I am looking for a nice reference (a review, tutorial, or maybe a book) that has tips and their explanations about general issues of floating-point arithmetic in scientific computations.
Some that I've seen are Numerical Recipes and Goldberg, D. E. (1991). What every computer scientist should know about floating-point arithmetic. ACM Computing Surveys, 23(1), 5.
Numerical Recipes is a very large book with specific algorithms. I am looking for general rules of thumb, of general applicability.
The article by Goldberg has a bit too much emphasis on multiple hardware implementations. I really only care about C/C++ programming on modern PCs, which has a pretty standard way of dealing with floats.