[turbofan] Unwind and jump to the catch handler in the deoptimizer.

[turbofan] Unwind and jump to the catch handler in the deoptimizer.

The idea here is to perform the handler lookup in the deoptimizer, and then take the information from the handler table to build the catch handler frame in the deoptimizer. Specifically, we use the pc offset, context location and stack height (in full-code) to tweak the output frame.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it keeps variables alive if they are used in the catch handler.

Committed: https://crrev.com/ab3b3bec86f2c61e3640a2a3f01b5d6d319ad7aa
Cr-Commit-Position: refs/heads/master@{#33936}

[Jarin, 2015-11-05 10:21:56]

Description was changed from

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer.

The idea here is to leave the handler lookup and the stack massage
to the deoptimizer.

Sadly, this still requires nasty woodoo for the liveness analyzer.
==========

to

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer [WIP]

The idea here is to leave the handler lookup and the stack massage
to the deoptimizer.

Sadly, this still requires nasty woodoo for the liveness analyzer.
==========

[Jarin, 2016-02-05 11:35:28]

Description was changed from

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer [WIP]

The idea here is to leave the handler lookup and the stack massage
to the deoptimizer.

Sadly, this still requires nasty woodoo for the liveness analyzer.
==========

to

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer.

The idea here is to leave the handler lookup and the stack massageto the
deoptimizer.

Sadly, this still requires nasty woodoo for the liveness analyzer.
==========

[Jarin, 2016-02-05 11:35:28]

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

[Jarin, 2016-02-05 11:56:30]

Description was changed from

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer.

The idea here is to leave the handler lookup and the stack massageto the
deoptimizer.

Sadly, this still requires nasty woodoo for the liveness analyzer.
==========

to

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer.

The idea here is to leave the handler lookup and the stack massaging to the
deoptimizer, using the existing information in the handler table.

Sadly, this still requires nasty woodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

[Jarin, 2016-02-05 11:56:45]

Could you take a look, please?

[Benedikt Meurer, 2016-02-05 12:05:53]

LGTM on ast.

[Jarin, 2016-02-08 10:35:51]

Description was changed from

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer.

The idea here is to leave the handler lookup and the stack massaging to the
deoptimizer, using the existing information in the handler table.

Sadly, this still requires nasty woodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

to

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer.

The idea here is to leave the handler lookup and the stack massaging to the
deoptimizer, using the existing information in the handler table.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

[Jarin, 2016-02-10 08:11:01]

Description was changed from

==========
[turbofan] Deoptimize to the catch handler in the deoptimizer.

The idea here is to leave the handler lookup and the stack massaging to the
deoptimizer, using the existing information in the handler table.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

to

==========
[turbofan] Unwind and jump to the catch handler in the deoptimizer.

The idea here is to perform the handler lookup in the deoptimizer, and then take
the information from the handler table to build the catch handler frame in the
deoptimizer. Specifically, we use the pc offset, context location and stack
height (in full-code) to tweak the output frame.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

[Jarin, 2016-02-10 08:13:58]

Michi, I believe the interpreter deopt works now, too. I have also fixed the
context deopt and added some tests for it (I have basically just tweaked your
try-context.js tests to do {lazy-deopt, throw} rather than just {throw}).

Could you take a look now?

[Michael Starzinger, 2016-02-12 09:32:41]

LGTM (squinting eyes).

https://codereview.chromium.org/1416543006/diff/140001/src/ast/ast-numbering.cc
File src/ast/ast-numbering.cc (right):

https://codereview.chromium.org/1416543006/diff/140001/src/ast/ast-numbering....
src/ast/ast-numbering.cc:308: DisableCrankshaft(kTryCatchStatement);
nit: Should we land this as a separate CL, so that it will be easier to revert
without reverting the whole thing? WDYT?

https://codereview.chromium.org/1416543006/diff/140001/src/compiler/ast-graph...
File src/compiler/ast-graph-builder.cc (right):

https://codereview.chromium.org/1416543006/diff/140001/src/compiler/ast-graph...
src/compiler/ast-graph-builder.cc:414: bool NodeHasExceptionUse(Node* node) {
Please use NodeProperties::IsExceptionalCall instead of introducing this new
predicate. I would be fine with adding the fast property pre-check to that
predicate if you think it makes sense.

