diff --git a/dpctl/tensor/__init__.py b/dpctl/tensor/__init__.py
index 2e720cba92..1c58f616a1 100644
--- a/dpctl/tensor/__init__.py
+++ b/dpctl/tensor/__init__.py
@@ -107,6 +107,7 @@
     log,
     log1p,
     multiply,
+    not_equal,
     proj,
     real,
     sin,
@@ -210,5 +211,6 @@
     "multiply",
     "subtract",
     "equal",
+    "not_equal",
     "sum",
 ]
diff --git a/dpctl/tensor/_elementwise_funcs.py b/dpctl/tensor/_elementwise_funcs.py
index cb24929b76..171c150b50 100644
--- a/dpctl/tensor/_elementwise_funcs.py
+++ b/dpctl/tensor/_elementwise_funcs.py
@@ -480,7 +480,33 @@
 # FIXME: implement U25
 
 # B20: ==== NOT_EQUAL   (x1, x2)
-# FIXME: implement B20
+_not_equal_docstring_ = """
+not_equal(x1, x2, out=None, order='K')
+
+Calculates inequality test results for each element `x1_i` of the
+input array `x1` with the respective element `x2_i` of the input array `x2`.
+
+Args:
+    x1 (usm_ndarray):
+        First input array, expected to have numeric data type.
+    x2 (usm_ndarray):
+        Second input array, also expected to have numeric 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 result of element-wise inequality comparison.
+        The data type of the returned array is determined by the
+        Type Promotion Rules.
+"""
+
+not_equal = BinaryElementwiseFunc(
+    "not_equal", ti._not_equal_result_type, ti._not_equal, _not_equal_docstring_
+)
 
 # U26: ==== POSITIVE    (x)
 # FIXME: implement U26
diff --git a/dpctl/tensor/libtensor/include/kernels/elementwise_functions/not_equal.hpp b/dpctl/tensor/libtensor/include/kernels/elementwise_functions/not_equal.hpp
new file mode 100644
index 0000000000..d7caa6c089
--- /dev/null
+++ b/dpctl/tensor/libtensor/include/kernels/elementwise_functions/not_equal.hpp
@@ -0,0 +1,266 @@
+//=== not_equal.hpp -   Binary function NOT_EQUAL        ------  *-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 inequality of
+/// tensor elements.
+//===---------------------------------------------------------------------===//
+
+#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 not_equal
+{
+
+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 NotEqualFunctor
+{
+    static_assert(std::is_same_v<resT, bool>);
+
+    using supports_sg_loadstore = std::negation<
+        std::disjunction<tu_ns::is_complex<argT1>, tu_ns::is_complex<argT2>>>;
+    using supports_vec = std::conjunction<
+        std::is_same<argT1, argT2>,
+        std::negation<std::disjunction<tu_ns::is_complex<argT1>,
+                                       tu_ns::is_complex<argT2>>>>;
+
+    resT operator()(const argT1 &in1, const argT2 &in2)
+    {
+        if constexpr (std::is_same_v<argT1, std::complex<float>> &&
+                      std::is_same_v<argT2, float>)
+        {
+            return (std::real(in1) != in2 || std::imag(in1) != 0.0f);
+        }
+        else if constexpr (std::is_same_v<argT1, float> &&
+                           std::is_same_v<argT2, std::complex<float>>)
+        {
+            return (in1 != std::real(in2) || std::imag(in2) != 0.0f);
+        }
+        else {
+            return (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)
+    {
+        auto tmp = (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 NotEqualContigFunctor =
+    elementwise_common::BinaryContigFunctor<argT1,
+                                            argT2,
+                                            resT,
+                                            NotEqualFunctor<argT1, argT2, resT>,
+                                            vec_sz,
+                                            n_vecs>;
+
