16126000227

Committed 07 Jul 2025 07:24PM UTC coverage: 22.684% (-49.4%) from 72.051%

Build # 16126000227

Build Type

Pull #2519

github

Committed by

web-flow

Commit Message

Merge bd753a3a3 into 624f14f20

Pull Request Pull Request #2519: tmp changes

Coverage Stats

889 of 9756 branches covered (9.11%)

Branch coverage included in aggregate %.

6317 of 22011 relevant lines covered (28.7%)

35.96 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

28.75

/dpnp/backend/extensions/vm/div.cpp

//*****************************************************************************
// Copyright (c) 2024-2025, Intel Corporation
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
// - Redistributions of source code must retain the above copyright notice,
//   this list of conditions and the following disclaimer.
// - Redistributions in binary form must reproduce the above copyright notice,
//   this list of conditions and the following disclaimer in the documentation
//   and/or other materials provided with the distribution.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
// THE POSSIBILITY OF SUCH DAMAGE.
//*****************************************************************************

#include <stdexcept>

#include <oneapi/mkl.hpp>
#include <sycl/sycl.hpp>

#include "dpctl4pybind11.hpp"

#include "common.hpp"
#include "div.hpp"

// include a local copy of elementwise common header from dpctl tensor:
// dpctl/tensor/libtensor/source/elementwise_functions/elementwise_functions.hpp
// TODO: replace by including dpctl header once available
#include "../elementwise_functions/elementwise_functions.hpp"

// dpctl tensor headers
#include "kernels/elementwise_functions/common.hpp"
#include "utils/type_dispatch.hpp"
#include "utils/type_utils.hpp"

namespace dpnp::extensions::vm
{
namespace py = pybind11;
namespace py_int = dpnp::extensions::py_internal;
namespace td_ns = dpctl::tensor::type_dispatch;

namespace impl
{
namespace ew_cmn_ns = dpctl::tensor::kernels::elementwise_common;
namespace mkl_vm = oneapi::mkl::vm; // OneMKL namespace with VM functions
namespace tu_ns = dpctl::tensor::type_utils;

/**
 * @brief A factory to define pairs of supported types for which
 * MKL VM library provides support in oneapi::mkl::vm::div<T> function.
 *
 * @tparam T Type of input vectors `a` and `b` and of result vector `y`.
 */
template <typename T1, typename T2>
struct OutputType
{
    using value_type = typename std::disjunction<
        td_ns::BinaryTypeMapResultEntry<T1,
                                        std::complex<double>,
                                        T2,
                                        std::complex<double>,
                                        std::complex<double>>,
        td_ns::BinaryTypeMapResultEntry<T1,
                                        std::complex<float>,
                                        T2,
                                        std::complex<float>,
                                        std::complex<float>>,
        td_ns::BinaryTypeMapResultEntry<T1, double, T2, double, double>,
        td_ns::BinaryTypeMapResultEntry<T1, float, T2, float, float>,
        td_ns::DefaultResultEntry<void>>::result_type;
};

template <typename T1, typename T2>
static sycl::event div_contig_impl(sycl::queue &exec_q,
                                   std::size_t in_n,
                                   const char *in_a,
                                   py::ssize_t a_offset,
                                   const char *in_b,
                                   py::ssize_t b_offset,
                                   char *out_y,
                                   py::ssize_t out_offset,
                                   const std::vector<sycl::event> &depends)
{
    tu_ns::validate_type_for_device<T1>(exec_q);
    tu_ns::validate_type_for_device<T2>(exec_q);

    if ((a_offset != 0) || (b_offset != 0) || (out_offset != 0)) {
        throw std::runtime_error("Arrays offsets have to be equals to 0");
    }

    std::int64_t n = static_cast<std::int64_t>(in_n);
    const T1 *a = reinterpret_cast<const T1 *>(in_a);
    const T2 *b = reinterpret_cast<const T2 *>(in_b);

    using resTy = typename OutputType<T1, T2>::value_type;
    resTy *y = reinterpret_cast<resTy *>(out_y);

    return mkl_vm::div(exec_q,
                       n, // number of elements to be calculated
                       a, // pointer `a` containing 1st input vector of size n
                       b, // pointer `b` containing 2nd input vector of size n
                       y, // pointer `y` to the output vector of size n
                       depends);
}

using ew_cmn_ns::binary_contig_impl_fn_ptr_t;
using ew_cmn_ns::binary_contig_matrix_contig_row_broadcast_impl_fn_ptr_t;
using ew_cmn_ns::binary_contig_row_contig_matrix_broadcast_impl_fn_ptr_t;
using ew_cmn_ns::binary_strided_impl_fn_ptr_t;

static int output_typeid_vector[td_ns::num_types][td_ns::num_types];
static binary_contig_impl_fn_ptr_t contig_dispatch_vector[td_ns::num_types]
                                                         [td_ns::num_types];

