[LangRef] make consequences of NaN rules for pow(i) more explicit

Author: RalfJung

llvm/docs/LangRef.rst

@@ -16269,6 +16269,12 @@ Semantics:
 This function returns the first value raised to the second power with an
 unspecified sequence of rounding operations.
 
+Note that the `powi` function is unusual in that NaN inputs can lead to non-NaN
+results, and this depends on the kind of NaN (quiet vs signaling). Due to how
+:ref:`LLVM treats NaN values <floatnan>`, the function may non-deterministically
+treat signaling NaNs as quiet NaNs. For example, `powi(QNaN, 0)` returns `1.0`,
+and `pow(SNaN, 0)` may non-deterministically return `1.0` or a NaN.
+
 .. _t_llvm_sin:
 
 '``llvm.sin.*``' Intrinsic
@@ -16831,6 +16837,12 @@ trapping or setting ``errno``.
 When specified with the fast-math-flag 'afn', the result may be approximated
 using a less accurate calculation.
 
+Note that the `pow` function is unusual in that NaN inputs can lead to non-NaN
+results, and this depends on the kind of NaN (quiet vs signaling). Due to how
+:ref:`LLVM treats NaN values <floatnan>`, the function may non-deterministically
+treat signaling NaNs as quiet NaNs. For example, `pow(QNaN, 0.0)` returns `1.0`,
+and `pow(SNaN, 0.0)` may non-deterministically return `1.0` or a NaN.
+
 .. _int_exp:
 
 '``llvm.exp.*``' Intrinsic