[Jarin, 2016-02-12 09:58:32]

Thanks!

https://codereview.chromium.org/1416543006/diff/140001/src/ast/ast-numbering.cc
File src/ast/ast-numbering.cc (right):

https://codereview.chromium.org/1416543006/diff/140001/src/ast/ast-numbering....
src/ast/ast-numbering.cc:308: DisableCrankshaft(kTryCatchStatement);
On 2016/02/12 09:32:40, Michael Starzinger wrote:
> nit: Should we land this as a separate CL, so that it will be easier to revert
> without reverting the whole thing? WDYT?

Done.

https://codereview.chromium.org/1416543006/diff/140001/src/compiler/ast-graph...
File src/compiler/ast-graph-builder.cc (right):

https://codereview.chromium.org/1416543006/diff/140001/src/compiler/ast-graph...
src/compiler/ast-graph-builder.cc:414: bool NodeHasExceptionUse(Node* node) {
On 2016/02/12 09:32:40, Michael Starzinger wrote:
> Please use NodeProperties::IsExceptionalCall instead of introducing this new
> predicate. I would be fine with adding the fast property pre-check to that
> predicate if you think it makes sense.

Done.

[Jarin, 2016-02-12 09:58:37]

The CQ bit was checked by jarin@chromium.org

[Jarin, 2016-02-12 09:58:37]

The patchset sent to the CQ was uploaded after l-g-t-m from
bmeurer@chromium.org, mstarzinger@chromium.org
Link to the patchset: https://codereview.chromium.org/1416543006/#ps160001
(title: "Rebase, review comments")

[commit-bot: I haz the power, 2016-02-12 09:58:42]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1416543006/160001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1416543006/160001

[commit-bot: I haz the power, 2016-02-12 10:14:46]

Description was changed from

==========
[turbofan] Unwind and jump to the catch handler in the deoptimizer.

The idea here is to perform the handler lookup in the deoptimizer, and then take
the information from the handler table to build the catch handler frame in the
deoptimizer. Specifically, we use the pc offset, context location and stack
height (in full-code) to tweak the output frame.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

to

==========
[turbofan] Unwind and jump to the catch handler in the deoptimizer.

The idea here is to perform the handler lookup in the deoptimizer, and then take
the information from the handler table to build the catch handler frame in the
deoptimizer. Specifically, we use the pc offset, context location and stack
height (in full-code) to tweak the output frame.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

[commit-bot: I haz the power, 2016-02-12 10:14:46]

Committed patchset #9 (id:160001)

[commit-bot: I haz the power, 2016-02-12 10:15:40]

Description was changed from

==========
[turbofan] Unwind and jump to the catch handler in the deoptimizer.

The idea here is to perform the handler lookup in the deoptimizer, and then take
the information from the handler table to build the catch handler frame in the
deoptimizer. Specifically, we use the pc offset, context location and stack
height (in full-code) to tweak the output frame.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.
==========

to

==========
[turbofan] Unwind and jump to the catch handler in the deoptimizer.

The idea here is to perform the handler lookup in the deoptimizer, and then take
the information from the handler table to build the catch handler frame in the
deoptimizer. Specifically, we use the pc offset, context location and stack
height (in full-code) to tweak the output frame.

Sadly, this still requires nasty voodoo for the liveness analyzer so that it
keeps variables alive if they are used in the catch handler.

Committed: https://crrev.com/ab3b3bec86f2c61e3640a2a3f01b5d6d319ad7aa
Cr-Commit-Position: refs/heads/master@{#33936}
==========

[commit-bot: I haz the power, 2016-02-12 10:15:41]

Patchset 9 (id:??) landed as
https://crrev.com/ab3b3bec86f2c61e3640a2a3f01b5d6d319ad7aa
Cr-Commit-Position: refs/heads/master@{#33936}