[turbofan] Do not set node's mark to min on get

[turbofan] Do not set node's mark to min on get

This simplifies the mark getter enough to generate a branch-free check
on many architectures (e.g. using cmov on x64 or csel on ARM).

As a drive-by, we can now make the mark access const.

Review-Url: https://codereview.chromium.org/2583043004
Cr-Commit-Position: refs/heads/master@{#41833}
Committed: https://chromium.googlesource.com/v8/v8/+/0a5bffe24d18bbad321c56e9125325b6b7589d50

[Leszek Swirski, 2016-12-19 17:45:00]

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

[commit-bot: I haz the power, 2016-12-19 17:45:08]

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

[Leszek Swirski, 2016-12-19 17:47:43]

leszeks@chromium.org changed reviewers:
+ bmeurer@chromium.org, jarin@chromium.org

[Leszek Swirski, 2016-12-19 17:47:44]

Hi Benedikt/Jaro,

I was seeing a lot of branch mispredictions (25%) on NodeMarkerBase::Get, and
this silly change is giving me a ~1% increase in compile speed (measured with
--turbo-stats on x64).

[Benedikt Meurer, 2016-12-19 17:51:20]

Nice catch! LGTM.

But I doubt that using cmov would be beneficial here, as cmov has a terrible
latency.

[commit-bot: I haz the power, 2016-12-19 18:12:34]

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

[commit-bot: I haz the power, 2016-12-19 18:12:34]

Dry run: This issue passed the CQ dry run.

[Leszek Swirski, 2016-12-20 08:51:14]

The CQ bit was checked by leszeks@chromium.org

[commit-bot: I haz the power, 2016-12-20 08:51:17]

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, 2016-12-20 08:52:55]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1482223874194250, "parent_rev":
"a141898108df9b01d524c57d15dea6a1220f31e2", "commit_rev":
"0a5bffe24d18bbad321c56e9125325b6b7589d50"}

[commit-bot: I haz the power, 2016-12-20 08:53:01]

Description was changed from

==========
[turbofan] Do not set node's mark to min on get

This simplifies the mark getter enough to generate a branch-free check
on many architectures (e.g. using cmov on x64 or csel on ARM).

As a drive-by, we can now make the mark access const.
==========

to

==========
[turbofan] Do not set node's mark to min on get

This simplifies the mark getter enough to generate a branch-free check
on many architectures (e.g. using cmov on x64 or csel on ARM).

As a drive-by, we can now make the mark access const.

Review-Url: https://codereview.chromium.org/2583043004
Cr-Commit-Position: refs/heads/master@{#41833}
Committed:
https://chromium.googlesource.com/v8/v8/+/0a5bffe24d18bbad321c56e9125325b6b75...
==========

[commit-bot: I haz the power, 2016-12-20 08:53:02]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/v8/v8/+/0a5bffe24d18bbad321c56e9125325b6b75...

[Leszek Swirski, 2016-12-20 08:55:11]

On 2016/12/19 17:51:20, Benedikt Meurer wrote:
> Nice catch! LGTM.
> 
> But I doubt that using cmov would be beneficial here, as cmov has a terrible
> latency.

Well, profiling shows a benefit, but I couldn't tell you if it's because of jmp
vs cmov, branch misprediction, smaller code/icache, or something I've completely
not thought of. Regardless, giving the compiler more chances to optimize never
hurts.