[vm/libs] Improve JsonUtf8Decoder performance.

This CL focuses on improving parsing of white space (between
JSON tokens) and simple strings which don't contain escape
sequences inside.

Improvements are achieved by changing the code to
table driven implementation instead of if-cascade: we
have a 256 element table which stores attributes for characters
(e.g. whether it is a white space or a terminal token for
a simple string) and use this table to make decisions
on whether to advance through characters or stop a loop and
do something else.

We also suppress bounds checks and interrupt checks in tight
loops - in tight loops like this a bound checks can cost
30% in overhead.

This CL brings 28% geomean improvement on benchmarks from the linked issue. (All measurements are done in X64 Product AOT)

Individual measurements are:

| Input JSON | ms/iter | vs HEAD | vs V8 |
| ---------- | ------- | ------- | ----- |
| apache_builds.json | 0.44 | 61.06% | 136.86% |
| canada.json | 31.04 | 96.54% | 187.15% |
| citm_catalog.json | 6.43 | 64.44% | 93.94% |
| github_events.json | 0.23 | 59.02% | 128.86% |
| google_maps_api_compact_response.json | 0.10 | 82.12% | 133.83% |
| google_maps_api_response.json | 0.12 | 68.79% | 140.07% |
| gsoc-2018.json | 9.25 | 44.89% | 147.43% |
| instruments.json | 1.08 | 70.18% | 167.38% |
| marine_ik.json | 21.07 | 88.25% | 142.58% |
| mesh.json | 4.51 | 94.56% | 136.57% |
| mesh.pretty.json | 9.97 | 83.76% | 193.79% |
| numbers.json | 0.57 | 91.88% | 83.37% |
| random.json | 3.79 | 78.32% | 107.18% |
| repeat.json | 0.06 | 71.51% | 118.47% |
| semanticscholar-corpus.json | 37.65 | 54.82% | 57.81% |
| tree-pretty.json | 0.17 | 68.68% | 162.33% |
| twitter_api_compact_response.json | 0.06 | 75.23% | 126.11% |
| twitter_api_response.json | 0.08 | 70.64% | 123.60% |
| twitterescaped.json | 3.88 | 84.66% | 177.94% |
| twitter.json | 3.54 | 73.01% | 105.33% |
| twitter_timeline.json | 0.37 | 81.52% | 271.37% |
| update-center.json | 2.85 | 66.75% | 89.01% |

vs HEAD (geomean): 72.94%
vs V8 (geomean): 130.99%
HEAD vs V8 (geomean): 179.59%

Issue https://github.com/dart-lang/sdk/issues/55522

TEST=covered by co19

Change-Id: Id673118c19250ab7781cc98c7656b972debc60ff
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/365803
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
2 files changed
tree: 29997e0dd71383ae8e9ef744bb9b1ee4677d63d1
  1. .dart_tool/
  2. .github/
  3. benchmarks/
  4. build/
  5. docs/
  6. pkg/
  7. runtime/
  8. samples/
  9. sdk/
  10. tests/
  11. third_party/
  12. tools/
  13. utils/
  14. .clang-format
  15. .gitattributes
  16. .gitconfig
  17. .gitignore
  18. .gn
  19. .mailmap
  20. .style.yapf
  21. AUTHORS
  22. BUILD.gn
  23. CHANGELOG.md
  24. codereview.settings
  25. CONTRIBUTING.md
  26. DEPS
  27. LICENSE
  28. OWNERS
  29. PATENT_GRANT
  30. PRESUBMIT.py
  31. README.dart-sdk
  32. README.md
  33. sdk.code-workspace
  34. sdk_args.gni
  35. sdk_packages.yaml
  36. SECURITY.md
  37. WATCHLISTS
README.md

Dart

An approachable, portable, and productive language for high-quality apps on any platform

Dart is:

  • Approachable: Develop with a strongly typed programming language that is consistent, concise, and offers modern language features like null safety and patterns.

  • Portable: Compile to ARM, x64, or RISC-V machine code for mobile, desktop, and backend. Compile to JavaScript or WebAssembly for the web.

  • Productive: Make changes iteratively: use hot reload to see the result instantly in your running app. Diagnose app issues using DevTools.

Dart's flexible compiler technology lets you run Dart code in different ways, depending on your target platform and goals:

  • Dart Native: For programs targeting devices (mobile, desktop, server, and more), Dart Native includes both a Dart VM with JIT (just-in-time) compilation and an AOT (ahead-of-time) compiler for producing machine code.

  • Dart Web: For programs targeting the web, Dart Web includes both a development time compiler (dartdevc) and a production time compiler (dart2js).

Dart platforms illustration

License & patents

Dart is free and open source.

See LICENSE and PATENT_GRANT.

Using Dart

Visit dart.dev to learn more about the language, tools, and to find codelabs.

Browse pub.dev for more packages and libraries contributed by the community and the Dart team.

Our API reference documentation is published at api.dart.dev, based on the stable release. (We also publish docs from our beta and dev channels, as well as from the primary development branch).

Building Dart

If you want to build Dart yourself, here is a guide to getting the source, preparing your machine to build the SDK, and building.

There are more documents on our wiki.

Contributing to Dart

The easiest way to contribute to Dart is to file issues.

You can also contribute patches, as described in Contributing.

Roadmap

Future plans for Dart are included in the combined Dart and Flutter roadmap on the Flutter wiki.