+template <typename argT1, typename argT2, typename resT, typename IndexerT>
+using NotEqualStridedFunctor = elementwise_common::BinaryStridedFunctor<
+    argT1,
+    argT2,
+    resT,
+    IndexerT,
+    NotEqualFunctor<argT1, argT2, resT>>;
+
+template <typename T1, typename T2> struct NotEqualOutputType
+{
+    using value_type = typename std::disjunction< // disjunction is C++17
+                                                  // feature, supported by DPC++
+        td_ns::BinaryTypeMapResultEntry<T1, bool, T2, bool, bool>,
+        td_ns::
+            BinaryTypeMapResultEntry<T1, std::uint8_t, T2, std::uint8_t, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1, std::int8_t, T2, std::int8_t, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1,
+                                        std::uint16_t,
+                                        T2,
+                                        std::uint16_t,
+                                        bool>,
+        td_ns::
+            BinaryTypeMapResultEntry<T1, std::int16_t, T2, std::int16_t, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1,
+                                        std::uint32_t,
+                                        T2,
+                                        std::uint32_t,
+                                        bool>,
+        td_ns::
+            BinaryTypeMapResultEntry<T1, std::int32_t, T2, std::int32_t, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1,
+                                        std::uint64_t,
+                                        T2,
+                                        std::uint64_t,
+                                        bool>,
+        td_ns::
+            BinaryTypeMapResultEntry<T1, std::int64_t, T2, std::int64_t, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1, sycl::half, T2, sycl::half, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1, float, T2, float, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1, double, T2, double, bool>,
+        td_ns::BinaryTypeMapResultEntry<T1,
+                                        std::complex<float>,
+                                        T2,
+                                        std::complex<float>,
+                                        bool>,
+        td_ns::BinaryTypeMapResultEntry<T1,
+                                        std::complex<double>,
+                                        T2,
+                                        std::complex<double>,
+                                        bool>,
+        td_ns::
+            BinaryTypeMapResultEntry<T1, float, T2, std::complex<float>, bool>,
+        td_ns::
+            BinaryTypeMapResultEntry<T1, std::complex<float>, T2, float, bool>,
+        td_ns::DefaultResultEntry<void>>::result_type;
+};
+
+template <typename argT1,
+          typename argT2,
+          typename resT,
+          unsigned int vec_sz,
+          unsigned int n_vecs>
+class not_equal_contig_kernel;
+
+template <typename argTy1, typename argTy2>
+sycl::event not_equal_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, NotEqualOutputType, NotEqualContigFunctor,
+        not_equal_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 NotEqualContigFactory
+{
+    fnT get()
+    {
+        if constexpr (std::is_same_v<
+                          typename NotEqualOutputType<T1, T2>::value_type,
+                          void>)
+        {
+            fnT fn = nullptr;
+            return fn;
+        }
+        else {
+            fnT fn = not_equal_contig_impl<T1, T2>;
+            return fn;
+        }
+    }
+};
+
+template <typename fnT, typename T1, typename T2> struct NotEqualTypeMapFactory
+{
