[turbofan] Initial support for escape analysis.

[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also 8 new mjsunit tests, some of which are
skiped as they require features not yet implemented.

BUG=v8:4586
LOG=n

Committed: https://crrev.com/aa0ddf7db41eee028cd86ba23f74ac56e5c7e84f
Cr-Commit-Position: refs/heads/master@{#32498}

[sigurds, 2015-11-18 13:08:02]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

There are a 8 new mjsunit tests.

BUG=
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

There are a 8 new mjsunit tests.

BUG=
==========

[sigurds, 2015-11-18 13:08:02]

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

[sigurds, 2015-11-19 10:51:05]

Patchset #2 (id:20001) has been deleted

[sigurds, 2015-11-19 11:58:56]

Patchset #2 (id:40001) has been deleted

[sigurds, 2015-11-19 12:05:09]

Patchset #3 (id:80001) has been deleted

[sigurds, 2015-11-19 12:05:18]

Patchset #2 (id:60001) has been deleted

[sigurds, 2015-11-19 12:26:45]

Patchset #2 (id:100001) has been deleted

[sigurds, 2015-11-19 12:27:35]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

There are a 8 new mjsunit tests.

BUG=
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

There are a 8 new mjsunit tests, which are mask due to required changes in
https://codereview.chromium.org/1447323005/

There are also 3 unittests.

BUG=
==========

[sigurds, 2015-11-19 12:42:24]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

There are a 8 new mjsunit tests, which are mask due to required changes in
https://codereview.chromium.org/1447323005/

There are also 3 unittests.

BUG=
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

The CL includes 3 unittests.

There are also a 8 new mjsunit tests, which are skiped as they trigger bugs (see
https://codereview.chromium.org/1447323005/)


BUG=
==========

[sigurds, 2015-11-19 14:07:57]

Please take a first look. I guess there will be a lot of comments.

[sigurds, 2015-11-20 11:43:20]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

The CL includes 3 unittests.

There are also a 8 new mjsunit tests, which are skiped as they trigger bugs (see
https://codereview.chromium.org/1447323005/)


BUG=
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are skiped as they trigger
bugs in the implementation.


BUG=
==========

[sigurds, 2015-11-20 11:43:40]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are skiped as they trigger
bugs in the implementation.


BUG=
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are skiped as they require
features not yet implemented.


BUG=
==========

[sigurds, 2015-11-20 11:44:00]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan. At the moment, there is
no deopt support, and support for loops is partial (only binary Phis are
handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are skiped as they require
features not yet implemented.


BUG=
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are
skiped as they require features not yet implemented.


BUG=
==========

[Michael Starzinger, 2015-11-30 10:05:39]

On a high level this is looking good already. Mostly nits or smaller comments. I
haven't dug through all of the escape-analysis.cc file, but the rest is fine
after this round of comments. Cool stuff!

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis-reducer.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:47: DCHECK(node->opcode() ==
IrOpcode::kLoadField);
nit: DCHECK_EQ, here and throughout the file.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:48: if (Node* rep =
escape_analysis()->representation(node, node->id())) {
The term "representation" is already pretty overloaded in V8, I would vote for
using a different name for this method. How would you feel about calling this
"replacement" instead?

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:72: void
EscapeAnalysisReducer::RemoveFromEffectChain(Node* node) {
This should be equivalent with calling AdvancedReducer::RelaxEffectsAndControls
on the affected node. This helper would become obsolete and we could drop it.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:158: Reduction
EscapeAnalysisReducer::ReplaceWithDeoptDummy(Node* node) {
nit: Let's add a TODO comment here, saying that this is temporary until
materialization during deoptimization is handled.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:181: Graph*
EscapeAnalysisReducer::graph() const { return jsgraph()->graph(); }
nit: Method seems to be unused, can we drop it?

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis-reducer.h (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.h:22: class SimplifiedOperatorBuilder;
nit: Some of the forward declarations seem not to be used in this file anymore,
let's drop those.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.h:62: #endif  //
V8_COMPILER_JS_NATIVE_CONTEXT_SPECIALIZATION_H_
nit: s/JS_NATIVE_CONTEXT_SPECIALIZATION_H_/COMPILER_ESCAPE_ANALYSIS_REDUCER_H/

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis.cc:306: //
-----------------------------EscapeVirtualAnalysis---------------------------
nit: s/EscapeVirtualAnalysis/EscapeObjectAnalysis/

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis.h (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis.h:16: class TypeCache;
nit: Some of the forward declarations seem not to be used in this file anymore,
let's drop those.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/pipeline.cc
File src/compiler/pipeline.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/pipeline....
src/compiler/pipeline.cc:632: if (!FLAG_turbo_escape) return;
nit: Better hoist the flag check out to the call-site instead of having an early
return here (e.g. similar to FLAG_turbo_cf_optimization).

