Algorithms for accurate, validated and fast polynomial evaluation
We survey a class of algorithms to evaluate polynomials with floating point coefficients and for computation performed with IEEE-754 floating point arithmetic. The principle is to apply, once or recursively, an error-free transformation of the polynomial evaluation with the Horner algorithm and to accurately sum the final decomposition. These compensated algorithms are as accurate as the Horner algorithm perforned inK times the working precision, forK an arbitrary positive integer. We prove this accuracy property with an a priori error analysis. We also provide validated dynamic bounds and apply these results to compute a faithfully rounded evaluation. These compensated algorithms are fast. We illustrate their practical efficiency with numerical experiments on significant environments. Comparing to existing alternatives theseK-times compensated algorithms are competitive forK up to 4, i.e., up to 212 mantissa bits.
Key wordspolynomial evaluation compensated algorithm floating-point arithmetic IEEE-754
Unable to display preview. Download preview PDF.
- High-Precision Software Directory. http://crd.lbl.gov/~dhbailey/mpdist.Google Scholar
- S. Graillat, P. Langlois and N. Louvet, Compensated Horner scheme. Algebraic and Numerical Algorithms and Computer-Assisted Proofs, B. Buchberger, S. Oishi, M. Plum and S.M. Rump (eds.), Dagstuhl Seminar Proceedings, No. 05391, Internationales Begenungs-und Forschungszentrum (IBFI), Schloss Dagstuhl, Germany, 2006.Google Scholar
- S. Graillat, P. Langlois and N. Louvet, Improving the compensated Horner scheme with a fused multiply and add. Proceedings of the 21st Annual ACM Symposium on Applied Computing, Vol. 2, Association for Computing Machinery, 2006, 1323–1327.Google Scholar
- Y. Hida, X.S. Li and D.H. Bailey, Quad-double arithmetic: Algorithms, implementation, and application. 15th IEEE Symposium on Computer Arithmetic, N. Burgess and L. Ciminiera (eds.), IEEE Computer Society, 2001, 155–162.Google Scholar
- C.M. Hoffmann, G. Park, J.-R. Simard and N.F. Stewart, Residual iteration, and accurate polynomial evaluation for shape-interrogation applications. Proceedings of the 9th ACM Symposium on Solid Modeling and Applications, 2004, 9–14.Google Scholar
- IEEE Standards Committee 754, IEEE Standard for Binary Floating-Point Arithmetic, ANSI/IEEE Standard 754–1985. Institute of Electrical and Electronics Engineers, Los Alamitos, CA, USA, 1985, Reprinted in SIGPLAN Notices,22 (1987), 9–25.Google Scholar
- P. Langlois and N. Louvet, How to ensure a faithful polynomial evaluation with the compensated Horner algorithm. 18th IEEE International Symposium on Computer Arithmetic, P. Kornerup and J.-M. Muller (eds.), IEEE Computer Society, 2007, 141–149.Google Scholar
- P. Langlois and N. Louvet, More instruction level parallelism explains the actual efficiency of compensated algorithms. Technical Report hal-00165020, DALI Research Project, HALCCSD 2007.Google Scholar
- P. Langlois and N. Louvet, Compensated Horner algorithm inK times the working precision. RNC-8, Real Numbers and Computer Conference, J. Brugera and M. Daumas (eds.), Santiago de Compostela, Spain, 2008.Google Scholar
- The MPFR Library (version 2.2.1). Available at http://www.mpfr.org.Google Scholar