chromium.org

[commit-bot: I haz the power, 2015-01-20 13:54:48]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/860003002/1

[Benedikt Meurer, 2015-01-20 13:59:01]

The CQ bit was unchecked by bmeurer@chromium.org

[Benedikt Meurer, 2015-01-20 13:59:31]

Committed patchset #1 (id:1) manually as
622be8f71e70b6ece4ea6a89bcfa1bc4be5e70c1 (presubmit successful).

[commit-bot: I haz the power, 2015-01-20 13:59:34]

Patchset 1 (id:??) landed as
https://crrev.com/622be8f71e70b6ece4ea6a89bcfa1bc4be5e70c1
Cr-Commit-Position: refs/heads/master@{#26159}

[Michael Achenbach, 2015-01-22 21:26:49]

A revert of this CL (patchset #1 id:1) has been created in
https://codereview.chromium.org/863423002/ by machenbach@chromium.org.

The reason for reverting is: Breaks chromium vista and XP browser tests:
http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_vist....

[Benedikt Meurer, 2015-01-23 08:25:32]

Committed patchset #1 (id:1) manually as
03fbb053b39b4765a5b848e4b3dddbccb8d19889 (presubmit successful).

[commit-bot: I haz the power, 2015-01-23 08:25:36]

Patchset 1 (id:??) landed as
https://crrev.com/03fbb053b39b4765a5b848e4b3dddbccb8d19889
Cr-Commit-Position: refs/heads/master@{#26235}

[Johann, 2015-01-23 19:10:25]

FYI - the cpu detection in v8/src/base/cpu.cc doesn't appear to take into
account the OSXSAVE bit:

https://gerrit.chromium.org/gerrit/#/c/73502/
and
https://software.intel.com/en-us/articles/how-to-detect-new-instruction-suppo...

of course, even with those checks in place libvpx is failing in a potentially
similar way. if you know of access to a winxp machine i could use to verify
this, i would appreciate it.