Skip to content

Commit

Permalink
Add GPU support to airy functions
Browse files Browse the repository at this point in the history 
Add SYCL testing of ariy functions

Add CUDA testing of airy functions

Add NVRTC testing of airy functions

Add GPU support to ellint rc

Add GPU support to ellint rd

Add GPU support to ellint rf

Add GPU support to ellint rg

Add GPU support to ellint rj

Add GPU support to ellint d

Add GPU support to ellint_1

Markup forward and add ellint_3 return type def for NVRTC platform

Add CUDA testing of ellint 1

NVRTC fixes

Add NVRTC testing of ellint_1

Add GPU support to ellint_2

Add CUDA testing of ellint_2

Fix NVRTC errors

Add NVRTC testing of ellint_2

Add GPU support to atanh

Add GPU support to ellint_3

Add NVRTC testing of ellint_3

Add CUDA testing of ellint_3

Replace use of static const char*

Add SYCL testing of ellint_1

Add SYCL testing of ellint 2 with slight tolerance bump

Remove recursion from ellint_rj

Add ellint_d CUDA testing

Add NVRTC testing of ellint_d

Add SYCL testing of ellint_d

Remove SYCL ellint_3 support

Update docs

Add GPU support to jacobi zeta

Add CUDA testing of jacobi zeta

Add NVRTC testing of jacobi zeta

Add SYCL testing of jacobi zeta

Add GPU support to heuman_lambda

Add NVRTC testing of heuman lambda

Add CUDA testing of heuman_lambda

Add SYCL testing of heuman lambda

Add markers to docs

Add marker for CUDA only functions in the docs
  • Loading branch information
@mborland
mborland committed Sep 16, 2024
1 parent ff72eaa commit 1324401
Show file tree
Hide file tree
Showing 72 changed files with 6,436 additions and 336 deletions.
5 changes: 3 additions & 2 deletions doc/overview/gpu.qbk
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,8 +3,9 @@
[h4 GPU Support]

Selected functions, distributions, tools, etc. support running on both host and devices.
These functions will have the annotation `BOOST_MATH_GPU_ENABLED` next to their individual documentation.
We test using CUDA (both NVCC and NVRTC) as well as SYCL to provide a wide range of support.
These functions will have the annotation `BOOST_MATH_GPU_ENABLED` or `BOOST_MATH_CUDA_ENABLED` next to their individual documentation.
Functions marked with `BOOST_MATH_GPU_ENABLED` are tested using CUDA (both NVCC and NVRTC) as well as SYCL to provide a wide range of support.
Functions marked with `BOOST_MATH_CUDA_ENABLED` are few, but due to its restrictions SYCL is unsupported.

[h4 Policies]

Expand Down
40 changes: 20 additions & 20 deletions doc/sf/ellint_carlson.qbk
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,10 +17,10 @@ LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
namespace boost { namespace math {

template <class T1, class T2, class T3>
``__sf_result`` ellint_rf(T1 x, T2 y, T3 z)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rf(T1 x, T2 y, T3 z)

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_rf(T1 x, T2 y, T3 z, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rf(T1 x, T2 y, T3 z, const ``__Policy``&)

}} // namespaces

Expand All @@ -32,10 +32,10 @@ LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
namespace boost { namespace math {

template <class T1, class T2, class T3>
``__sf_result`` ellint_rd(T1 x, T2 y, T3 z)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rd(T1 x, T2 y, T3 z)

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_rd(T1 x, T2 y, T3 z, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rd(T1 x, T2 y, T3 z, const ``__Policy``&)

}} // namespaces

Expand All @@ -47,10 +47,10 @@ LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
namespace boost { namespace math {

template <class T1, class T2, class T3, class T4>
``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p)

template <class T1, class T2, class T3, class T4, class ``__Policy``>
``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p, const ``__Policy``&)

}} // namespaces

Expand All @@ -62,10 +62,10 @@ LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
namespace boost { namespace math {

template <class T1, class T2>
``__sf_result`` ellint_rc(T1 x, T2 y)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rc(T1 x, T2 y)

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_rc(T1 x, T2 y, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rc(T1 x, T2 y, const ``__Policy``&)

}} // namespaces

Expand All @@ -76,10 +76,10 @@ LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
namespace boost { namespace math {

template <class T1, class T2, class T3>
``__sf_result`` ellint_rg(T1 x, T2 y, T3 z)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rg(T1 x, T2 y, T3 z)

