[LLVMdev] Proposal for Poison Semantics

[John Regehr]

> Having said that I won't be surprised if fast-math flags suffer from
> issues similar to poison -- "Allow algebraically equivalent
> transformations that may dramatically change results in floating
> point" sounds somewhat vague to me.

The vaguenss should be expected since nobody seems to know what "gcc 
-ffast-math" or "clang -ffast-math" mean at the source level either.

John