Ah, sorry for that I must have seen the updated one.

I agree with you that switching programs that are all using the IEEE floating-point formats with the same underlying hardware implementations should result in the same answers.