+    /*! @brief get typeid for output type of operator()!=(x, y), always bool */
+    std::enable_if_t<std::is_same<fnT, int>::value, int> get()
+    {
+        using rT = typename NotEqualOutputType<T1, T2>::value_type;
+        return td_ns::GetTypeid<rT>{}.get();
+    }
+};
+
+template <typename T1, typename T2, typename resT, typename IndexerT>
+class not_equal_strided_strided_kernel;
+
+template <typename argTy1, typename argTy2>
+sycl::event
+not_equal_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, NotEqualOutputType, NotEqualStridedFunctor,
+        not_equal_strided_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 NotEqualStridedFactory
+{
+    fnT get()
+    {
+        if constexpr (std::is_same_v<
+                          typename NotEqualOutputType<T1, T2>::value_type,
+                          void>)
+        {
+            fnT fn = nullptr;
+            return fn;
+        }
+        else {
+            fnT fn = not_equal_strided_impl<T1, T2>;
+            return fn;
+        }
+    }
+};
+
+} // namespace not_equal
+} // namespace kernels
+} // namespace tensor
+} // namespace dpctl
diff --git a/dpctl/tensor/libtensor/source/elementwise_functions.cpp b/dpctl/tensor/libtensor/source/elementwise_functions.cpp
index 88597512bc..b0e58e0f6c 100644
--- a/dpctl/tensor/libtensor/source/elementwise_functions.cpp
+++ b/dpctl/tensor/libtensor/source/elementwise_functions.cpp
@@ -46,6 +46,7 @@
 #include "kernels/elementwise_functions/log.hpp"
 #include "kernels/elementwise_functions/log1p.hpp"
 #include "kernels/elementwise_functions/multiply.hpp"
+#include "kernels/elementwise_functions/not_equal.hpp"
 #include "kernels/elementwise_functions/proj.hpp"
 #include "kernels/elementwise_functions/real.hpp"
 #include "kernels/elementwise_functions/sin.hpp"
@@ -924,7 +925,39 @@ namespace impl
 // B20: ==== NOT_EQUAL   (x1, x2)
 namespace impl
 {
-// FIXME: add code for B20
+namespace not_equal_fn_ns = dpctl::tensor::kernels::not_equal;
+
+static binary_contig_impl_fn_ptr_t
+    not_equal_contig_dispatch_table[td_ns::num_types][td_ns::num_types];
+static int not_equal_output_id_table[td_ns::num_types][td_ns::num_types];
+
+static binary_strided_impl_fn_ptr_t
+    not_equal_strided_dispatch_table[td_ns::num_types][td_ns::num_types];
+
+void populate_not_equal_dispatch_tables(void)
+{
+    using namespace td_ns;
+    namespace fn_ns = not_equal_fn_ns;
+
+    // which input types are supported, and what is the type of the result
+    using fn_ns::NotEqualTypeMapFactory;
+    DispatchTableBuilder<int, NotEqualTypeMapFactory, num_types> dtb1;
+    dtb1.populate_dispatch_table(not_equal_output_id_table);
+
+    // function pointers for operation on general strided arrays
+    using fn_ns::NotEqualStridedFactory;
+    DispatchTableBuilder<binary_strided_impl_fn_ptr_t, NotEqualStridedFactory,
+                         num_types>
+        dtb2;
+    dtb2.populate_dispatch_table(not_equal_strided_dispatch_table);
+
+    // function pointers for operation on contiguous inputs and output
+    using fn_ns::NotEqualContigFactory;
+    DispatchTableBuilder<binary_contig_impl_fn_ptr_t, NotEqualContigFactory,
+                         num_types>
+        dtb3;
+    dtb3.populate_dispatch_table(not_equal_contig_dispatch_table);
+};
 } // namespace impl
 
 // U26: ==== POSITIVE    (x)
@@ -1702,7 +1735,45 @@ void init_elementwise_functions(py::module_ m)
     // FIXME:
 
     // B20: ==== NOT_EQUAL   (x1, x2)
-    // FIXME:
+    {
+        impl::populate_not_equal_dispatch_tables();
+        using impl::not_equal_contig_dispatch_table;
+        using impl::not_equal_output_id_table;
