Resurrect compiler perf improvements PRs

[vzarytovskii]

It's a shame those didn't get merged, I think we should resurrect it, they look like great changes

#5360
#5307
#5278
#5112

Going to be working on those past .NET8 release (November+)

[charlesroddie]

This should be done instead: fsharp/fslang-suggestions#1280 . Doesn't matter whether it is opt-in form or opt-out form: anyone who cares about performance and/or internal consistency of the language will enable it. Last 3 linked PRs are overcomplicated since they don't bite the bullet. One of them even considers using reflection as an additional workaround to the problem. At runtime? Even if not, it still runs the risk and indicates the degree of hacks involved.

[vzarytovskii]

This should be done instead: fsharp/fslang-suggestions#1280 . Doesn't matter whether it is opt-in form or opt-out form: anyone who cares about performance and/or internal consistency of the language will enable it. Last 3 linked PRs are overcomplicated since they don't bite the bullet. One of them even considers using reflection as an additional workaround to the problem. At runtime? Even if not, it still runs the risk and indicates the degree of hacks involved.

They cover a bunch more, how equality improvements are achieved is an implementation detail. It will be decided during when work will be actually happening.

[vzarytovskii]

If anyone wants to experiment, feel free, I don't have interest in those anymore. I might reuse some parts for the fslib extended library I'm maintaining for my own projects.