[turbofan] Call FastNewContextStub for function context.

[turbofan] Call FastNewContextStub for function context.

This lowers JSCreateFunctionContext nodes to call the above stub for
help with allocating function contexts when possible. It also contains
an implementation for inlined allocations of such contexts, which is
still behind a flag until inlined allocations are ready for prime time.

TEST=unittests/JSTypedLoweringTest.JSCreateFunctionContext
R=mvstanton@chromium.org

Committed: https://crrev.com/ea26401266f4ae7c157026c96c7b6456c0489e93
Cr-Commit-Position: refs/heads/master@{#31068}

[Michael Starzinger, 2015-10-01 12:05:55]

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

[Michael Starzinger, 2015-10-01 12:05:55]

Michael: PTAL.
Benedikt: FYI.

[Benedikt Meurer, 2015-10-01 13:10:17]

Awesome! LGTM with one nit.

https://codereview.chromium.org/1380113002/diff/1/test/unittests/compiler/js-...
File test/unittests/compiler/js-typed-lowering-unittest.cc (right):

https://codereview.chromium.org/1380113002/diff/1/test/unittests/compiler/js-...
test/unittests/compiler/js-typed-lowering-unittest.cc:1094: FLAG_turbo_allocate
= true;
Mhm, global flags in unittests are bad, since you can run any number of
unittests in the same process (our test runner currently doesn't take advantage
of that, but that's more of a bug/shortcoming in the test runner).
I could live with this if you put a big fat TODO on that line and promise to
remove it ASAP :-)

[Michael Starzinger, 2015-10-01 16:18:59]

https://codereview.chromium.org/1380113002/diff/1/test/unittests/compiler/js-...
File test/unittests/compiler/js-typed-lowering-unittest.cc (right):

https://codereview.chromium.org/1380113002/diff/1/test/unittests/compiler/js-...
test/unittests/compiler/js-typed-lowering-unittest.cc:1094: FLAG_turbo_allocate
= true;
On 2015/10/01 13:10:16, Benedikt Meurer wrote:
> Mhm, global flags in unittests are bad, since you can run any number of
> unittests in the same process (our test runner currently doesn't take
advantage
> of that, but that's more of a bug/shortcoming in the test runner).
> I could live with this if you put a big fat TODO on that line and promise to
> remove it ASAP :-)

Done. I don't really want to put a TODO here, because that won't prevent silent
breakage should we ever switch to multiple unit-tests per process before I get
to addressing this. But I changed it so that this unit-test is skipped when
inlined allocation is disabled. We do this in other unit-tests as well. Please
advise if this works for you. WDYT?

Also note that counters-unittest.cc still suffers from the same problem.

[Benedikt Meurer, 2015-10-01 16:57:16]

On 2015/10/01 16:18:59, Michael Starzinger wrote:
>
https://codereview.chromium.org/1380113002/diff/1/test/unittests/compiler/js-...
> File test/unittests/compiler/js-typed-lowering-unittest.cc (right):
> 
>
https://codereview.chromium.org/1380113002/diff/1/test/unittests/compiler/js-...
> test/unittests/compiler/js-typed-lowering-unittest.cc:1094:
FLAG_turbo_allocate
> = true;
> On 2015/10/01 13:10:16, Benedikt Meurer wrote:
> > Mhm, global flags in unittests are bad, since you can run any number of
> > unittests in the same process (our test runner currently doesn't take
> advantage
> > of that, but that's more of a bug/shortcoming in the test runner).
> > I could live with this if you put a big fat TODO on that line and promise to
> > remove it ASAP :-)
> 
> Done. I don't really want to put a TODO here, because that won't prevent
silent
> breakage should we ever switch to multiple unit-tests per process before I get
> to addressing this. But I changed it so that this unit-test is skipped when
> inlined allocation is disabled. We do this in other unit-tests as well. Please
> advise if this works for you. WDYT?
> 
> Also note that counters-unittest.cc still suffers from the same problem.

Skipping is fine with me. I think unittesting and dependency on global state
don't work together, that's why I tend to dislike every FLAG by default.
For unittests, we usually introduce a flag/mode for the unit under test
and set that flag/mode in production depending on some global FLAG
or compilationinfo flag. But hopefully it's not worth the effort for
inline allocate, since we will hopefully be able to kill that FLAG really
soonish.

[mvstanton, 2015-10-02 07:10:03]

LGTM from here.

[Michael Starzinger, 2015-10-02 08:14:21]

The CQ bit was checked by mstarzinger@chromium.org

[Michael Starzinger, 2015-10-02 08:14:21]

The patchset sent to the CQ was uploaded after l-g-t-m from bmeurer@chromium.org
Link to the patchset: https://codereview.chromium.org/1380113002/#ps20001
(title: "Addressed comment.")

[commit-bot: I haz the power, 2015-10-02 08:14:30]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1380113002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1380113002/20001

[commit-bot: I haz the power, 2015-10-02 09:30:19]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-10-02 09:30:34]

Patchset 2 (id:??) landed as
https://crrev.com/ea26401266f4ae7c157026c96c7b6456c0489e93
Cr-Commit-Position: refs/heads/master@{#31068}