Skip to content
Open
Show file tree
Hide file tree
Changes from 9 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
580 changes: 352 additions & 228 deletions include/xsf/mathieu.h

Large diffs are not rendered by default.

39 changes: 39 additions & 0 deletions include/xsf/mathieu/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,39 @@
This is an implementation of the Mathieu fcns in C/C++. The
implementation follows the prototype algos created in Matlab and
maintained on GitHub at
https://github.com/brorson/MathieuFcnsFourier. This impl is a
header-only library for compatability with Scipy's xsf library.

The following Mathieu fcns are implemented:

* Angular fcn ce(n,q,v)
* Angular fcn se(n,q,v)
* Radial (modified) fcn of first kind mc1(n,q,u)
* Radial (modified) fcn of first kind ms1(n,q,u)
* Radial (modified) fcn of second kind mc2(n,q,u)
* Radial (modified) fcn of second kind ms2(n,q,u)

Here, n = fcn order, q = frequency (geometry) parmeter, v = angular
coord (radians), u = radial coord (au).

I also provide the following utility fcns:

* Eigenvalue a_n(q)
* Eigenvalue b_n(q)
* Fourier coeffs A_n^k(q) for ce fcns
* Fourier coeffs B_n^k(q) for se fcns

The goal is to provide a replacement of the Mathieu fcn suite used by
Scipy.

These programs may be built the usual way on a Linux system using the
usual GNU build tools. The main() function runs some simple sanity
checks on the functions. In particular, it verifies some output
values against those computed by the Matlab programs. I did a lot of
verification and accuracy testing on the Matlab implementations.
Therefore, tests run here just make sure the C implementation's
outputs match those from Matlab. The code in main() also shows how to
invoke the various fcns.

Summer 2025, SDB

84 changes: 84 additions & 0 deletions include/xsf/mathieu/besseljyd.h
Original file line number Diff line number Diff line change
@@ -0,0 +1,84 @@
#ifndef BESSELJYD_H
#define BESSELJYD_H

#include "../bessel.h"
#include "../config.h"

/*
*
* This is part of the Mathieu function suite -- a reimplementation
* of the Mathieu functions for Scipy. This file holds helpers
* to the Bessel J and Y functions and also returns derivatives
* of those fcns.
*
*/

namespace xsf {
namespace mathieu {

//==================================================================
double besselj(int k, double z) {
// This is just a thin wrapper around the Bessel impl in the
// std library.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
// std library.
// xsf library.

double v = (double)k;
return xsf::cyl_bessel_j(v, z);
}

//==================================================================
double bessely(int k, double z) {
// This is just a thin wrapper around the Bessel impl in the
// std library.
double v = (double)k;
return xsf::cyl_bessel_y(v, z);
}

//==================================================================
double besseljd(int k, double z) {
// This returns the derivative of besselj. The deriv is
// computed using common identities.
double y;

if (k == 0) {
double v = 1.0;
y = -besselj(v, z);
} else {
double kp1 = (double)(k + 1);
double km1 = (double)(k - 1);
y = (besselj(km1, z) - besselj(kp1, z)) / 2.0;
}

// Must flip sign for negative k and odd k.
if (k < 0 && ((k % 2) != 0)) {
y = -y;
}

return y;
}

//==================================================================
double besselyd(int k, double z) {
// This returns the derivative of besselj. The deriv is
// computed using common identities.
double y;

if (k == 0) {
double v = 1.0;
y = -bessely(v, z);
} else {
double kp1 = (double)(k + 1);
double km1 = (double)(k - 1);
y = (bessely(km1, z) - bessely(kp1, z)) / 2.0;
}

// Must flip sign for negative k and odd k.
if (k < 0 && ((k % 2) != 0)) {
y = -y;
}

return y;
}

} // namespace mathieu
} // namespace xsf

#endif // #ifndef BESSELJYD_H
198 changes: 198 additions & 0 deletions include/xsf/mathieu/make_matrix.h
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There's a backport of std::mdspan (from C++23) in kokkos/mdspan.hpp that sounds like a perfect fit for these functions. It's already used quite a lot in xsf, e.g. in legendre.h.

Original file line number Diff line number Diff line change
@@ -0,0 +1,198 @@
#ifndef MAKE_MATRIX_H
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

could a #pragma once help here?

#define MAKE_MATRIX_H

#include "../config.h"
#include "../error.h"
#include "matrix_utils.h"

/*
*
* This is part of the Mathieu function suite -- a reimplementation
* of the Mathieu functions for Scipy. This file holds the functions
* which make the recursion matrices.
*
* Stuart Brorson, Summer 2025.
*
*/

#define SQRT2 1.414213562373095
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There's a M_SQRT2 in config.h you could use instead


namespace xsf {
namespace mathieu {

/*-----------------------------------------------
This creates the recurrence relation matrix for
the even-even Mathieu fcns (ce_2n).
Inputs:
N = matrix size (related to max order desired).
q = shape parameter.
Output:
A = recurrence matrix (must be calloc'ed in caller).
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It might help to document the size constraints. Because some naive user 1 might try to pass double[1] for A.