+        using impl::not_equal_strided_dispatch_table;
+
+        auto not_equal_pyapi = [&](dpctl::tensor::usm_ndarray src1,
+                                   dpctl::tensor::usm_ndarray src2,
+                                   dpctl::tensor::usm_ndarray dst,
+                                   sycl::queue exec_q,
+                                   const std::vector<sycl::event> &depends =
+                                       {}) {
+            return py_binary_ufunc(
+                src1, src2, dst, exec_q, depends, not_equal_output_id_table,
+                // function pointers to handle operation on contiguous arrays
+                // (pointers may be nullptr)
+                not_equal_contig_dispatch_table,
+                // function pointers to handle operation on strided arrays (most
+                // general case)
+                not_equal_strided_dispatch_table,
+                // function pointers to handle operation of c-contig matrix and
+                // c-contig row with broadcasting (may be nullptr)
+                td_ns::NullPtrTable<
+                    binary_contig_matrix_contig_row_broadcast_impl_fn_ptr_t>{},
+                // function pointers to handle operation of c-contig matrix and
+                // c-contig row with broadcasting (may be nullptr)
+                td_ns::NullPtrTable<
+                    binary_contig_row_contig_matrix_broadcast_impl_fn_ptr_t>{});
+        };
+        auto not_equal_result_type_pyapi = [&](py::dtype dtype1,
+                                               py::dtype dtype2) {
+            return py_binary_ufunc_result_type(dtype1, dtype2,
+                                               not_equal_output_id_table);
+        };
+        m.def("_not_equal", not_equal_pyapi, "", py::arg("src1"),
+              py::arg("src2"), py::arg("dst"), py::arg("sycl_queue"),
+              py::arg("depends") = py::list());
+        m.def("_not_equal_result_type", not_equal_result_type_pyapi, "");
+    }
 
     // U26: ==== POSITIVE    (x)
     // FIXME:
diff --git a/dpctl/tests/elementwise/test_equal.py b/dpctl/tests/elementwise/test_equal.py
index cdd26a32d0..502d0d4886 100644
--- a/dpctl/tests/elementwise/test_equal.py
+++ b/dpctl/tests/elementwise/test_equal.py
@@ -127,6 +127,10 @@ def test_equal_broadcasting():
     r2 = dpt.equal(v, m)
     assert (dpt.asnumpy(r2) == expected).all()
 
+    r3 = dpt.empty_like(m, dtype="?")
+    dpt.equal(m, v, r3)
+    assert (dpt.asnumpy(r3) == expected).all()
+
 
 @pytest.mark.parametrize("arr_dt", _all_dtypes)
 def test_equal_python_scalar(arr_dt):
diff --git a/dpctl/tests/elementwise/test_not_equal.py b/dpctl/tests/elementwise/test_not_equal.py
new file mode 100644
index 0000000000..6baef3f145
--- /dev/null
+++ b/dpctl/tests/elementwise/test_not_equal.py
@@ -0,0 +1,190 @@
+#                       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.
+
+import ctypes
+
+import numpy as np
+import pytest
+
+import dpctl
+import dpctl.tensor as dpt
+from dpctl.tests.helper import get_queue_or_skip, skip_if_dtype_not_supported
+
+from .utils import _all_dtypes, _compare_dtypes, _usm_types
+
+
+@pytest.mark.parametrize("op1_dtype", _all_dtypes)
+@pytest.mark.parametrize("op2_dtype", _all_dtypes)
+def test_not_equal_dtype_matrix(op1_dtype, op2_dtype):
+    q = get_queue_or_skip()
+    skip_if_dtype_not_supported(op1_dtype, q)
+    skip_if_dtype_not_supported(op2_dtype, q)
+
+    sz = 127
+    ar1 = dpt.ones(sz, dtype=op1_dtype)
+    ar2 = dpt.ones_like(ar1, dtype=op2_dtype)
+
+    r = dpt.not_equal(ar1, ar2)
+    assert isinstance(r, dpt.usm_ndarray)
+    expected_dtype = np.not_equal(
