Stabilize escape analysis (without deopt)

[turbofan] Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n

Committed: https://crrev.com/3161c17126e73f965a24f43172581f4e09fdd80e
Cr-Commit-Position: refs/heads/master@{#32833}

[sigurds, 2015-12-10 15:31:53]

Description was changed from

==========
Stabilize escape analysis (without deopt).

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
==========

to

==========
Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
==========

[sigurds, 2015-12-10 15:40:17]

PTAL

https://codereview.chromium.org/1510973006/diff/40001/src/compiler/pipeline.cc
File src/compiler/pipeline.cc (right):

https://codereview.chromium.org/1510973006/diff/40001/src/compiler/pipeline.c...
src/compiler/pipeline.cc:1170: Run<EarlyGraphTrimmingPhase>();
Please advise what to do here.

[Michael Starzinger, 2015-12-14 10:41:49]

LGTM.

https://codereview.chromium.org/1510973006/diff/40001/src/compiler/pipeline.cc
File src/compiler/pipeline.cc (right):

https://codereview.chromium.org/1510973006/diff/40001/src/compiler/pipeline.c...
src/compiler/pipeline.cc:1170: Run<EarlyGraphTrimmingPhase>();
On 2015/12/10 15:40:17, sigurds wrote:
> Please advise what to do here.

Acknowledged. For now I am fine with running the graph trimming as a pre-pass to
the escape analysis. Not sure yet how to exactly tackle this, but we can do it
after stabilization. Leaving a TODO in this regard might make sense though.

https://codereview.chromium.org/1510973006/diff/80001/src/compiler/escape-ana...
File src/compiler/escape-analysis-reducer.cc (right):

https://codereview.chromium.org/1510973006/diff/80001/src/compiler/escape-ana...
src/compiler/escape-analysis-reducer.cc:154: if
(escape_analysis()->IsAllocation(input)) {
IIUC, for the escape analysis it would suffice to only do this reduction if
"IsVirtual(input)". So we could limit the reducer to that case and leave the
general optimization (i.e. ObjectIsSmi(Allocate) nodes) to another reducer. Then
we would not need the additional "IsAllocation" function either I think.

[sigurds, 2015-12-14 11:58:09]

Thanks for the comments!

https://codereview.chromium.org/1510973006/diff/40001/src/compiler/pipeline.cc
File src/compiler/pipeline.cc (right):

https://codereview.chromium.org/1510973006/diff/40001/src/compiler/pipeline.c...
src/compiler/pipeline.cc:1170: Run<EarlyGraphTrimmingPhase>();
On 2015/12/14 10:41:49, Michael Starzinger wrote:
> On 2015/12/10 15:40:17, sigurds wrote:
> > Please advise what to do here.
> 
> Acknowledged. For now I am fine with running the graph trimming as a pre-pass
to
> the escape analysis. Not sure yet how to exactly tackle this, but we can do it
> after stabilization. Leaving a TODO in this regard might make sense though.

Done.

https://codereview.chromium.org/1510973006/diff/80001/src/compiler/escape-ana...
File src/compiler/escape-analysis-reducer.cc (right):

https://codereview.chromium.org/1510973006/diff/80001/src/compiler/escape-ana...
src/compiler/escape-analysis-reducer.cc:154: if
(escape_analysis()->IsAllocation(input)) {
On 2015/12/14 10:41:49, Michael Starzinger wrote:
> IIUC, for the escape analysis it would suffice to only do this reduction if
> "IsVirtual(input)". So we could limit the reducer to that case and leave the
> general optimization (i.e. ObjectIsSmi(Allocate) nodes) to another reducer.
Then
> we would not need the additional "IsAllocation" function either I think.

Done.

[sigurds, 2015-12-14 12:23:12]

Description was changed from

==========
Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
==========

to

==========
[turbofan] Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
==========

[sigurds, 2015-12-14 12:53:52]

The CQ bit was checked by sigurds@chromium.org

[sigurds, 2015-12-14 12:53:53]

The patchset sent to the CQ was uploaded after l-g-t-m from
mstarzinger@chromium.org
Link to the patchset: https://codereview.chromium.org/1510973006/#ps120001
(title: "Rebase")

[commit-bot: I haz the power, 2015-12-14 12:53:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1510973006/120001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1510973006/120001

[commit-bot: I haz the power, 2015-12-14 13:13:29]

Description was changed from

==========
[turbofan] Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
==========

to

==========
[turbofan] Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
==========

[commit-bot: I haz the power, 2015-12-14 13:13:30]

Committed patchset #7 (id:120001)

[commit-bot: I haz the power, 2015-12-14 13:14:24]

Description was changed from

==========
[turbofan] Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
==========

to

==========
[turbofan] Stabilize escape analysis (without deopt)

Bugfixes and improvements in escape analysis include:

* Handling of ObjectIsSmi (non-escaping)
* Handling of nested phi replacements
* Handling of phis with arity > 2
* Resilience against effectful nodes dangling from start
* Allocations escape now, if non-const load/store is performed
* Fixed a bug where non-allocated objects where tracked
* Allow fixed double arrays to be tracked

R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n

Committed: https://crrev.com/3161c17126e73f965a24f43172581f4e09fdd80e
Cr-Commit-Position: refs/heads/master@{#32833}
==========

[commit-bot: I haz the power, 2015-12-14 13:14:25]

Patchset 7 (id:??) landed as
https://crrev.com/3161c17126e73f965a24f43172581f4e09fdd80e
Cr-Commit-Position: refs/heads/master@{#32833}