Electronic Proceedings of the Seventh Annual International Conference on Technology in Collegiate MathematicsOrlando, Florida, November 17-20, 1994Paper C024Should We Be Concerned About Roundoff Error? |
Anthony P. LeclercDepartment of Computer Science The College of Charleston 66 George Street Charleston, SC 29424 USA Phone: (803) 953-5963 leclerc@cs.cofc.edu |
| Click to access this paper: |
ABSTRACT
Floating point computations, whether performed in a programming language, on a calculator, or on a computer mathematic system such as Derive, Macsyma, Maple, or Mathematica, are susceptible to roundoff error. Some might contend that, today, roundoff error is not a critical issue because of the extended precision floating point numbers available on many computers. However, roundoff error can be severe, even for innocuous-looking computations.This paper will consider what can be done to alleviate roundoff error. In addition, the following questions will be addressed:
- How can one know the severity of roundoff error for a given computation?
- Are there ``tools'' on systems such as Derive, Macsyma, Maple, or Mathematica to control the roundoff error problem?
- Do such roundoff error control mechanisms exist as extensions or libraries within certain programming languages?
- Is there an educational opportunity in teaching students about the
roundoff error issue or is this something the instructor should
shield from students?
Keyword(s): numerical methods, computer algebra systems