Implements copysign and exp2 elementwise funcs

Author: ndgrigorian

dpctl/tensor/__init__.py

@@ -113,11 +113,13 @@
     cbrt,
     ceil,
     conj,
+    copysign,
     cos,
     cosh,
     divide,
     equal,
     exp,
+    exp2,
     expm1,
     floor,
     floor_divide,
@@ -316,4 +318,6 @@
     "argmin",
     "prod",
     "cbrt",
+    "exp2",
+    "copysign",
 ]

dpctl/tensor/_elementwise_funcs.py

@@ -1788,3 +1788,62 @@
 cbrt = UnaryElementwiseFunc(
     "cbrt", ti._cbrt_result_type, ti._cbrt, _cbrt_docstring_
 )
+
+
+# U34: ==== EXP2        (x)
+_exp2_docstring_ = """
+exp2(x, out=None, order='K')
+
+Computes the base-2 exponential for each element `x_i` for input array `x`.
+
+Args:
+    x (usm_ndarray):
+        Input array, expected to have a floating-point data type.
+    out ({None, usm_ndarray}, optional):
+        Output array to populate.
+        Array have the correct shape and the expected data type.
+    order ("C","F","A","K", optional):
+        Memory layout of the newly output array, if parameter `out` is `None`.
+        Default: "K".
+Returns:
+    usm_narray:
+        An array containing the element-wise base-2 exponentials.
+        The data type of the returned array is determined by
+        the Type Promotion Rules.
+"""
+
+exp2 = UnaryElementwiseFunc(
+    "exp2", ti._exp2_result_type, ti._exp2, _exp2_docstring_
+)
+
+
+# B23: ==== COPYSIGN    (x1, x2)
+_copysign_docstring_ = """
+copysign(x1, x2, out=None, order='K')
+
+Composes a floating-point value with the magnitude of `x1_i` and the sign of
+`x2_i` for each element of input arrays `x1` and `x2`.
+
+Args:
+    x1 (usm_ndarray):
+        First input array, expected to have a real floating-point data type.
+    x2 (usm_ndarray):
+        Second input array, also expected to have a real floating-point data
+        type.
+    out ({None, usm_ndarray}, optional):
+        Output array to populate.
+        Array have the correct shape and the expected data type.
+    order ("C","F","A","K", optional):
+        Memory layout of the newly output array, if parameter `out` is `None`.
+        Default: "K".
+Returns:
+    usm_narray:
+        An array containing the element-wise results. The data type
+        of the returned array is determined by the Type Promotion Rules.
+"""
+copysign = BinaryElementwiseFunc(
+    "copysign",
+    ti._copysign_result_type,
+    ti._copysign,
+    _copysign_docstring_,
+)

