Lazy deoptimization for comparisons in Turbofan.

Lazy deoptimization for comparisons in Turbofan.

BUG=
R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=23594

[Jarin, 2014-09-01 11:38:11]

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

[Jarin, 2014-09-01 11:38:27]

Could you take a look, please?

[Michael Starzinger, 2014-09-01 23:10:46]

LGTM, just minor comments.

https://codereview.chromium.org/526953004/diff/1/src/compiler/js-generic-lowe...
File src/compiler/js-generic-lowering.cc (right):

https://codereview.chromium.org/526953004/diff/1/src/compiler/js-generic-lowe...
src/compiler/js-generic-lowering.cc:314: DCHECK(!has_frame_state);
Ouch, this is getting complex, but we can always clean it up later. But could we
preserve the comment about why we add effect and control for pure nodes?

https://codereview.chromium.org/526953004/diff/1/src/compiler/js-generic-lowe...
src/compiler/js-generic-lowering.cc:383: Operator* new_op =
common()->Call(desc);
nit: Please keep this consistent with the other ReplaceWithFoo helpers, let's
either hoist the operator into a local variable everywhere or undo this change
here.

https://codereview.chromium.org/526953004/diff/1/src/compiler/verifier.cc
File src/compiler/verifier.cc (right):

https://codereview.chromium.org/526953004/diff/1/src/compiler/verifier.cc#new...
src/compiler/verifier.cc:78:
CHECK(NodeProperties::GetFrameStateInput(node)->op()->opcode() ==
nit: There should be Node::opcode() directly IIRC.

https://codereview.chromium.org/526953004/diff/1/src/compiler/verifier.cc#new...
src/compiler/verifier.cc:79: IrOpcode::kFrameState);
suggestion: Do you think it makes sense to check that def-use and use-def chains
are present (i.e. Is[DefUse|UseDef]ChainLinkPresent)?

https://codereview.chromium.org/526953004/diff/1/test/cctest/cctest.status
File test/cctest/cctest.status (left):

https://codereview.chromium.org/526953004/diff/1/test/cctest/cctest.status#ol...
test/cctest/cctest.status:105: 'test-deoptimization/DeoptimizeCompare': [PASS,
NO_VARIANTS],
Nice!

[Jarin, 2014-09-02 10:38:51]

Committed patchset #2 (id:20001) manually as 23594 (presubmit successful).