Test for (IN) equality, subtraction, proof of correctness

ABSTRACT

As normally implemented, (in)equality tests such as "a < b" can fail to execute correctly and thus algorithms that have been "proven" may fail.

REFERENCES

• 1: Richard P. Brent, A Fortran Multiple-Precision Arithmetic Package, ACM Transactions on Mathematical Software (TOMS), v.4 n.1, p.57-70, March 1978 [doi>10.1145/355769.355775]
• 2: J. Demmel, Underflow and the reliability of numerical software, SIAM J. Sci. Stat. Comput. <u>5</u> (1984), 887--919.
• 3: C. Dunham, Floating point with rounding before normalization, Congressus Numerantium <u>46</u> (1985), (Proceedings 14th Manitoba Conf. on Numerical Math. and Computing), 91-102. Reprints available.
• 4: Robert L. Johnston, Numerical Methods: A Software Approach, John Wiley & Sons, Inc., New York, NY, 1982
• 5: P. Naur, Proof of algorithms by general snapshots, BIT <u>6</u> (1966), 310--316.
• 6: F. W. J. Olver, Error bounds for arithmetic operations on computers without guard digits, IMA J. Numer. Anal. <u>3</u> (1983), 153--160.
• 7: P. Sterbenz, Floating-point Computation, Prentice-Hall, 1974.
• 8: James H. Wilkinson, Rounding Errors in Algebraic Processes, Dover Publications, Incorporated, 1994
• 9: W. Kahan, A survey of error analysis, in Information Processing 71, North Holland, 1214--1239.
• 10: W. Kahan, A proposed standard for binary floating-point arithmetic. Computer <u>14</u> (March 1981), 51--62.