dpctl/tensor/libtensor/include/kernels/elementwise_functions/copysign.hpp

@@ -0,0 +1,215 @@
+//=== copysign.hpp -   Binary function COPYSIGN         ------  *-C++-*--/===//
+//
+//                      Data Parallel Control (dpctl)
+//
+// Copyright 2020-2023 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//    http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+//===---------------------------------------------------------------------===//
+///
+/// \file
+/// This file defines kernels for elementwise evaluation of COPYSIGN(x1, x2)
+/// function.
+//===---------------------------------------------------------------------===//
+
+#pragma once
+#include <CL/sycl.hpp>
+#include <cstddef>
+#include <cstdint>
+#include <type_traits>
+
+#include "utils/offset_utils.hpp"
+#include "utils/type_dispatch.hpp"
+#include "utils/type_utils.hpp"
+
+#include "kernels/elementwise_functions/common.hpp"
+#include <pybind11/pybind11.h>
+
+namespace dpctl
+{
+namespace tensor
+{
+namespace kernels
+{
+namespace copysign
+{
+
+namespace py = pybind11;
+namespace td_ns = dpctl::tensor::type_dispatch;
+namespace tu_ns = dpctl::tensor::type_utils;
+
+template <typename argT1, typename argT2, typename resT> struct CopysignFunctor
+{
+
+    using supports_sg_loadstore = std::true_type;
+    using supports_vec = std::true_type;
+
+    resT operator()(const argT1 &in1, const argT2 &in2) const
+    {
+        return sycl::copysign(in1, in2);
+    }
+
+    template <int vec_sz>
+    sycl::vec<resT, vec_sz>
+    operator()(const sycl::vec<argT1, vec_sz> &in1,
+               const sycl::vec<argT2, vec_sz> &in2) const
+    {
+        auto tmp = sycl::copysign(in1, in2);
+        if constexpr (std::is_same_v<resT,
+                                     typename decltype(tmp)::element_type>) {
+            return tmp;
+        }
+        else {
+            using dpctl::tensor::type_utils::vec_cast;
+
+            return vec_cast<resT, typename decltype(tmp)::element_type, vec_sz>(
+                tmp);
+        }
+    }
+};
+
+template <typename argT1,
+          typename argT2,
+          typename resT,
+          unsigned int vec_sz = 4,
+          unsigned int n_vecs = 2>
+using CopysignContigFunctor =
+    elementwise_common::BinaryContigFunctor<argT1,
+                                            argT2,
+                                            resT,
+                                            CopysignFunctor<argT1, argT2, resT>,
+                                            vec_sz,
+                                            n_vecs>;
+
+template <typename argT1, typename argT2, typename resT, typename IndexerT>
+using CopysignStridedFunctor = elementwise_common::BinaryStridedFunctor<
+    argT1,
+    argT2,
+    resT,
+    IndexerT,
+    CopysignFunctor<argT1, argT2, resT>>;
+
+template <typename T1, typename T2> struct CopysignOutputType
+{
+    using value_type = typename std::disjunction< // disjunction is C++17
+                                                  // feature, supported by DPC++
+        td_ns::BinaryTypeMapResultEntry<T1,
+                                        sycl::half,
+                                        T2,
+                                        sycl::half,
+                                        sycl::half>,
+        td_ns::BinaryTypeMapResultEntry<T1, float, T2, float, float>,
+        td_ns::BinaryTypeMapResultEntry<T1, double, T2, double, double>,
+        td_ns::DefaultResultEntry<void>>::result_type;
+};
+
+template <typename argT1,
+          typename argT2,
+          typename resT,
+          unsigned int vec_sz,
+          unsigned int n_vecs>
+class copysign_contig_kernel;
+
+template <typename argTy1, typename argTy2>
+sycl::event copysign_contig_impl(sycl::queue &exec_q,
+                                 size_t nelems,
+                                 const char *arg1_p,
+                                 py::ssize_t arg1_offset,
+                                 const char *arg2_p,
+                                 py::ssize_t arg2_offset,
+                                 char *res_p,
+                                 py::ssize_t res_offset,
+                                 const std::vector<sycl::event> &depends = {})
+{
+    return elementwise_common::binary_contig_impl<
+        argTy1, argTy2, CopysignOutputType, CopysignContigFunctor,
+        copysign_contig_kernel>(exec_q, nelems, arg1_p, arg1_offset, arg2_p,
+                                arg2_offset, res_p, res_offset, depends);
+}
+
+template <typename fnT, typename T1, typename T2> struct CopysignContigFactory
+{
+    fnT get()
+    {
+        if constexpr (std::is_same_v<
+                          typename CopysignOutputType<T1, T2>::value_type,
+                          void>)
+        {
+            fnT fn = nullptr;
+            return fn;
+        }
+        else {
+            fnT fn = copysign_contig_impl<T1, T2>;
+            return fn;
+        }
+    }
+};
+
+template <typename fnT, typename T1, typename T2> struct CopysignTypeMapFactory
+{
+    /*! @brief get typeid for output type of divide(T1 x, T2 y) */
+    std::enable_if_t<std::is_same<fnT, int>::value, int> get()
+    {
+        using rT = typename CopysignOutputType<T1, T2>::value_type;
+        return td_ns::GetTypeid<rT>{}.get();
+    }
+};
+
+template <typename T1, typename T2, typename resT, typename IndexerT>
+class copysign_strided_kernel;
+
+template <typename argTy1, typename argTy2>
+sycl::event
+copysign_strided_impl(sycl::queue &exec_q,
+                      size_t nelems,
+                      int nd,
+                      const py::ssize_t *shape_and_strides,
+                      const char *arg1_p,
+                      py::ssize_t arg1_offset,
+                      const char *arg2_p,
+                      py::ssize_t arg2_offset,
+                      char *res_p,
+                      py::ssize_t res_offset,
+                      const std::vector<sycl::event> &depends,
+                      const std::vector<sycl::event> &additional_depends)
+{
+    return elementwise_common::binary_strided_impl<
+        argTy1, argTy2, CopysignOutputType, CopysignStridedFunctor,
+        copysign_strided_kernel>(exec_q, nelems, nd, shape_and_strides, arg1_p,
+                                 arg1_offset, arg2_p, arg2_offset, res_p,
+                                 res_offset, depends, additional_depends);
+}
+
+template <typename fnT, typename T1, typename T2> struct CopysignStridedFactory
+{
+    fnT get()
+    {
+        if constexpr (std::is_same_v<
+                          typename CopysignOutputType<T1, T2>::value_type,
+                          void>)
+        {
+            fnT fn = nullptr;
+            return fn;
+        }
+        else {
+            fnT fn = copysign_strided_impl<T1, T2>;
+            return fn;
+        }
+    }
+};
+
+} // namespace copysign
+} // namespace kernels
+} // namespace tensor
+} // namespace dpctl