Remark on algorithm 723
Concerning evaluation of Fresnel Cosine and Sine integrals \(C(x)\) and \(S(x)\) for \(x>1.6\), the last sentence in the description of Algorithm 723 by the author [(*) ibid. 19, No. 4, 452-456 (1993; reviewed above)] states ``Extensive accuracy testing would simply have validated the trigonometric function routines. \textit{A. H. Morris} jun. [NSWC library of mathematics subroutines, Tech. Rep. NSWCDD/TR-92/425, Naval Surface Warfare Center, Dahlgren, Va. Jan. (1993)] has correctly pointed out that the procedures described in (*) to compute \(C(x)\) and \(S(x)\) for \(x>1.6\) require computing \(\sin\pi x^2/2\) and \(\cos\pi x^2/2\). Thus the quoted statement begs the question whether one might compute these trigonometric functions more accurately than can be done by naive application of the Fortran intrinsic trigonometric function routines, when \(x\) is large. The answer is in the affirmative. By separating \(x\) into its integer and fractional parts one can compute these functions with substantially better accuracy.
- Rational approximations, software and test methods for sine and cosine integrals
- Evaluation of Fresnel integrals based on the continued fractions method
- Algorithm 723: Fresnel integrals
- Computing Fresnel integrals via modified trapezium rules
- Corrigendum to: ``Remark on Algorithm 723: Fresnel integrals
- Algorithm for the Computation of the Generalized Fresnel Integral
- Calculation of Fresnel integrals of real and complex arguments up to 28 significant digits
This page was built for publication: Remark on algorithm 723
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4371118)
