Potential to add #[inline} attributes where possible

[shssoichiro]

Potential to add #[inline} attributes where possible

[Amanieu]

If you've found a specific case where the lack of #[inline] causes a performance problem for you then we are happy to accept a PR to add the #[inline] back.

[nia-e]

I benchmarked every function with #[inline] vs. as-is, and there are a few considerable improvements. On my machine (x86_64 Arch Linux, Ryzen 7 5800H) I got these significant results - though I'd appreciate someone on a different system/architecture running a bench as well:

• atan: 3 -> 2 ns/iter
• atan2: 6 -> 5 ns/iter
• cosf: 4 -> 2 ns/iter
• coshf: 4 -> 3 ns/iter
• expm1f: 3 -> 2 ns/iter
• hypot: 4 -> 2 ns/iter
• hypotf: 2 -> 1 ns/iter
• sinf: 5 -> 2 ns/iter
• tan: 4 -> 3 ns/iter
• tanf: 7 -> 2 ns/iter
• tanh: 8 -> 6 ns/iter

I only included results where #[inline] helped by >10% but I will do more runs and average out results before doing any proper PR, for now just want to put this out there