[Code Stubs] Convert FastNewClosureStub to a TurboFanCodeStub.

[Code Stubs] Convert FastNewClosureStub to a TurboFanCodeStub.

Converts FastNewClosureStub from a Hydrogen to a TurboFan code stub.
The plan is to start using this in the Interpreter CreateClosure
bytecode handler (in a follow-up CL).

BUG=v8:4280
Committed: https://crrev.com/f0a430e5dcfaaa27ab7cef19c935cd255d0c42d1
Cr-Commit-Position: refs/heads/master@{#37429}

[rmcilroy, 2016-06-27 16:07:45]

The CQ bit was checked by rmcilroy@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-06-27 16:07:49]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-06-27 16:33:38]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-27 16:33:38]

Dry run: This issue passed the CQ dry run.

[rmcilroy, 2016-06-27 16:37:52]

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

[rmcilroy, 2016-06-27 16:37:53]

Benedikt, PTAL thanks.

[Benedikt Meurer, 2016-06-27 18:33:55]

https://codereview.chromium.org/2100883003/diff/1/src/code-stubs.h
File src/code-stubs.h (right):

https://codereview.chromium.org/2100883003/diff/1/src/code-stubs.h#newcode1043
src/code-stubs.h:1043: minor_key_ = (LanguageModeBits::encode(language_mode) |
So both the language_mode and the kind are deriv(ed|able) from the
SharedFunctionInfo and we pass in the SharedFunctionInfo. How about computing
the map index at runtime (making the encoding more cleverish could help)? Then
we don't need any minor key at all, but a single code object can handle
everything (similar to the byte code handler then).

[rmcilroy, 2016-06-29 15:52:08]

The CQ bit was checked by rmcilroy@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-06-29 15:52:14]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-06-29 16:11:38]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-29 16:11:39]

Dry run: Try jobs failed on following builders:
  v8_mac_rel_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_mac_rel_ng/builds/4181)
  v8_mac_rel_ng_triggered on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_mac_rel_ng_triggered/bui...)

[Benedikt Meurer, 2016-06-29 17:42:16]

Nice, looks promising modulo test failures.

https://codereview.chromium.org/2100883003/diff/20001/src/code-stubs.cc
File src/code-stubs.cc (right):

https://codereview.chromium.org/2100883003/diff/20001/src/code-stubs.cc#newco...
src/code-stubs.cc:4553: // TODO(rmcilroy): Should we set the code entry from the
SharedFunctionInfo
We want to go to optimized code if it's already available in the optimized code
map. Installing the baseline code would force us to baseline first, and might
even result in unnecessary OSR.

[rmcilroy, 2016-06-29 21:11:48]

https://codereview.chromium.org/2100883003/diff/1/src/code-stubs.h
File src/code-stubs.h (right):

https://codereview.chromium.org/2100883003/diff/1/src/code-stubs.h#newcode1043
src/code-stubs.h:1043: minor_key_ = (LanguageModeBits::encode(language_mode) |
On 2016/06/27 18:33:55, Benedikt Meurer wrote:
> So both the language_mode and the kind are deriv(ed|able) from the
> SharedFunctionInfo and we pass in the SharedFunctionInfo. How about computing
> the map index at runtime (making the encoding more cleverish could help)? Then
> we don't need any minor key at all, but a single code object can handle
> everything (similar to the byte code handler then).

Done. PTAL, thanks.

https://codereview.chromium.org/2100883003/diff/20001/src/code-stubs.cc
File src/code-stubs.cc (right):

https://codereview.chromium.org/2100883003/diff/20001/src/code-stubs.cc#newco...
src/code-stubs.cc:4553: // TODO(rmcilroy): Should we set the code entry from the
SharedFunctionInfo
On 2016/06/29 17:42:16, Benedikt Meurer wrote:
> We want to go to optimized code if it's already available in the optimized
code
> map. Installing the baseline code would force us to baseline first, and might
> even result in unnecessary OSR.

Ahh yes, of course. Removed this TODO.

Is there any thought that we could enable generated code to look up the
optimized code map and avoid the roundtrip to the runtime for new functions, or
is this likely not worth the effort?

[rmcilroy, 2016-06-29 21:11:54]

The CQ bit was checked by rmcilroy@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-06-29 21:12:07]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-06-29 21:44:34]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-29 21:44:35]

Dry run: This issue passed the CQ dry run.

[Benedikt Meurer, 2016-06-30 03:30:16]

https://codereview.chromium.org/2100883003/diff/20001/src/code-stubs.cc
File src/code-stubs.cc (right):

https://codereview.chromium.org/2100883003/diff/20001/src/code-stubs.cc#newco...
src/code-stubs.cc:4553: // TODO(rmcilroy): Should we set the code entry from the
SharedFunctionInfo
CompileLazy doesn't do a round trip to the runtime if code is already available.

[Benedikt Meurer, 2016-06-30 03:30:21]

lgtm

[rmcilroy, 2016-06-30 09:40:19]

The CQ bit was checked by rmcilroy@chromium.org

[commit-bot: I haz the power, 2016-06-30 09:40:30]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-06-30 09:42:35]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-06-30 09:45:17]

Description was changed from

==========
[Code Stubs] Convert FastNewClosureStub to a TurboFanCodeStub.

Converts FastNewClosureStub from a Hydrogen to a TurboFan code stub.
The plan is to start using this in the Interpreter CreateClosure
bytecode handler (in a follow-up CL).

BUG=v8:4280
==========

to

==========
[Code Stubs] Convert FastNewClosureStub to a TurboFanCodeStub.

Converts FastNewClosureStub from a Hydrogen to a TurboFan code stub.
The plan is to start using this in the Interpreter CreateClosure
bytecode handler (in a follow-up CL).

BUG=v8:4280

Committed: https://crrev.com/f0a430e5dcfaaa27ab7cef19c935cd255d0c42d1
Cr-Commit-Position: refs/heads/master@{#37429}
==========

[commit-bot: I haz the power, 2016-06-30 09:45:18]

Patchset 3 (id:??) landed as
https://crrev.com/f0a430e5dcfaaa27ab7cef19c935cd255d0c42d1
Cr-Commit-Position: refs/heads/master@{#37429}