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/pipeline....
src/compiler/pipeline.cc:639: EscapeAnalysisReducer red(&graph_reducer,
data->jsgraph(), &escape_status,
nit: s/red/escape_reducer/

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/compiler/...
File test/mjsunit/compiler/escape-analysis-1.js (right):

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/compiler/...
test/mjsunit/compiler/escape-analysis-1.js:28: // Flags: --allow-natives-syntax
--turbo --turbo-escape
We could drop the "--turbo" from this line, because one of our variants will
contain it and we get broader coverage by also being able to run it against the
default variant that is shipping at the moment.

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/mjsunit.s...
File test/mjsunit/mjsunit.status (right):

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/mjsunit.s...
test/mjsunit/mjsunit.status:174: 'compiler/escape-analysis-1': [SKIP],
If the "--turbo" is dropped from the test flags, then "[PASS, NO_VARIANTS]"
here, ensuring that these tests are run only in the default variant for now.

https://codereview.chromium.org/1457683003/diff/170001/test/unittests/compile...
File test/unittests/compiler/escape-analysis-unittest.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/test/unittests/compile...
test/unittests/compiler/escape-analysis-unittest.cc:43: escape_status_ = new
EscapeStatusAnalysis(graph(), zone());
Would it be possible to just embed these two classes in this test class (i.e.
change from "EscapeStatusAnalysis*" to "EscapeStatusAnalysis"). Thereby no
allocation would be necessary, and also the deletion in the destructor would
disappear. Leaks would be ruled out. Currently, calling Analysis() twice for
example would cause a leak.

[sigurds, 2015-11-30 13:50:24]

Thanks for the comments, I addressed all of them.

escape-analysis.cc has been refactored.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis-reducer.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:47: DCHECK(node->opcode() ==
IrOpcode::kLoadField);
On 2015/11/30 10:05:38, Michael Starzinger wrote:
> nit: DCHECK_EQ, here and throughout the file.

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:48: if (Node* rep =
escape_analysis()->representation(node, node->id())) {
On 2015/11/30 10:05:38, Michael Starzinger wrote:
> The term "representation" is already pretty overloaded in V8, I would vote for
> using a different name for this method. How would you feel about calling this
> "replacement" instead?

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:72: void
EscapeAnalysisReducer::RemoveFromEffectChain(Node* node) {
On 2015/11/30 10:05:38, Michael Starzinger wrote:
> This should be equivalent with calling
AdvancedReducer::RelaxEffectsAndControls
> on the affected node. This helper would become obsolete and we could drop it.

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:158: Reduction
EscapeAnalysisReducer::ReplaceWithDeoptDummy(Node* node) {
On 2015/11/30 10:05:38, Michael Starzinger wrote:
> nit: Let's add a TODO comment here, saying that this is temporary until
> materialization during deoptimization is handled.

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.cc:181: Graph*
EscapeAnalysisReducer::graph() const { return jsgraph()->graph(); }
On 2015/11/30 10:05:38, Michael Starzinger wrote:
> nit: Method seems to be unused, can we drop it?

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis-reducer.h (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.h:22: class SimplifiedOperatorBuilder;
On 2015/11/30 10:05:38, Michael Starzinger wrote:
> nit: Some of the forward declarations seem not to be used in this file
anymore,
> let's drop those.

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.h:62: #endif  //
V8_COMPILER_JS_NATIVE_CONTEXT_SPECIALIZATION_H_
On 2015/11/30 10:05:38, Michael Starzinger wrote:
> nit: s/JS_NATIVE_CONTEXT_SPECIALIZATION_H_/COMPILER_ESCAPE_ANALYSIS_REDUCER_H/

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis.cc:306: //
-----------------------------EscapeVirtualAnalysis---------------------------
On 2015/11/30 10:05:39, Michael Starzinger wrote:
> nit: s/EscapeVirtualAnalysis/EscapeObjectAnalysis/

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
File src/compiler/escape-analysis.h (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/escape-an...
src/compiler/escape-analysis.h:16: class TypeCache;
On 2015/11/30 10:05:39, Michael Starzinger wrote:
> nit: Some of the forward declarations seem not to be used in this file
anymore,
> let's drop those.

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/pipeline.cc
File src/compiler/pipeline.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/pipeline....
src/compiler/pipeline.cc:632: if (!FLAG_turbo_escape) return;
On 2015/11/30 10:05:39, Michael Starzinger wrote:
> nit: Better hoist the flag check out to the call-site instead of having an
early
> return here (e.g. similar to FLAG_turbo_cf_optimization).

