[interpreter] First implementation of stack unwinding.

[interpreter] First implementation of stack unwinding.

This implements a first prototype of stack unwinding for interpreted
frames. The unwinding machinery performs a range-based lookup in the
given handler table and potentially continues dispatching at the handler
offset. Note that this does not yet correctly restore the context to the
correct value when the handler is being entered.

R=rmcilroy@chromium.org,oth@chromium.org
BUG=v8:4674
LOG=n

Committed: https://crrev.com/0b3066b8f5d914bb1a8c5dd0404a866d389d7798
Cr-Commit-Position: refs/heads/master@{#33414}

[Michael Starzinger, 2016-01-19 18:21:51]

Ports are not yet done, but I wanted to get a first round of comments before I
do the architecture ports. Otherwise I think this is a reasonable check-point to
be landed. WDYT?

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc
File src/x64/builtins-x64.cc (right):

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc#new...
src/x64/builtins-x64.cc:901: void
Builtins::Generate_InterpreterEnterExceptionHandler(MacroAssembler* masm) {
Note that this is an unmodified copy of the code from notify-deoptimized helper
above. Please advise on whether you want me to factor this out now, or in a
later CL.

[oth, 2016-01-20 07:49:08]

Superb! Very clean. This would be good to land as soon as other platforms are
ready.

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc
File src/x64/builtins-x64.cc (right):

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc#new...
src/x64/builtins-x64.cc:901: void
Builtins::Generate_InterpreterEnterExceptionHandler(MacroAssembler* masm) {
On 2016/01/19 18:21:51, Michael Starzinger wrote:
> Note that this is an unmodified copy of the code from notify-deoptimized
helper
> above. Please advise on whether you want me to factor this out now, or in a
> later CL.

For now I'd settle for a comment at the start of the function explaining it's an
unmodified copy and then focus on other platforms.

[rmcilroy, 2016-01-20 08:52:13]

Very nice, pleasantly surprised how little extra support you needed to add.
Lgtm.

https://codereview.chromium.org/1605633003/diff/1/src/frames.cc
File src/frames.cc (right):

https://codereview.chromium.org/1605633003/diff/1/src/frames.cc#newcode1143
src/frames.cc:1143: int raw_offset = Smi::cast(GetExpression(1))->value();
Nit - could you pull the '1' out to a constant somewhere? It would just make
things easier if we need to change the frame layout later.

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc
File src/x64/builtins-x64.cc (right):

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc#new...
src/x64/builtins-x64.cc:901: void
Builtins::Generate_InterpreterEnterExceptionHandler(MacroAssembler* masm) {
On 2016/01/19 18:21:51, Michael Starzinger wrote:
> Note that this is an unmodified copy of the code from notify-deoptimized
helper
> above. Please advise on whether you want me to factor this out now, or in a
> later CL.

If it's unmodified let's just factor this out in this CL if it's not too much
trouble.

[Michael Starzinger, 2016-01-20 11:45:27]

Thanks. Comments addressed. Going for ports now.

https://codereview.chromium.org/1605633003/diff/1/src/frames.cc
File src/frames.cc (right):

https://codereview.chromium.org/1605633003/diff/1/src/frames.cc#newcode1143
src/frames.cc:1143: int raw_offset = Smi::cast(GetExpression(1))->value();
On 2016/01/20 08:52:13, rmcilroy wrote:
> Nit - could you pull the '1' out to a constant somewhere? It would just make
> things easier if we need to change the frame layout later.

Done. Added the constant to the InterpreterFrameConstants class.

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc
File src/x64/builtins-x64.cc (right):

https://codereview.chromium.org/1605633003/diff/1/src/x64/builtins-x64.cc#new...
src/x64/builtins-x64.cc:901: void
Builtins::Generate_InterpreterEnterExceptionHandler(MacroAssembler* masm) {
On 2016/01/20 08:52:13, rmcilroy wrote:
> On 2016/01/19 18:21:51, Michael Starzinger wrote:
> > Note that this is an unmodified copy of the code from notify-deoptimized
> helper
> > above. Please advise on whether you want me to factor this out now, or in a
> > later CL.
> 
> If it's unmodified let's just factor this out in this CL if it's not too much
> trouble.

Done. Since it's unmodified, I went with a minimal refactoring in this CL.

[Michael Starzinger, 2016-01-20 14:21:42]

PTAL, this is ready for final review.

Note that I had to disable a handful of tests. They started failing now, because
the finally block does not yet rethrow exceptions correctly. Before we wouldn't
got through the finally on exceptions, just propagate the exception out. Now we
properly go through the finally, but then forget to re-throw the exception.

[oth, 2016-01-20 14:48:43]

very cool, lgtm.

[Michael Starzinger, 2016-01-20 17:49:02]

The CQ bit was checked by mstarzinger@chromium.org

[Michael Starzinger, 2016-01-20 17:49:02]

The patchset sent to the CQ was uploaded after l-g-t-m from
rmcilroy@chromium.org
Link to the patchset: https://codereview.chromium.org/1605633003/#ps150001
(title: "Rebase and skip one more test.")

[commit-bot: I haz the power, 2016-01-20 17:49:15]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1605633003/150001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1605633003/150001

[commit-bot: I haz the power, 2016-01-20 18:10:22]

Committed patchset #9 (id:150001)

[commit-bot: I haz the power, 2016-01-20 18:10:47]

Description was changed from

==========
[interpreter] First implementation of stack unwinding.

This implements a first prototype of stack unwinding for interpreted
frames. The unwinding machinery performs a range-based lookup in the
given handler table and potentially continues dispatching at the handler
offset. Note that this does not yet correctly restore the context to the
correct value when the handler is being entered.

R=rmcilroy@chromium.org,oth@chromium.org
BUG=v8:4674
LOG=n
==========

to

==========
[interpreter] First implementation of stack unwinding.

This implements a first prototype of stack unwinding for interpreted
frames. The unwinding machinery performs a range-based lookup in the
given handler table and potentially continues dispatching at the handler
offset. Note that this does not yet correctly restore the context to the
correct value when the handler is being entered.

R=rmcilroy@chromium.org,oth@chromium.org
BUG=v8:4674
LOG=n

Committed: https://crrev.com/0b3066b8f5d914bb1a8c5dd0404a866d389d7798
Cr-Commit-Position: refs/heads/master@{#33414}
==========

[commit-bot: I haz the power, 2016-01-20 18:10:48]

Patchset 9 (id:??) landed as
https://crrev.com/0b3066b8f5d914bb1a8c5dd0404a866d389d7798
Cr-Commit-Position: refs/heads/master@{#33414}