Add device only impl

Add function for device only impl

Fix function signatures

Fix arrays

Add basic test for compilation

Allow serial implementation to be run on host under NVCC

Add verification steps

Add arrays of levels coefficient sizes

Cleanup test set

Add double test set

Add structure for the doubles support

Save space by using pointer to different size arrays rather than 2d

Separate the double precision weights into their own arrays

Remove stray call to std::abs

Add NVRTC testing

Add documentation section

Add device function signature for sinh_sinh_integrate

Add float coefficients

Add double coeffs

Add device specific impl

Add sinh_sinh CUDA testing

Add sinh_sinh NVRTC testing

doc/quadrature/double_exponential.qbk

@@ -1,5 +1,6 @@
 [/
 Copyright (c) 2017 Nick Thompson
+Copyright (c) 2024 Matt Borland
 Use, modification and distribution are subject to 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)
@@ -538,6 +539,30 @@ This form integrates just fine over (-log([pi]/2), +[infin]) using either the `t
 
 [endsect] [/section:de_caveats Caveats]
 
+[section:gpu_usage GPU Usage]
+
+``
+    #include <boost/math/quadrature/exp_sinh.hpp>
+
+    namespace boost{ namespace math{ namespace quadrature {
+
+    template <class F, class Real, class Policy = policies::policy<> >
+    __device__ auto exp_sinh_integrate(const F& f, Real a, Real b, Real tolerance, Real* error, Real* L1, boost::math::size_t* levels)
+
+    template <class F, class Real, class Policy = policies::policy<> >
+    __device__ auto exp_sinh_integrate(const F& f, Real tolerance, Real* error, Real* L1, boost::math::size_t* levels)
+
+}}}
+``
+
+Quadrature is additionally able to run on CUDA (NVCC and NVRTC) platforms.
+The major difference is outlined in the above function signatures.
+When used on device these are free standing functions instead of using OOP like on the host.
+The tables of abscissas and weights are stored in shared read only memory on the device instead of being initialized when the class is constructed.
+An example use case would be in the finite elements method computing a stiffness matrix since it would consist of many different functions.
+
+[endsect] [/section:gpu_usage Usage]
+
 [section:de_refes References]
 
 * Hidetosi Takahasi and Masatake Mori, ['Double Exponential Formulas for Numerical Integration] Publ. Res. Inst. Math. Sci., 9 (1974), pp. 721-741.