Done.

https://codereview.chromium.org/1457683003/diff/170001/src/compiler/pipeline....
src/compiler/pipeline.cc:639: EscapeAnalysisReducer red(&graph_reducer,
data->jsgraph(), &escape_status,
On 2015/11/30 10:05:39, Michael Starzinger wrote:
> nit: s/red/escape_reducer/

Done.

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/compiler/...
File test/mjsunit/compiler/escape-analysis-1.js (right):

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/compiler/...
test/mjsunit/compiler/escape-analysis-1.js:28: // Flags: --allow-natives-syntax
--turbo --turbo-escape
On 2015/11/30 10:05:39, Michael Starzinger wrote:
> We could drop the "--turbo" from this line, because one of our variants will
> contain it and we get broader coverage by also being able to run it against
the
> default variant that is shipping at the moment.

Done.

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/mjsunit.s...
File test/mjsunit/mjsunit.status (right):

https://codereview.chromium.org/1457683003/diff/170001/test/mjsunit/mjsunit.s...
test/mjsunit/mjsunit.status:174: 'compiler/escape-analysis-1': [SKIP],
On 2015/11/30 10:05:39, Michael Starzinger wrote:
> If the "--turbo" is dropped from the test flags, then "[PASS, NO_VARIANTS]"
> here, ensuring that these tests are run only in the default variant for now.

Done.

https://codereview.chromium.org/1457683003/diff/170001/test/unittests/compile...
File test/unittests/compiler/escape-analysis-unittest.cc (right):

https://codereview.chromium.org/1457683003/diff/170001/test/unittests/compile...
test/unittests/compiler/escape-analysis-unittest.cc:43: escape_status_ = new
EscapeStatusAnalysis(graph(), zone());
On 2015/11/30 10:05:39, Michael Starzinger wrote:
> Would it be possible to just embed these two classes in this test class (i.e.
> change from "EscapeStatusAnalysis*" to "EscapeStatusAnalysis"). Thereby no
> allocation would be necessary, and also the deletion in the destructor would
> disappear. Leaks would be ruled out. Currently, calling Analysis() twice for
> example would cause a leak.

Done.

[Benedikt Meurer, 2015-11-30 17:21:16]

https://codereview.chromium.org/1457683003/diff/230001/src/compiler/escape-an...
File src/compiler/escape-analysis.cc (right):

https://codereview.chromium.org/1457683003/diff/230001/src/compiler/escape-an...
src/compiler/escape-analysis.cc:511: ++count;
Nit: if (++count > 1) return true;
and then return false; below.

[sigurds, 2015-12-01 10:13:17]

Adressed the comments, more try jobs running.

https://codereview.chromium.org/1457683003/diff/230001/src/compiler/escape-an...
File src/compiler/escape-analysis.cc (right):

https://codereview.chromium.org/1457683003/diff/230001/src/compiler/escape-an...
src/compiler/escape-analysis.cc:511: ++count;
On 2015/11/30 17:21:16, Benedikt Meurer wrote:
> Nit: if (++count > 1) return true;
> and then return false; below.

Done.

[sigurds, 2015-12-01 15:10:07]

https://codereview.chromium.org/1457683003/diff/260001/src/compiler/escape-an...
File src/compiler/escape-analysis.cc (right):

https://codereview.chromium.org/1457683003/diff/260001/src/compiler/escape-an...
src/compiler/escape-analysis.cc:298: case IrOpcode::kLoadField: {
This is the bugfix.

[Michael Starzinger, 2015-12-01 17:18:38]

LGTM after last comments are addressed. Did a thorough review of all of the
integration pieces and the reducer. I haven't done a deep dive into all of
escape-analysis.cc yet, but on a high level it looks good too.

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
File src/compiler/escape-analysis-reducer.h (right):

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.h:10: #include "src/compiler/js-graph.h"
nit: It looks like the "js-graph.h" include is not necessary. Either way is fine
with me, but only forward declaring "JSGraph" or including the header should do
it.

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
File src/compiler/escape-analysis.h (right):

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
src/compiler/escape-analysis.h:79: class VirtualState : public ZoneObject {
Both the VirtualObject as well as the VirtualState classes are only used by the
"escape-analysis.cc" implementation itself. They are never passed around to
different components. I like that very much. I would even go a step further and
move those two classes into the .cc file. Thereby the chance of someone mocking
with internal state of the escape analysis are reduced even further.

https://codereview.chromium.org/1457683003/diff/280001/test/unittests/compile...
File test/unittests/compiler/escape-analysis-unittest.cc (right):

https://codereview.chromium.org/1457683003/diff/280001/test/unittests/compile...
test/unittests/compiler/escape-analysis-unittest.cc:146: void
ExpectRepresentation(Node* node, Node* rep) {
nit: s/Representation/Replacement/

https://codereview.chromium.org/1457683003/diff/280001/test/unittests/compile...
test/unittests/compiler/escape-analysis-unittest.cc:150: void
ExpectRepresentationPhi(Node* node, Node* left, Node* right) {
nit: Likewise.

