Handle HCheckInstanceType and HIsStringAndBranch in check elimination.

Handle HCheckInstanceType and HIsStringAndBranch in check elimination.

R=ishell@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=21593

Committed: https://code.google.com/p/v8/source/detail?r=21606

[Benedikt Meurer, 2014-05-29 11:31:42]

PTAL

[Igor Sheludko, 2014-05-30 10:00:41]

Mostly nits + what about adding a couple of tests that triggers this
optimization as a seed for the fuzzer?

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
File src/hydrogen-check-elimination.cc (right):

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.cc:294: }
Suggestion: What about adding Subtract() method to UniqueSet?

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.cc:475: }
Does it make sense to update set of entry maps if the instruction ensures string
or non-string type even if the instruction itself can't be eliminated?

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.cc:623:
instr->block()->MarkSuccEdgeUnreachable(unreachable_succ);
I would suggest to avoid duplication of the three lines above.

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
File src/hydrogen-check-elimination.h (right):

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.h:31: removed_cho_ = 0;
removed_cit_ = 0;

[Benedikt Meurer, 2014-05-31 11:15:46]

Comments addressed. There are several tests for HCheckInstanceType already,
because it's used for string adds in BuildBinaryOperation(). HIsStringAndBranch
isn't really used right now, but will be used with Array.indexOf inlining
(separate CL).

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
File src/hydrogen-check-elimination.cc (right):

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.cc:294: }
On 2014/05/30 10:00:41, Igor Sheludko wrote:
> Suggestion: What about adding Subtract() method to UniqueSet?

Done.

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.cc:475: }
On 2014/05/30 10:00:41, Igor Sheludko wrote:
> Does it make sense to update set of entry maps if the instruction ensures
string
> or non-string type even if the instruction itself can't be eliminated?

Done.

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.cc:623:
instr->block()->MarkSuccEdgeUnreachable(unreachable_succ);
On 2014/05/30 10:00:41, Igor Sheludko wrote:
> I would suggest to avoid duplication of the three lines above.

Done.

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
File src/hydrogen-check-elimination.h (right):

https://codereview.chromium.org/300423003/diff/20001/src/hydrogen-check-elimi...
src/hydrogen-check-elimination.h:31: removed_cho_ = 0;
On 2014/05/30 10:00:41, Igor Sheludko wrote:
> removed_cit_ = 0;

Done.

[Igor Sheludko, 2014-06-02 08:46:21]

lgtm

[Benedikt Meurer, 2014-06-02 08:51:32]

Committed patchset #2 manually as r21593 (presubmit successful).

[Benedikt Meurer, 2014-06-02 12:17:16]

Committed patchset #2 manually as r21606 (presubmit successful).