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_rg(T1 x, T2 y, T3 z, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rg(T1 x, T2 y, T3 z, const ``__Policy``&)

}} // namespaces

Expand All @@ -98,10 +98,10 @@ when the arguments are of different types: otherwise the return is the same type
as the arguments.

template <class T1, class T2, class T3>
``__sf_result`` ellint_rf(T1 x, T2 y, T3 z)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rf(T1 x, T2 y, T3 z)

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_rf(T1 x, T2 y, T3 z, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rf(T1 x, T2 y, T3 z, const ``__Policy``&)

Returns Carlson's Elliptic Integral ['R[sub F]]:

Expand All @@ -113,10 +113,10 @@ one may be zero. Otherwise returns the result of __domain_error.
[optional_policy]

template <class T1, class T2, class T3>
``__sf_result`` ellint_rd(T1 x, T2 y, T3 z)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rd(T1 x, T2 y, T3 z)

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_rd(T1 x, T2 y, T3 z, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rd(T1 x, T2 y, T3 z, const ``__Policy``&)

Returns Carlson's elliptic integral R[sub D]:

Expand All @@ -128,10 +128,10 @@ zero, and that z >= 0. Otherwise returns the result of __domain_error.
[optional_policy]

template <class T1, class T2, class T3, class T4>
``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p)

template <class T1, class T2, class T3, class T4, class ``__Policy``>
``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rj(T1 x, T2 y, T3 z, T4 p, const ``__Policy``&)

Returns Carlson's elliptic integral R[sub J]:

Expand All @@ -149,10 +149,10 @@ using the relation:
[equation ellint17]

template <class T1, class T2>
``__sf_result`` ellint_rc(T1 x, T2 y)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rc(T1 x, T2 y)

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_rc(T1 x, T2 y, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rc(T1 x, T2 y, const ``__Policy``&)

Returns Carlson's elliptic integral R[sub C]:

Expand All @@ -170,10 +170,10 @@ using the relation:
[equation ellint18]

template <class T1, class T2, class T3>
``__sf_result`` ellint_rg(T1 x, T2 y, T3 z)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rg(T1 x, T2 y, T3 z)

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_rg(T1 x, T2 y, T3 z, const ``__Policy``&)
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_rg(T1 x, T2 y, T3 z, const ``__Policy``&)

Returns Carlson's elliptic integral ['R[sub G]:]

Expand Down
72 changes: 36 additions & 36 deletions doc/sf/ellint_legendre.qbk
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,16 +17,16 @@ LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
namespace boost { namespace math {

template <class T1, class T2>
``__sf_result`` ellint_1(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_1(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T1 k, T2 phi, const ``__Policy``&);

template <class T>
``__sf_result`` ellint_1(T k);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T k);

template <class T, class ``__Policy``>
``__sf_result`` ellint_1(T k, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T k, const ``__Policy``&);

}} // namespaces

Expand All @@ -42,10 +42,10 @@ when T1 and T2 are different types: when they are the same type then the result
is the same type as the arguments.

template <class T1, class T2>
``__sf_result`` ellint_1(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_1(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T1 k, T2 phi, const ``__Policy``&);

Returns the incomplete elliptic integral of the first kind ['F([phi], k)]:

Expand All @@ -56,10 +56,10 @@ Requires k[super 2]sin[super 2](phi) < 1, otherwise returns the result of __doma
[optional_policy]

template <class T>
``__sf_result`` ellint_1(T k);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T k);

template <class T>
``__sf_result`` ellint_1(T k, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_1(T k, const ``__Policy``&);

Returns the complete elliptic integral of the first kind ['K(k)]:

Expand Down Expand Up @@ -123,16 +123,16 @@ and
namespace boost { namespace math {

template <class T1, class T2>
``__sf_result`` ellint_2(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_2(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T1 k, T2 phi, const ``__Policy``&);

template <class T>
``__sf_result`` ellint_2(T k);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T k);

template <class T, class ``__Policy``>
``__sf_result`` ellint_2(T k, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T k, const ``__Policy``&);

}} // namespaces

Expand All @@ -148,10 +148,10 @@ when T1 and T2 are different types: when they are the same type then the result
is the same type as the arguments.

template <class T1, class T2>
``__sf_result`` ellint_2(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_2(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T1 k, T2 phi, const ``__Policy``&);