[sigurds, 2015-12-02 09:30:52]

Patchset #12 (id:300001) has been deleted

[sigurds, 2015-12-02 10:12:20]

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
File src/compiler/escape-analysis-reducer.h (right):

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
src/compiler/escape-analysis-reducer.h:10: #include "src/compiler/js-graph.h"
On 2015/12/01 17:18:37, Michael Starzinger wrote:
> nit: It looks like the "js-graph.h" include is not necessary. Either way is
fine
> with me, but only forward declaring "JSGraph" or including the header should
do
> it.

Done.

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
File src/compiler/escape-analysis.h (right):

https://codereview.chromium.org/1457683003/diff/280001/src/compiler/escape-an...
src/compiler/escape-analysis.h:79: class VirtualState : public ZoneObject {
On 2015/12/01 17:18:38, Michael Starzinger wrote:
> Both the VirtualObject as well as the VirtualState classes are only used by
the
> "escape-analysis.cc" implementation itself. They are never passed around to
> different components. I like that very much. I would even go a step further
and
> move those two classes into the .cc file. Thereby the chance of someone
mocking
> with internal state of the escape analysis are reduced even further.

Done.

https://codereview.chromium.org/1457683003/diff/280001/test/unittests/compile...
File test/unittests/compiler/escape-analysis-unittest.cc (right):

https://codereview.chromium.org/1457683003/diff/280001/test/unittests/compile...
test/unittests/compiler/escape-analysis-unittest.cc:146: void
ExpectRepresentation(Node* node, Node* rep) {
On 2015/12/01 17:18:38, Michael Starzinger wrote:
> nit: s/Representation/Replacement/

Done.

https://codereview.chromium.org/1457683003/diff/280001/test/unittests/compile...
test/unittests/compiler/escape-analysis-unittest.cc:150: void
ExpectRepresentationPhi(Node* node, Node* left, Node* right) {
On 2015/12/01 17:18:38, Michael Starzinger wrote:
> nit: Likewise.

Done.

[sigurds, 2015-12-02 10:12:49]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are
skiped as they require features not yet implemented.


BUG=
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are
skiped as they require features not yet implemented.


BUG=v8:4586
LOG=n
==========

[sigurds, 2015-12-02 10:45:41]

The CQ bit was checked by sigurds@chromium.org

[sigurds, 2015-12-02 10:45:42]

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/1457683003/#ps340001
(title: "Rebase")

[commit-bot: I haz the power, 2015-12-02 10:45:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1457683003/340001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1457683003/340001

[sigurds, 2015-12-02 10:48:28]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also a 8 new mjsunit tests, two of which are
skiped as they require features not yet implemented.


BUG=v8:4586
LOG=n
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also 8 new mjsunit tests, some of which are
skiped as they require features not yet implemented.


BUG=v8:4586
LOG=n
==========

[commit-bot: I haz the power, 2015-12-02 10:53:31]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also 8 new mjsunit tests, some of which are
skiped as they require features not yet implemented.


BUG=v8:4586
LOG=n
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also 8 new mjsunit tests, some of which are
skiped as they require features not yet implemented.


BUG=v8:4586
LOG=n
==========

[commit-bot: I haz the power, 2015-12-02 10:53:32]

Committed patchset #13 (id:340001)

[commit-bot: I haz the power, 2015-12-02 10:53:55]

Description was changed from

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support 
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also 8 new mjsunit tests, some of which are
skiped as they require features not yet implemented.


BUG=v8:4586
LOG=n
==========

to

==========
[turbofan] Initial support for escape analysis.

This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support
for loops is partial (only binary Phis are handled).

The CL includes 4 unittests.

There are also 8 new mjsunit tests, some of which are
skiped as they require features not yet implemented.

BUG=v8:4586
LOG=n

Committed: https://crrev.com/aa0ddf7db41eee028cd86ba23f74ac56e5c7e84f
Cr-Commit-Position: refs/heads/master@{#32498}
==========

[commit-bot: I haz the power, 2015-12-02 10:53:56]

Patchset 13 (id:??) landed as
https://crrev.com/aa0ddf7db41eee028cd86ba23f74ac56e5c7e84f
Cr-Commit-Position: refs/heads/master@{#32498}