MACRO_POPULATE_DISPATCH_TABLES(div);
} // namespace impl

void init_div(py::module_ m)
{
    using arrayT = dpctl::tensor::usm_ndarray;
    using event_vecT = std::vector<sycl::event>;

    impl::populate_dispatch_tables();
    using impl::contig_dispatch_vector;
    using impl::output_typeid_vector;

    auto div_pyapi = [&](sycl::queue &exec_q, const arrayT &src1,
                         const arrayT &src2, const arrayT &dst,
                         const event_vecT &depends = {}) {
        return py_int::py_binary_ufunc(
            src1, src2, dst, exec_q, depends, output_typeid_vector,
            contig_dispatch_vector,
            // no support of strided implementation in OneMKL
            td_ns::NullPtrTable<impl::binary_strided_impl_fn_ptr_t>{},
            // no support of C-contig row with broadcasting in OneMKL
            td_ns::NullPtrTable<
                impl::
                    binary_contig_matrix_contig_row_broadcast_impl_fn_ptr_t>{},
            td_ns::NullPtrTable<
                impl::
                    binary_contig_row_contig_matrix_broadcast_impl_fn_ptr_t>{});
    };
    m.def("_div", div_pyapi,
          "Call `div` function from OneMKL VM library to performs element "
          "by element division of vector `src1` by vector `src2` "
          "to resulting vector `dst`",
          py::arg("sycl_queue"), py::arg("src1"), py::arg("src2"),
          py::arg("dst"), py::arg("depends") = py::list());

    auto div_need_to_call_pyapi = [&](sycl::queue &exec_q, const arrayT &src1,
                                      const arrayT &src2, const arrayT &dst) {
        return py_internal::need_to_call_binary_ufunc(exec_q, src1, src2, dst,
                                                      output_typeid_vector,
                                                      contig_dispatch_vector);
    };
    m.def("_mkl_div_to_call", div_need_to_call_pyapi,
          "Check input arguments to answer if `div` function from "
          "OneMKL VM library can be used",
          py::arg("sycl_queue"), py::arg("src1"), py::arg("src2"),
          py::arg("dst"));
}
} // namespace dpnp::extensions::vm

