Ability to invoke all std benchmarks via jmh

[Akirathan]

Closes #7489

Pull Request Description

Important Notes

• All the benchmarks have configured phases (number of iterations/seconds for warmup or measurement) such that the total number of benchmark invocations per warmup and measurement is somewhere around 100 at least and their score is not less than 1 ms.
• There is yet no API for providing setup or teardown methods.
  • An example of how lazy setup can be done for Text/compare is in 3ca6cd7
  • Without the lazy setup, the collection of benchmarks in Text/compare takes 7 seconds, with lazy setup, the collection takes 160 ms.

Checklist

Please ensure that the following checklist has been satisfied before submitting the PR:

• The documentation has been updated, if necessary.
• Screenshots/screencasts have been attached, if there are any visual changes. For interactive or animated visual changes, a screencast is preferred.
• All code follows the
  Scala,
  Java,
  and
  Rust
  style guides. In case you are using a language not listed above, follow the Rust style guide.
• All code has been tested:
  • Unit tests have been written where possible.
  • If GUI codebase was changed, the GUI was tested when built using ./run ide build.

[JaroslavTulach]

Perfect and sufficient, in my opinion.

[radeusgd]

This looks perfect! ❤️

[JaroslavTulach]

As far as I can say we don't need teardown right now. As such, please let's keep this pull request simple and focused. Let's use existing Enso language lazy atom fields concept to get proper setup behavior without changing the BenchProcessor, without introducing set_setup and set_teardown and without having problems to invoke these methods from JMH.

[Akirathan]

As far as I can say we don't need teardown right now. As such, please let's keep this pull request simple and focused. Let's use existing Enso language lazy atom fields concept to get proper setup behavior without changing the BenchProcessor, without introducing set_setup and set_teardown and without having problems to invoke these methods from JMH.

Sounds like a good idea. I reverted all the commits that introduced the special API for setup and teardown methods. For future reference, it is kept in a new issue #7566.

[radeusgd]

The direction looks good and I agree with @JaroslavTulach on setup/teardown - let's keep it simple. I think computing the lazy values on first warmup iteration should be acceptable - it's warmup after all. We just should make sure that they are indeed not recomputed in the measurement phase.

And let's ensure the data for benchmarks is always computed lazily so it is not computed in the 'gathering benchmark structure' phase, as it may slow it down.

[JaroslavTulach]

I see the CI is green. I think the current state is good. Approving.

[JaroslavTulach]

Shouldn't all the collect_benches method accept a Bench.builder? Then this would be:

Bench.build builder ->
  Array_Proxy_Bench.collect_benches builder
  Distinct.callect_benches builder
  ...

Then we wouldn't need to iterate Vector, but rather:

main =
    all_benchmarks . run_main

[Akirathan]

The GH Workflow is added in #7597. Since we need to first merge #7597 to try the new action, let's already merge this PR. I will also revert the commits that contain the CI-related changes.