[regexp] Port test, match, and search

[regexp] Port test, match, and search

This CL ports RegExp.prototype.test, RegExp.prototype.match
and RegExp.prototype.search to C++.

Performance regressions are expected but should be improved
in an upcoming CL.

BUG=v8:5339
Committed: https://crrev.com/80f3c6ce7702c0cc28388ae4f084c23870a9aa4f
Cr-Commit-Position: refs/heads/master@{#40122}

[jgruber, 2016-10-05 09:35:11]

The CQ bit was checked by jgruber@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-10-05 09:35:20]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-10-05 09:54:29]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-10-05 09:54:29]

Dry run: Try jobs failed on following builders:
  v8_mac_rel_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_mac_rel_ng/builds/10147)
  v8_mac_rel_ng_triggered on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_mac_rel_ng_triggered/bui...)

[jgruber, 2016-10-06 13:56:36]

The CQ bit was checked by jgruber@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-10-06 13:56:45]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[jgruber, 2016-10-06 14:00:47]

jgruber@chromium.org changed reviewers:
+ yangguo@chromium.org

[jgruber, 2016-10-06 14:00:48]

I'm expecting fairly high performance regressions for these methods since this
is a straight-forward C++ port without any optimizations. 

We don't tank octane/regexp at all since it only cares about replace and exec.

These can be optimized later if needed through fast-paths in C++ or by
conversion to TF.

[commit-bot: I haz the power, 2016-10-06 14:27:04]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-10-06 14:27:04]

Dry run: This issue passed the CQ dry run.

[Yang, 2016-10-10 09:49:01]

https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
File src/builtins/builtins-regexp.cc (right):

https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
src/builtins/builtins-regexp.cc:1008: Execution::Call(isolate, exec, regexp,
argc, argv.start()), Object);
Would we gain performance if we could recognize the original RegExpPrototypeExec
stub and stay in C++ instead? Same further below.

Or maybe at least for atomic regexps, where we do not compile regexp code, but
always go to C++ anyways?

https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
src/builtins/builtins-regexp.cc:1071: increment = 2;
We could just return 2 here and return 1 below.

Also it would probably be closer to the name of this function to return index +
increment, and change the call sites accordingly.

[Yang, 2016-10-10 09:49:21]

On 2016/10/10 09:49:01, Yang wrote:
>
https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
> File src/builtins/builtins-regexp.cc (right):
> 
>
https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
> src/builtins/builtins-regexp.cc:1008: Execution::Call(isolate, exec, regexp,
> argc, argv.start()), Object);
> Would we gain performance if we could recognize the original
RegExpPrototypeExec
> stub and stay in C++ instead? Same further below.
> 
> Or maybe at least for atomic regexps, where we do not compile regexp code, but
> always go to C++ anyways?
> 
>
https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
> src/builtins/builtins-regexp.cc:1071: increment = 2;
> We could just return 2 here and return 1 below.
> 
> Also it would probably be closer to the name of this function to return index
+
> increment, and change the call sites accordingly.

otherwise LGTM.

[jgruber, 2016-10-10 13:01:41]

https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
File src/builtins/builtins-regexp.cc (right):

https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
src/builtins/builtins-regexp.cc:1008: Execution::Call(isolate, exec, regexp,
argc, argv.start()), Object);
On 2016/10/10 09:49:01, Yang wrote:
> Would we gain performance if we could recognize the original
RegExpPrototypeExec
> stub and stay in C++ instead? Same further below.
> 
> Or maybe at least for atomic regexps, where we do not compile regexp code, but
> always go to C++ anyways?

For IRREGEXP regexps, using RegExpImpl::Exec as it is at the moment would
probably be slower than calling into RegExpExecStub. For ATOM regexps, it
definitely makes sense.

Once all of this has landed, I'd like to take one more optimization pass through
all of these functions, and this one of the things I want to take another look
at then.

https://codereview.chromium.org/2394713003/diff/20001/src/builtins/builtins-r...
src/builtins/builtins-regexp.cc:1071: increment = 2;
On 2016/10/10 09:49:01, Yang wrote:
> We could just return 2 here and return 1 below.
> 
> Also it would probably be closer to the name of this function to return index
+
> increment, and change the call sites accordingly.

Good point, it's also closer to the spec that way. Again, I'd like to defer
until the remaining regexp CLs have landed since upcoming CLs push this code
around a bit and I don't want to miss any use site during rebasing.

[jgruber, 2016-10-10 13:15:49]

The CQ bit was checked by jgruber@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-10-10 13:15:58]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[jgruber, 2016-10-10 13:19:23]

The CQ bit was checked by jgruber@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-10-10 13:19:28]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[jgruber, 2016-10-10 13:36:27]

Description was changed from

==========
[regexp] Port test, match, and search

BUG=v8:5339
==========

to

==========
[regexp] Port test, match, and search

This CL ports RegExp.prototype.test, RegExp.prototype.match
and RegExp.prototype.search to C++.

Performance regressions are expected but should be improved
in an upcoming CL.

BUG=v8:5339
==========

[commit-bot: I haz the power, 2016-10-10 13:46:03]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-10-10 13:46:04]

Dry run: This issue passed the CQ dry run.

[jgruber, 2016-10-10 13:54:00]

The CQ bit was checked by jgruber@chromium.org

[jgruber, 2016-10-10 13:54:00]

The patchset sent to the CQ was uploaded after l-g-t-m from yangguo@chromium.org
Link to the patchset: https://codereview.chromium.org/2394713003/#ps60001
(title: "Rebaseline bytecode expectations")

[commit-bot: I haz the power, 2016-10-10 13:54:10]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-10-10 13:56:50]

Description was changed from

==========
[regexp] Port test, match, and search

This CL ports RegExp.prototype.test, RegExp.prototype.match
and RegExp.prototype.search to C++.

Performance regressions are expected but should be improved
in an upcoming CL.

BUG=v8:5339
==========

to

==========
[regexp] Port test, match, and search

This CL ports RegExp.prototype.test, RegExp.prototype.match
and RegExp.prototype.search to C++.

Performance regressions are expected but should be improved
in an upcoming CL.

BUG=v8:5339
==========

[commit-bot: I haz the power, 2016-10-10 13:56:51]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2016-10-10 13:57:27]

Description was changed from

==========
[regexp] Port test, match, and search

This CL ports RegExp.prototype.test, RegExp.prototype.match
and RegExp.prototype.search to C++.

Performance regressions are expected but should be improved
in an upcoming CL.

BUG=v8:5339
==========

to

==========
[regexp] Port test, match, and search

This CL ports RegExp.prototype.test, RegExp.prototype.match
and RegExp.prototype.search to C++.

Performance regressions are expected but should be improved
in an upcoming CL.

BUG=v8:5339

Committed: https://crrev.com/80f3c6ce7702c0cc28388ae4f084c23870a9aa4f
Cr-Commit-Position: refs/heads/master@{#40122}
==========

[commit-bot: I haz the power, 2016-10-10 13:57:28]

Patchset 4 (id:??) landed as
https://crrev.com/80f3c6ce7702c0cc28388ae4f084c23870a9aa4f
Cr-Commit-Position: refs/heads/master@{#40122}

[Michael Achenbach, 2016-10-11 06:50:31]

A revert of this CL (patchset #4 id:60001) has been created in
https://codereview.chromium.org/2404223002/ by machenbach@chromium.org.

The reason for reverting is: Seems to have made regexp-property-lu-ui more than
twice as slow in gc stress (and flaky slow):
https://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20gc%20stress/....