1	//*****************************************************************************
2	// Copyright (c) 2024-2025, Intel Corporation
3	// All rights reserved.
4	//
5	// Redistribution and use in source and binary forms, with or without
6	// modification, are permitted provided that the following conditions are met:
7	// - Redistributions of source code must retain the above copyright notice,
8	// this list of conditions and the following disclaimer.
9	// - Redistributions in binary form must reproduce the above copyright notice,
10	// this list of conditions and the following disclaimer in the documentation
11	// and/or other materials provided with the distribution.
12	//
13	// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
14	// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15	// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16	// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
17	// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
18	// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
19	// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
20	// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
21	// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
22	// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
23	// THE POSSIBILITY OF SUCH DAMAGE.
24	//*****************************************************************************
25
26	#include <stdexcept>
27
28	#include <oneapi/mkl.hpp>
29	#include <sycl/sycl.hpp>
30
31	#include "dpctl4pybind11.hpp"
32
33	#include "common.hpp"
34	#include "div.hpp"
35
36	// include a local copy of elementwise common header from dpctl tensor:
37	// dpctl/tensor/libtensor/source/elementwise_functions/elementwise_functions.hpp
38	// TODO: replace by including dpctl header once available
39	#include "../elementwise_functions/elementwise_functions.hpp"
40
41	// dpctl tensor headers
42	#include "kernels/elementwise_functions/common.hpp"
43	#include "utils/type_dispatch.hpp"
44	#include "utils/type_utils.hpp"
45
46	namespace dpnp::extensions::vm
47	{
48	namespace py = pybind11;
49	namespace py_int = dpnp::extensions::py_internal;
50	namespace td_ns = dpctl::tensor::type_dispatch;
51
52	namespace impl
53	{
54	namespace ew_cmn_ns = dpctl::tensor::kernels::elementwise_common;
55	namespace mkl_vm = oneapi::mkl::vm; // OneMKL namespace with VM functions
56	namespace tu_ns = dpctl::tensor::type_utils;
57
58	/**
59	* @brief A factory to define pairs of supported types for which
60	* MKL VM library provides support in oneapi::mkl::vm::div<T> function.
61	*
62	* @tparam T Type of input vectors `a` and `b` and of result vector `y`.
63	*/
64	template <typename T1, typename T2>
65	struct OutputType
66	{
67	using value_type = typename std::disjunction<
68	td_ns::BinaryTypeMapResultEntry<T1,
69	std::complex<double>,
70	T2,
71	std::complex<double>,
72	std::complex<double>>,
73	td_ns::BinaryTypeMapResultEntry<T1,
74	std::complex<float>,
75	T2,
76	std::complex<float>,
77	std::complex<float>>,
78	td_ns::BinaryTypeMapResultEntry<T1, double, T2, double, double>,
79	td_ns::BinaryTypeMapResultEntry<T1, float, T2, float, float>,
80	td_ns::DefaultResultEntry<void>>::result_type;
81	};
82
83	template <typename T1, typename T2>
84	static sycl::event div_contig_impl(sycl::queue &exec_q,
85	std::size_t in_n,
86	const char *in_a,
87	py::ssize_t a_offset,
88	const char *in_b,
89	py::ssize_t b_offset,
90	char *out_y,
91	py::ssize_t out_offset,
92	const std::vector<sycl::event> &depends)
93	{	×
94	tu_ns::validate_type_for_device<T1>(exec_q);	×
95	tu_ns::validate_type_for_device<T2>(exec_q);	×
96
97	if ((a_offset != 0) \|\| (b_offset != 0) \|\| (out_offset != 0)) {	×
98	throw std::runtime_error("Arrays offsets have to be equals to 0");	×
99	}	×
100
101	std::int64_t n = static_cast<std::int64_t>(in_n);	×
102	const T1 a = reinterpret_cast<const T1 >(in_a);	×
103	const T2 b = reinterpret_cast<const T2 >(in_b);	×
104
105	using resTy = typename OutputType<T1, T2>::value_type;	×
106	resTy y = reinterpret_cast<resTy >(out_y);	×
107
108	return mkl_vm::div(exec_q,	×
109	n, // number of elements to be calculated	×
110	a, // pointer `a` containing 1st input vector of size n	×
111	b, // pointer `b` containing 2nd input vector of size n	×
112	y, // pointer `y` to the output vector of size n	×
113	depends);	×
114	}	×
115
116	using ew_cmn_ns::binary_contig_impl_fn_ptr_t;
117	using ew_cmn_ns::binary_contig_matrix_contig_row_broadcast_impl_fn_ptr_t;
118	using ew_cmn_ns::binary_contig_row_contig_matrix_broadcast_impl_fn_ptr_t;
119	using ew_cmn_ns::binary_strided_impl_fn_ptr_t;
120
121	static int output_typeid_vector[td_ns::num_types][td_ns::num_types];
122	static binary_contig_impl_fn_ptr_t contig_dispatch_vector[td_ns::num_types]
123	[td_ns::num_types];
124
125	MACRO_POPULATE_DISPATCH_TABLES(div);
126	} // namespace impl
127
128	void init_div(py::module_ m)
129	{	2✔
130	using arrayT = dpctl::tensor::usm_ndarray;	2✔
131	using event_vecT = std::vector<sycl::event>;	2✔
132
133	impl::populate_dispatch_tables();	2✔
134	using impl::contig_dispatch_vector;	2✔
135	using impl::output_typeid_vector;	2✔
136
137	auto div_pyapi = [&](sycl::queue &exec_q, const arrayT &src1,	2✔
138	const arrayT &src2, const arrayT &dst,	2✔
139	const event_vecT &depends = {}) {	2✔
140	return py_int::py_binary_ufunc(	×
141	src1, src2, dst, exec_q, depends, output_typeid_vector,	×
142	contig_dispatch_vector,	×
143	// no support of strided implementation in OneMKL
144	td_ns::NullPtrTable<impl::binary_strided_impl_fn_ptr_t>{},	×
145	// no support of C-contig row with broadcasting in OneMKL
146	td_ns::NullPtrTable<	×
147	impl::	×
148	binary_contig_matrix_contig_row_broadcast_impl_fn_ptr_t>{},	×
149	td_ns::NullPtrTable<	×
150	impl::	×
151	binary_contig_row_contig_matrix_broadcast_impl_fn_ptr_t>{});	×
152	};	×
153	m.def("_div", div_pyapi,	2✔
154	"Call `div` function from OneMKL VM library to performs element "	2✔
155	"by element division of vector `src1` by vector `src2` "	2✔
156	"to resulting vector `dst`",	2✔
157	py::arg("sycl_queue"), py::arg("src1"), py::arg("src2"),	2✔
158	py::arg("dst"), py::arg("depends") = py::list());	2✔
159
160	auto div_need_to_call_pyapi = [&](sycl::queue &exec_q, const arrayT &src1,	2✔
161	const arrayT &src2, const arrayT &dst) {	2✔
162	return py_internal::need_to_call_binary_ufunc(exec_q, src1, src2, dst,	×
163	output_typeid_vector,	×
164	contig_dispatch_vector);	×
165	};	×
166	m.def("_mkl_div_to_call", div_need_to_call_pyapi,	2✔
167	"Check input arguments to answer if `div` function from "	2✔
168	"OneMKL VM library can be used",	2✔
169	py::arg("sycl_queue"), py::arg("src1"), py::arg("src2"),	2✔
170	py::arg("dst"));	2✔
171	}	2✔
172	} // namespace dpnp::extensions::vm

IntelPython / dpnp / 16126000227

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous