Skip to main content

Desperately Needed Remedies for the Undebuggability of Large Floating-Point Computations in Science and Engineering (part 1)

William Kahan
How long does it take to either allay or confirm suspicions, should they arise, about the accuracy of a computed result? Often diagnosis has been overtaken by the end of a computing platform's service life. Diagnosis could be sped up by at least an order of magnitude if more users and developers of numerical software knew enough to demand the needed software tools. Almost all these have existed though not all of them together in one place at one time. These tools cope with vulnerabilities peculiar to Floating-Point, namely roundoff and arithmetic exceptions. Programming languages tend to turn exceptions into branches which are prone to error. In particular, unanticipated events deemed ERRORs are handled in obsolete ways inherited from the era of batch computing. There are better ways. They would have prevented the crash of Air France #447 in June 2009, among other things.

Speaker bio

Prof. W. Kahan (now retired) error-analyzes scientific and engineering floating-point computations on electronic computers, which he has programmed since 1953. Born and educated in Toronto, Canada, he spent two Post-doctoral years 1958-60 at Cambridge, England, before returning to teach at the Univ. of Toronto until 1969, when he moved to the Univ. of Calif. @ Berkeley. Among his handiwork: <> infallible algorithms for the Hewlett-Packard HP-12C financial calculator (still for sale since 1982). <> fast and accurate singular-value decompositions (with Gene H. Golub in 1964) now used very widely. <> the mathematical foundation for the near-ubiquitous IEEE Standard 754 for Binary (& now Decimal) Floating-Point. Among his trophies: the ACM Turing Award (1989), S.I.A.M. Von Neumann Lecture (1997), & the IEEE Piore Award (2000).



Share this: