[turbofan] Introduce DeoptimizeIf And DeoptimizeUnless common operators.

[turbofan] Introduce DeoptimizeIf And DeoptimizeUnless common operators.

These macro operators represent a conditional eager deoptimization exit
without explicit branching, which greatly reduces overhead of both
scheduling and register allocation, and thereby greatly reduces overall
compilation time, esp. when there are a lot of eager deoptimization
exits.

R=jarin@chromium.org

Committed: https://crrev.com/c129aa4d39364bf78d93bec8c16811bfde040a37
Cr-Commit-Position: refs/heads/master@{#34239}

[Benedikt Meurer, 2016-02-23 08:14:09]

Hey Jaro,

Here's the baby. It looks like two Kraken benchmarks actually benefitted from
the ability to schedule stuff in the Deoptimize block; we might need to look
into that later and see if we can do something about it (it's due to the hole
NaN, so we could also say we wait for NaN boxing to autosolve the problem for
us).
Please take a look.

Thanks,
Benedikt

[Benedikt Meurer, 2016-02-23 09:06:34]

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

[commit-bot: I haz the power, 2016-02-23 09:06:47]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1721103003/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1721103003/1

[commit-bot: I haz the power, 2016-02-23 10:47:58]

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

[commit-bot: I haz the power, 2016-02-23 10:47:59]

Dry run: Try jobs failed on following builders:
  v8_linux_arm64_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux_arm64_rel/builds/1...)

[Jarin, 2016-02-24 08:56:39]

lgtm. (Not very pretty, but gets the job done, I suppose.)

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/branch-eli...
File src/compiler/branch-elimination.cc (right):

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/branch-eli...
src/compiler/branch-elimination.cc:102: return Replace(control);
Perhaps add a comment saying that we do not need to update the conditions here
because we are replacing with the 'control' node and that node already contains
the right information.

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/js-native-...
File src/compiler/js-native-context-specialization.cc (right):

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/js-native-...
src/compiler/js-native-context-specialization.cc:566: if (--num_classes == 0 &&
num_transitions == 0 &&
Side-effect in shortcut operators, what could possibly go wrong?

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/js-native-...
src/compiler/js-native-context-specialization.cc:568:
this_controls.push_back(graph()->NewNode(common()->DeoptimizeUnless(),
Please, add a comment explaining what you are doing here. (I assume this is
handling the last conditional in the chain and the black magic with the counts
is there just to track which one is the last one).

[Benedikt Meurer, 2016-02-24 09:04:56]

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/branch-eli...
File src/compiler/branch-elimination.cc (right):

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/branch-eli...
src/compiler/branch-elimination.cc:102: return Replace(control);
On 2016/02/24 08:56:38, Jarin wrote:
> Perhaps add a comment saying that we do not need to update the conditions here
> because we are replacing with the 'control' node and that node already
contains
> the right information.

Done.

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/js-native-...
File src/compiler/js-native-context-specialization.cc (right):

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/js-native-...
src/compiler/js-native-context-specialization.cc:566: if (--num_classes == 0 &&
num_transitions == 0 &&
On 2016/02/24 08:56:38, Jarin wrote:
> Side-effect in shortcut operators, what could possibly go wrong?

Acknowledged.

https://codereview.chromium.org/1721103003/diff/20001/src/compiler/js-native-...
src/compiler/js-native-context-specialization.cc:568:
this_controls.push_back(graph()->NewNode(common()->DeoptimizeUnless(),
On 2016/02/24 08:56:38, Jarin wrote:
> Please, add a comment explaining what you are doing here. (I assume this is
> handling the last conditional in the chain and the black magic with the counts
> is there just to track which one is the last one).

Done.

[Benedikt Meurer, 2016-02-24 09:05:03]

The CQ bit was checked by bmeurer@chromium.org

[Benedikt Meurer, 2016-02-24 09:05:03]

The patchset sent to the CQ was uploaded after l-g-t-m from jarin@chromium.org
Link to the patchset: https://codereview.chromium.org/1721103003/#ps60001
(title: "Add comments")

[commit-bot: I haz the power, 2016-02-24 09:05:10]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1721103003/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1721103003/60001

[commit-bot: I haz the power, 2016-02-24 09:49:20]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2016-02-24 09:49:57]

Description was changed from

==========
[turbofan] Introduce DeoptimizeIf And DeoptimizeUnless common operators.

These macro operators represent a conditional eager deoptimization exit
without explicit branching, which greatly reduces overhead of both
scheduling and register allocation, and thereby greatly reduces overall
compilation time, esp. when there are a lot of eager deoptimization
exits.

R=jarin@chromium.org
==========

to

==========
[turbofan] Introduce DeoptimizeIf And DeoptimizeUnless common operators.

These macro operators represent a conditional eager deoptimization exit
without explicit branching, which greatly reduces overhead of both
scheduling and register allocation, and thereby greatly reduces overall
compilation time, esp. when there are a lot of eager deoptimization
exits.

R=jarin@chromium.org

Committed: https://crrev.com/c129aa4d39364bf78d93bec8c16811bfde040a37
Cr-Commit-Position: refs/heads/master@{#34239}
==========

[commit-bot: I haz the power, 2016-02-24 09:49:58]

Patchset 4 (id:??) landed as
https://crrev.com/c129aa4d39364bf78d93bec8c16811bfde040a37
Cr-Commit-Position: refs/heads/master@{#34239}