math/doc/sf/gegenbauer.qbk

112 lines
3.3 KiB
Plaintext

[/
Copyright 2019, Nick Thompson
Distributed under the Boost Software License, Version 1.0.
(See accompanying file LICENSE_1_0.txt or copy at
http://www.boost.org/LICENSE_1_0.txt).
]
[section:gegenbauer Gegenbauer Polynomials]
[h4 Synopsis]
``
#include <boost/math/special_functions/gegenbauer.hpp>
``
namespace boost{ namespace math{
template<typename Real>
Real gegenbauer(unsigned n, Real lambda, Real x);
template<typename Real>
Real gegenbauer_prime(unsigned n, Real lambda, Real x);
template<typename Real>
Real gegenbauer_derivative(unsigned n, Real lambda, Real x, unsigned k);
}} // namespaces
Gegenbauer polynomials are a family of orthogonal polynomials.
A basic usage is as follows:
using boost::math::gegenbauer;
double x = 0.5;
double lambda = 0.5;
unsigned n = 3;
double y = gegenbauer(n, lambda, x);
All derivatives of the Gegenbauer polynomials are available.
The /k/-th derivative of the /n/-th Gegenbauer polynomial is given by
using boost::math::gegenbauer_derivative;
double x = 0.5;
double lambda = 0.5;
unsigned n = 3;
unsigned k = 2;
double y = gegenbauer_derivative(n, lambda, x, k);
For consistency with the rest of the library, `gegenbauer_prime` is provided which simply returns `gegenbauer_derivative(n, lambda, x,1 )`.
[$../graphs/gegenbauer.svg]
[h3 Implementation]
The implementation uses the 3-term recurrence for the Gegenbauer polynomials, rising.
[h3 Performance]
Double precision timing on a consumer x86 laptop is shown below.
Included is the time to generate a random number argument in the interval \[-1, 1\] (which takes 11.5ns).
``
Run on (16 X 4300 MHz CPU s)
CPU Caches:
L1 Data 32K (x8)
L1 Instruction 32K (x8)
L2 Unified 1024K (x8)
L3 Unified 11264K (x1)
Load Average: 0.21, 0.33, 0.29
-----------------------------------------
Benchmark Time
-----------------------------------------
Gegenbauer<double>/1 12.5 ns
Gegenbauer<double>/2 13.5 ns
Gegenbauer<double>/3 14.6 ns
Gegenbauer<double>/4 16.0 ns
Gegenbauer<double>/5 17.5 ns
Gegenbauer<double>/6 19.2 ns
Gegenbauer<double>/7 20.7 ns
Gegenbauer<double>/8 22.2 ns
Gegenbauer<double>/9 23.6 ns
Gegenbauer<double>/10 25.2 ns
Gegenbauer<double>/11 26.9 ns
Gegenbauer<double>/12 28.7 ns
Gegenbauer<double>/13 30.5 ns
Gegenbauer<double>/14 32.5 ns
Gegenbauer<double>/15 34.3 ns
Gegenbauer<double>/16 36.3 ns
Gegenbauer<double>/17 38.0 ns
Gegenbauer<double>/18 39.9 ns
Gegenbauer<double>/19 41.8 ns
Gegenbauer<double>/20 43.8 ns
UniformReal<double> 11.5 ns
``
[h3 Accuracy]
Some representative ULP plots are shown below.
The relative accuracy cannot be controlled at the roots of the polynomial, as is to be expected.
[$../graphs/gegenbauer_ulp_3.svg]
[$../graphs/gegenbauer_ulp_5.svg]
[$../graphs/gegenbauer_ulp_9.svg]
[h3 Caveats]
Some programs define the Gegenbauer polynomial with \u03BB = 0 via renormalization (which makes them Chebyshev polynomials).
We do not follow this convention: In this case, only the zeroth Gegenbauer polynomial is nonzero.
[endsect]