Signals 5.0.0 weaker performance with Rainbench and changed behavior observed

[rydmike]

I updated all packages on the Rainbench benchmark to latest versions and rerun all benchmarks.

I found some odd behavior with Signals 5.0.0, that actually happens already after version 3.0.0 when using this benchmark. Well at least with version 4.5.1 it does, I did not test with all 4.x.x releases.

Additionally, for this particular benchmark, Signal version 5.0.0 performs poorer than before and is in this benchmark clearly the weakest signals performer. This is of course just one benchmark, but so far the only one anybody has bothered to write, so I'm using it! 😄

I did not yet create a PR of my updated findings and test results to the main repo, but I'm about to. I figured maybe @rodydavis wants to review it and comment (fix?) before I submit a PR or publish test results any further, e.g. on Twitter.

Check the readme in the fork here for more info: https://github.com/rydmike/rainbench?tab=readme-ov-file#updated-test-runs-april-13-2024

cc: @jinyus @nank1ro

[jinyus]

Re: Not being able to run the test twice

This is because it disposes the signal after the test and disposed signals cannot be reused...So that's the expected behavior. Creating a new signal in the dispose method should fix that(same done for state_beacon and value notifier)

[rydmike]

Gotcha, fixed it and removed the note about this from the readme. Submitted a PR with updated results, in case you are interested in them.

[rodydavis]

Yeah will test with the benchmark and fix anything if needed