[turbofan] Fix write barrier handling of map values.

[turbofan] Fix write barrier handling of map values.

We cannot omit flag check with kPointersToHereAreInterestingMask for maps because incremental marker dynamically sets and clears the flag.

BUG=chromium:587004
LOG=NO

Committed: https://crrev.com/5d06548208b338b6dd5f84cb6b94ccc47985abdc
Cr-Commit-Position: refs/heads/master@{#34165}

[ulan, 2016-02-19 16:31:31]

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

[ulan, 2016-02-19 16:31:31]

ptal

[Michael Starzinger, 2016-02-19 17:08:16]

LGTM, nice catch! Could you add a brief explanation to the CL description that
this is due to incremental marking, where pointers to maps become interesting as
well?

https://codereview.chromium.org/1714513003/diff/20001/src/compiler/arm/code-g...
File src/compiler/arm/code-generator-arm.cc (left):

https://codereview.chromium.org/1714513003/diff/20001/src/compiler/arm/code-g...
src/compiler/arm/code-generator-arm.cc:243: EMIT_REMEMBERED_SET, save_fp_mode);
As discussed offline: For map values we could actually call the RecordWriteStub
with OMIT_REMEMBERED_SET here. But lets do that in a separate CL.

[ulan, 2016-02-19 17:09:56]

Description was changed from

==========
[turbofan] Fix write barrier handling of map values.

BUG=chromium:587004
LOG=NO
==========

to

==========
[turbofan] Fix write barrier handling of map values.

We cannot omit flag check with kPointersToHereAreInterestingMask for maps
because incremental marker dynamically sets and clears the flag.


BUG=chromium:587004
LOG=NO
==========

[ulan, 2016-02-19 17:10:12]

On 2016/02/19 17:08:16, Michael Starzinger wrote:
> LGTM, nice catch! Could you add a brief explanation to the CL description that
> this is due to incremental marking, where pointers to maps become interesting
as
> well?
> 
>
https://codereview.chromium.org/1714513003/diff/20001/src/compiler/arm/code-g...
> File src/compiler/arm/code-generator-arm.cc (left):
> 
>
https://codereview.chromium.org/1714513003/diff/20001/src/compiler/arm/code-g...
> src/compiler/arm/code-generator-arm.cc:243: EMIT_REMEMBERED_SET,
save_fp_mode);
> As discussed offline: For map values we could actually call the
RecordWriteStub
> with OMIT_REMEMBERED_SET here. But lets do that in a separate CL.

Thanks! Updated the description.

[ulan, 2016-02-19 17:10:23]

The CQ bit was checked by ulan@chromium.org

[commit-bot: I haz the power, 2016-02-19 17:10:34]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1714513003/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1714513003/20001

[ulan, 2016-02-19 17:23:37]

The CQ bit was unchecked by ulan@chromium.org

[ulan, 2016-02-19 17:24:38]

The CQ bit was checked by ulan@chromium.org

[ulan, 2016-02-19 17:24:39]

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/1714513003/#ps40001
(title: "Add arm64")

[commit-bot: I haz the power, 2016-02-19 17:24:49]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1714513003/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1714513003/40001

[commit-bot: I haz the power, 2016-02-19 18:45:43]

Description was changed from

==========
[turbofan] Fix write barrier handling of map values.

We cannot omit flag check with kPointersToHereAreInterestingMask for maps
because incremental marker dynamically sets and clears the flag.


BUG=chromium:587004
LOG=NO
==========

to

==========
[turbofan] Fix write barrier handling of map values.

We cannot omit flag check with kPointersToHereAreInterestingMask for maps
because incremental marker dynamically sets and clears the flag.


BUG=chromium:587004
LOG=NO
==========

[commit-bot: I haz the power, 2016-02-19 18:45:44]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-02-19 18:46:31]

Description was changed from

==========
[turbofan] Fix write barrier handling of map values.

We cannot omit flag check with kPointersToHereAreInterestingMask for maps
because incremental marker dynamically sets and clears the flag.


BUG=chromium:587004
LOG=NO
==========

to

==========
[turbofan] Fix write barrier handling of map values.

We cannot omit flag check with kPointersToHereAreInterestingMask for maps
because incremental marker dynamically sets and clears the flag.

BUG=chromium:587004
LOG=NO

Committed: https://crrev.com/5d06548208b338b6dd5f84cb6b94ccc47985abdc
Cr-Commit-Position: refs/heads/master@{#34165}
==========

[commit-bot: I haz the power, 2016-02-19 18:46:32]

Patchset 3 (id:??) landed as
https://crrev.com/5d06548208b338b6dd5f84cb6b94ccc47985abdc
Cr-Commit-Position: refs/heads/master@{#34165}

[Benedikt Meurer, 2016-02-19 18:57:51]

Nice catch. Does the map write barrier make sense to have at all?

[Michael Starzinger, 2016-02-19 19:13:47]

On 2016/02/19 18:57:51, Benedikt Meurer wrote:
> Nice catch. Does the map write barrier make sense to have at all?

As mentioned in my comment, I think we could use the the RecordWriteStub with
OMIT_REMEMBERED_SET for maps, as they are guaranteed never to be in new-space.