Tracking issue for extra floating-point logarithm constants

[clarfonthey]

Specifically, LOG10_2 and LOG2_10. PR: #50539

[clarfonthey]

Has anyone used these constants at all and want to pitch in for stabilisation of these? These have landed for a while and I'm curious how many people are using them.

[kennytm]

+1 for stabilization. (Tried to see if log₂(10) exists and found that it is unstable 🙄.)

[clarfonthey]

(Mind providing the specific use case? Curious to see if people are using this for anything other than scientific computing.)

[kennytm]

@clarcharr to estimate the "−log₂(probability)" of collision of two values chosen randomly from a set, which will be used to derive the probability of collision of N values (≈ 1 − e^-pN2/2, p ≪ 1, N ≫ 1).

log₂ is chosen instead of natural log because the set size is often 2ⁿ (e.g. u32), but sometimes it is drawn from a set of n decimal digits which would then involve n·log₂(10).

[iliekturtles]

One hartley is = log2(10) shannons (bits). Currently the constant is duplicated in uom.

[clarfonthey]

I can't @ anyone from the libs team but do you think we'd be able to start an FCP for this?

[varkor]

cc @rust-lang/libs for a FCP.

[alexcrichton]

@rfcbot fcp merge

I believe the stabilization includes core::{f32, f64}::{LOG2_10, LOG10_2}.

[rfcbot]

Team member @alexcrichton has proposed to merge this. The next step is review by the rest of the tagged team members:

• @Amanieu
• @Kimundi
• @SimonSapin
• @alexcrichton
• @dtolnay
• @sfackler
• @withoutboats

No concerns currently listed.

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

See this document for info about what commands tagged team members can give me.