[llvm] LangRef: Clarify behaviors of nsz in fast math flag (PR #137567)

[Andy Kaylor via llvm-commits]

andykaylor wrote:

> The problem we are trying to solve is to say something more specific than this. I assert that if we cannot get the IEEE-754 2008 fuzzy minnum/maxnum behavior by taking a stronger intrinsic with ordered 0 handling, and adding nsz to it, the definition of nsz is broken

Trying to get IEEE-754 semantics by *adding* fast-math flags seems wrong to me. If we want IEEE-754 2008 semantics for an intrinsic, why don't we just make an intrinsic with the semantics we want.

https://github.com/llvm/llvm-project/pull/137567