David Goldberg. What every computer scientist should know about floating-point arithmetic. ACM Computing Surveys, 23(1):5--48, March 1991.
Not directly language related, but still quite relevant.
Most programming languages support some form of floating point artihmetic. Some languages have a well defined model, others simply provide a floating point type, but don't give formal semantics and don't specify a numeric model.
Aside from exploring the classic issues related to floating point arithmetic, this paper also talks about language implementation details (e.g., how optimizations affect numeric accuracy).
Posted to general by Ehud Lamm on 10/31/02; 3:27:10 PM