8383892: Shenandoah: Decouple allocation rate sampling from GC cycle

[earthling-amzn]

TL;DR;

The idea for this change started way back during the discussion of https://git.openjdk.org/jdk/pull/27398. I have been nursing the branch along for many months. The accelerated triggers PR threw it for a loop, but I have recently found the time to work through those changes.

• The bytes_allocated_since_last_gc and all of its usages are removed
• The old statistical based spike detector is removed
• Tracking gc times and allocation rates share a common ShenandoahWeightedSeq now

Motivation

Presently, Shenandoah accumulates allocations in a bytes_allocated_since_last_gc variable. Although it is sampled periodically, it is only reset when a gc cycles runs. Correctly managing this variable has proven to be complicated:

• https://bugs.openjdk.org/browse/JDK-8368015
• https://bugs.openjdk.org/browse/JDK-8371852
• https://bugs.openjdk.org/browse/JDK-8380459
• https://bugs.openjdk.org/browse/JDK-8381985

This approach leads to inconsistent allocation rate samples and predictions, and code complexity.

This task decouples the management of the allocation accumulation and sampling from GC cycles and moves it to the allocation thread.

Results

Consider these critical_jops results from specjbb 2015 on a small (10g) heap:

                               Category | Count | Total | GeoMean | Average | Trim 0.1 | StdDev | Minimum | Maximum
    control/small/genshen critical_jops | 6 | 43673.000 | 7278.234 | 7278.833 | 7278.833 | 102.597 | 7177.000 | 7409.000
  treatment/small/genshen critical_jops | 4 | 31176.000 | 7793.102 | 7794.000 | 7794.000 | 137.193 | 7698.000 | 7989.000
       control/small/shen critical_jops | 6 | 45773.000 | 7627.889 | 7628.833 | 7628.833 | 131.119 | 7409.000 | 7800.000
     treatment/small/shen critical_jops | 4 | 31391.000 | 7847.140 | 7847.750 | 7847.750 | 113.500 | 7791.000 | 8018.000

On a medium (30g) heap:

                               Category | Count | Total | GeoMean | Average | Trim 0.1 | StdDev | Minimum | Maximum
   control/medium/genshen critical_jops | 6 | 101680.000 | 16943.068 | 16946.667 | 16946.667 | 383.252 | 16564.000 | 17401.000
 treatment/medium/genshen critical_jops | 4 | 71153.000 | 17773.961 | 17788.250 | 17788.250 | 825.516 | 16911.000 | 18810.000
      control/medium/shen critical_jops | 6 | 104370.000 | 17383.564 | 17395.000 | 17395.000 | 698.918 | 16911.000 | 18603.000
    treatment/medium/shen critical_jops | 4 | 75240.000 | 18810.000 | 18810.000 | 18810.000 | 0.000 | 18810.000 | 18810.000

p100 latency measurements on the Extremem benchmark are also improved.

---

• I confirm that I make this contribution in accordance with the OpenJDK Interim AI Policy.
• I made this code with my own bare hands.

---

Progress

• Change must not contain extraneous whitespace
• Commit message must refer to an issue
• Change must be properly reviewed (2 reviews required, with at least 1 Reviewer, 1 Author)

Issue

• JDK-8383892: Shenandoah: Decouple allocation rate sampling from GC cycle (Sub-task - P4)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/31047/head:pull/31047
$ git checkout pull/31047

Update a local copy of the PR:
$ git checkout pull/31047
$ git pull https://git.openjdk.org/jdk.git pull/31047/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 31047

View PR using the GUI difftool:
$ git pr show -t 31047

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/31047.diff

Using Webrev

Link to Webrev Comment

[bridgekeeper]

👋 Welcome back wkemper! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

[openjdk]

❗ This change is not yet ready to be integrated.
See the Progress checklist in the description for automated requirements.

[openjdk]

@earthling-amzn The following labels will be automatically applied to this pull request:

• hotspot-gc
• shenandoah

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing lists. If you would like to change these labels, use the /label pull request command.

[openjdk]

The total number of required reviews for this PR has been set to 2 based on the presence of this label: hotspot-gc. This can be overridden with the /reviewers command.

[mlbridge]

Webrevs

• 00: Full (4f4f71a4)