Returns the incomplete elliptic integral of the second kind ['E([phi], k)]:

Expand All @@ -162,10 +162,10 @@ Requires k[super 2]sin[super 2](phi) < 1, otherwise returns the result of __doma
[optional_policy]

template <class T>
``__sf_result`` ellint_2(T k);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T k);

template <class T>
``__sf_result`` ellint_2(T k, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_2(T k, const ``__Policy``&);

Returns the complete elliptic integral of the second kind ['E(k)]:

Expand Down Expand Up @@ -230,16 +230,16 @@ and
namespace boost { namespace math {

template <class T1, class T2, class T3>
``__sf_result`` ellint_3(T1 k, T2 n, T3 phi);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n, T3 phi);

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_3(T1 k, T2 n, T3 phi, const ``__Policy``&);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n, T3 phi, const ``__Policy``&);

template <class T1, class T2>
``__sf_result`` ellint_3(T1 k, T2 n);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_3(T1 k, T2 n, const ``__Policy``&);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n, const ``__Policy``&);

}} // namespaces

Expand All @@ -255,10 +255,10 @@ when the arguments are of different types: when they are the same type then the
is the same type as the arguments.

template <class T1, class T2, class T3>
``__sf_result`` ellint_3(T1 k, T2 n, T3 phi);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n, T3 phi);

template <class T1, class T2, class T3, class ``__Policy``>
``__sf_result`` ellint_3(T1 k, T2 n, T3 phi, const ``__Policy``&);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n, T3 phi, const ``__Policy``&);

Returns the incomplete elliptic integral of the third kind ['[Pi](n, [phi], k)]:

Expand All @@ -271,10 +271,10 @@ would be complex).
[optional_policy]

template <class T1, class T2>
``__sf_result`` ellint_3(T1 k, T2 n);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_3(T1 k, T2 n, const ``__Policy``&);
BOOST_MATH_CUDA_ENABLED ``__sf_result`` ellint_3(T1 k, T2 n, const ``__Policy``&);

Returns the complete elliptic integral of the first kind ['[Pi](n, k)]:

Expand Down Expand Up @@ -355,16 +355,16 @@ and
namespace boost { namespace math {

template <class T1, class T2>
``__sf_result`` ellint_d(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_d(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_d(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_d(T1 k, T2 phi, const ``__Policy``&);

template <class T1>
``__sf_result`` ellint_d(T1 k);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_d(T1 k);

template <class T1, class ``__Policy``>
``__sf_result`` ellint_d(T1 k, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_d(T1 k, const ``__Policy``&);

}} // namespaces

Expand All @@ -378,10 +378,10 @@ when the arguments are of different types: when they are the same type then the
is the same type as the arguments.

template <class T1, class T2>
``__sf_result`` ellint_d(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_d(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` ellint_3(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_3(T1 k, T2 phi, const ``__Policy``&);

Returns the incomplete elliptic integral:

Expand All @@ -394,10 +394,10 @@ would be complex).
[optional_policy]

template <class T1>
``__sf_result`` ellint_d(T1 k);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_d(T1 k);

template <class T1, class ``__Policy``>
``__sf_result`` ellint_d(T1 k, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` ellint_d(T1 k, const ``__Policy``&);

Returns the complete elliptic integral ['D(k) = D([pi]/2, k)]

Expand Down Expand Up @@ -463,10 +463,10 @@ using the relation:
namespace boost { namespace math {

template <class T1, class T2>
``__sf_result`` jacobi_zeta(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` jacobi_zeta(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` jacobi_zeta(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` jacobi_zeta(T1 k, T2 phi, const ``__Policy``&);

}} // namespaces

Expand Down Expand Up @@ -543,10 +543,10 @@ is [@../../example/jacobi_zeta_example.cpp jacobi_zeta_example.cpp].
namespace boost { namespace math {

template <class T1, class T2>
``__sf_result`` heuman_lambda(T1 k, T2 phi);
BOOST_MATH_GPU_ENABLED ``__sf_result`` heuman_lambda(T1 k, T2 phi);

template <class T1, class T2, class ``__Policy``>
``__sf_result`` heuman_lambda(T1 k, T2 phi, const ``__Policy``&);
BOOST_MATH_GPU_ENABLED ``__sf_result`` heuman_lambda(T1 k, T2 phi, const ``__Policy``&);

}} // namespaces

Expand Down
Loading

0 comments on commit 1324401

Please sign in to comment.