[moonchild]

https://github.com/rutgers-apl/The-RLIBM-Project/blob/main/l... has a lot less and claims to be correctly rounded.

[pclmulqdq]

For 32-bit float, yes. For double precision, not so much.