Footnotes

  1. ... such as myself; since I actually made this mistake a couple of days ago, and debugging it took way too long haha

Return:
return code = 0 if OK.
-------------------------------------------------*/
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nitpick: I think it would a bit easier to read if it would be a bit wider, so that there aren't as many linebreaks needed.

int make_matrix_ee(int N, double q, double *A) {
int j;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nitpick: by moving this closer to the where it's used, it's slightly easier to read this (as human), as it requires less "mental ram".

int i;

// Symmetrize matrix here, then fix in caller.
i = MATRIX_IDX(N, 0, 1);
A[i] = SQRT2 * q;
i = MATRIX_IDX(N, 1, 0);
A[i] = SQRT2 * q;
i = MATRIX_IDX(N, 1, 1);
A[i] = 4.0;
i = MATRIX_IDX(N, 1, 2);
A[i] = q;

for (j = 2; j <= N - 2; j++) {
i = MATRIX_IDX(N, j, j - 1);
A[i] = q;
i = MATRIX_IDX(N, j, j);
A[i] = (2.0 * j) * (2.0 * j);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I kinda like the symmetry here, but it took me a couple of seconds to interpret this (mental arithmetic + double checking that the parentheses are indeed redundant here), so I think it would help (human) readers if this would simplified as something like 4.0 * j * j

i = MATRIX_IDX(N, j, j + 1);
A[i] = q;
}

i = MATRIX_IDX(N, N - 1, N - 2);
A[i] = q;
i = MATRIX_IDX(N, N - 1, N - 1);
A[i] = (2.0 * (N - 1)) * (2.0 * (N - 1));
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

same nitpick as above


return 0;
}

/*-----------------------------------------------
This creates the recurrence relation matrix for
the even-odd Mathieu fcns (ce_2n+1).
Inputs:
N = matrix size (related to max order desired).
q = shape parameter.
Output:
A = recurrence matrix (calloc in caller).
Return:
return code = 0 if OK.
-------------------------------------------------*/
int make_matrix_eo(int N, double q, double *A) {
int j;
int i;

i = MATRIX_IDX(N, 0, 0);
A[i] = 1.0 + q;
i = MATRIX_IDX(N, 0, 1);
A[i] = q;
i = MATRIX_IDX(N, 1, 0);
A[i] = q;
i = MATRIX_IDX(N, 1, 1);
A[i] = 9.0;
i = MATRIX_IDX(N, 1, 2);
A[i] = q;

for (j = 2; j <= N - 2; j++) {
i = MATRIX_IDX(N, j, j - 1);
A[i] = q;
i = MATRIX_IDX(N, j, j);
A[i] = (2.0 * j + 1.0) * (2.0 * j + 1.0);
i = MATRIX_IDX(N, j, j + 1);
A[i] = q;
}

i = MATRIX_IDX(N, N - 1, N - 2);
A[i] = q;
i = MATRIX_IDX(N, N - 1, N - 1);
A[i] = (2.0 * (N - 1) + 1.0) * (2.0 * (N - 1) + 1.0);

return 0;
}

/*-----------------------------------------------
This creates the recurrence relation matrix for
the odd-even Mathieu fcns (se_2n) -- sometimes called
se_2n+2.
Inputs:
N = matrix size (related to max order desired).
q = shape parameter.
Output:
A = recurrence matrix (calloc in caller).
Return:
return code = 0 if OK.
-------------------------------------------------*/
int make_matrix_oe(int N, double q, double *A) {
int j;
int i;

i = MATRIX_IDX(N, 0, 0);
A[i] = 4.0;
i = MATRIX_IDX(N, 0, 1);
A[i] = q;
i = MATRIX_IDX(N, 1, 0);
A[i] = q;
i = MATRIX_IDX(N, 1, 1);
A[i] = 16.0;
i = MATRIX_IDX(N, 1, 2);
A[i] = q;

for (j = 2; j <= N - 2; j++) {
i = MATRIX_IDX(N, j, j - 1);
A[i] = q;
i = MATRIX_IDX(N, j, j);
A[i] = (2.0 * (j + 1)) * (2.0 * (j + 1));
i = MATRIX_IDX(N, j, j + 1);
A[i] = q;
}

i = MATRIX_IDX(N, N - 1, N - 2);
A[i] = q;
i = MATRIX_IDX(N, N - 1, N - 1);
A[i] = (2.0 * N) * (2.0 * N);

return 0;
}

/*-----------------------------------------------
This creates the recurrence relation matrix for
the odd-odd Mathieu fcns (se_2n+1).
Inputs:
N = matrix size (related to max order desired).
q = shape parameter.
Output:
A = recurrence matrix (calloc in caller).
Return:
return code = 0 if OK.
-------------------------------------------------*/
int make_matrix_oo(int N, double q, double *A) {
int j;
int i;

i = MATRIX_IDX(N, 0, 0);
A[i] = 1.0 - q;
i = MATRIX_IDX(N, 0, 1);
A[i] = q;
i = MATRIX_IDX(N, 1, 0);
A[i] = q;
i = MATRIX_IDX(N, 1, 1);
A[i] = 9.0;
i = MATRIX_IDX(N, 1, 2);
A[i] = q;

for (j = 2; j <= N - 2; j++) {
i = MATRIX_IDX(N, j, j - 1);
A[i] = q;
i = MATRIX_IDX(N, j, j);
A[i] = (2.0 * j + 1.0) * (2.0 * j + 1.0);
i = MATRIX_IDX(N, j, j + 1);
A[i] = q;
}

i = MATRIX_IDX(N, N - 1, N - 2);
A[i] = q;
i = MATRIX_IDX(N, N - 1, N - 1);
A[i] = (2.0 * N - 1.0) * (2.0 * N - 1.0);

return 0;
}

} // namespace mathieu
} // namespace xsf

#endif // #ifndef MAKE_MATRIX_H
Loading