hrubin@pop.stat.purdue.edu (Herman Rubin) (06/25/91)
In article <1991Jun24.211407.17389@ingres.Ingres.COM>, jpk@Ingres.COM (Jon Krueger) writes: > jbs@WATSON.IBM.COM (James B. Shearer) writes: > > Do you disagree that it is undesirable for optimization to affect > > a program's behavior? > > Yes. > > It is a good and joyful thing, always and everywhere, for > compilers to expose broken programs. > > Correct programs can be written in available languages. > The meaning of such programs does not depend on whether > compilers apply permissible optimizations. It is almost possible to write absolutely correct programs, but is it worthwhile? It is easy to give examples where the slightest breathing at a program will give different answers. One is always going to have problems with architectural limits, such as number of bits of precision, exponent range, how overflow and underflow are handled, etc. Cancellation can be a problem, and sometimes a careful rearrangement of the calculation, supposedly violating optimization rules, can avoid the problem. Only by using obscure languages do we have the availability needed for even producing reasonable programs. -- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399 Phone: (317)494-6054 hrubin@l.cc.purdue.edu (Internet, bitnet) {purdue,pur-ee}!l.cc!hrubin(UUCP)