+        np.zeros(1, dtype=op1_dtype), np.zeros(1, dtype=op2_dtype)
+    ).dtype
+    assert _compare_dtypes(r.dtype, expected_dtype, sycl_queue=q)
+    assert r.shape == ar1.shape
+    assert (dpt.asnumpy(r) == np.full(r.shape, False, dtype=r.dtype)).all()
+    assert r.sycl_queue == ar1.sycl_queue
+
+    ar3 = dpt.ones(sz, dtype=op1_dtype)
+    ar4 = dpt.ones(2 * sz, dtype=op2_dtype)
+
+    r = dpt.not_equal(ar3[::-1], ar4[::2])
+    assert isinstance(r, dpt.usm_ndarray)
+    expected_dtype = np.not_equal(
+        np.ones(1, dtype=op1_dtype), np.ones(1, dtype=op2_dtype)
+    ).dtype
+    assert _compare_dtypes(r.dtype, expected_dtype, sycl_queue=q)
+    assert r.shape == ar3.shape
+    assert (dpt.asnumpy(r) == np.full(r.shape, False, dtype=r.dtype)).all()
+
+
+@pytest.mark.parametrize("op1_usm_type", _usm_types)
+@pytest.mark.parametrize("op2_usm_type", _usm_types)
+def test_not_equal_usm_type_matrix(op1_usm_type, op2_usm_type):
+    get_queue_or_skip()
+
+    sz = 128
+    ar1 = dpt.ones(sz, dtype="i4", usm_type=op1_usm_type)
+    ar2 = dpt.ones_like(ar1, dtype="i4", usm_type=op2_usm_type)
+
+    r = dpt.not_equal(ar1, ar2)
+    assert isinstance(r, dpt.usm_ndarray)
+    expected_usm_type = dpctl.utils.get_coerced_usm_type(
+        (op1_usm_type, op2_usm_type)
+    )
+    assert r.usm_type == expected_usm_type
+
+
+def test_not_equal_order():
+    get_queue_or_skip()
+
+    ar1 = dpt.ones((20, 20), dtype="i4", order="C")
+    ar2 = dpt.ones((20, 20), dtype="i4", order="C")
+    r1 = dpt.not_equal(ar1, ar2, order="C")
+    assert r1.flags.c_contiguous
+    r2 = dpt.not_equal(ar1, ar2, order="F")
+    assert r2.flags.f_contiguous
+    r3 = dpt.not_equal(ar1, ar2, order="A")
+    assert r3.flags.c_contiguous
+    r4 = dpt.not_equal(ar1, ar2, order="K")
+    assert r4.flags.c_contiguous
+
+    ar1 = dpt.ones((20, 20), dtype="i4", order="F")
+    ar2 = dpt.ones((20, 20), dtype="i4", order="F")
+    r1 = dpt.not_equal(ar1, ar2, order="C")
+    assert r1.flags.c_contiguous
+    r2 = dpt.not_equal(ar1, ar2, order="F")
+    assert r2.flags.f_contiguous
+    r3 = dpt.not_equal(ar1, ar2, order="A")
+    assert r3.flags.f_contiguous
+    r4 = dpt.not_equal(ar1, ar2, order="K")
+    assert r4.flags.f_contiguous
+
+    ar1 = dpt.ones((40, 40), dtype="i4", order="C")[:20, ::-2]
+    ar2 = dpt.ones((40, 40), dtype="i4", order="C")[:20, ::-2]
+    r4 = dpt.not_equal(ar1, ar2, order="K")
+    assert r4.strides == (20, -1)
+
+    ar1 = dpt.ones((40, 40), dtype="i4", order="C")[:20, ::-2].mT
+    ar2 = dpt.ones((40, 40), dtype="i4", order="C")[:20, ::-2].mT
+    r4 = dpt.not_equal(ar1, ar2, order="K")
+    assert r4.strides == (-1, 20)
+
+
+def test_not_equal_broadcasting():
+    get_queue_or_skip()
+
+    m = dpt.ones((100, 5), dtype="i4")
+    v = dpt.arange(5, dtype="i4")
+
+    r = dpt.not_equal(m, v)
+    expected = np.full((100, 5), [True, False, True, True, True], dtype="?")
+
+    assert (dpt.asnumpy(r) == expected).all()
+
+    r2 = dpt.not_equal(v, m)
+    assert (dpt.asnumpy(r2) == expected).all()
+
+    r3 = dpt.empty_like(m, dtype="?")
+    dpt.not_equal(m, v, r3)
+    assert (dpt.asnumpy(r3) == expected).all()
+
+
+@pytest.mark.parametrize("arr_dt", _all_dtypes)
+def test_not_equal_python_scalar(arr_dt):
+    q = get_queue_or_skip()
+    skip_if_dtype_not_supported(arr_dt, q)
+
+    X = dpt.zeros((10, 10), dtype=arr_dt, sycl_queue=q)
+    py_zeros = (
+        bool(0),
+        int(0),
+        float(0),
+        complex(0),
+        np.float32(0),
+        ctypes.c_int(0),
+    )
+    for sc in py_zeros:
+        R = dpt.not_equal(X, sc)
+        assert isinstance(R, dpt.usm_ndarray)
+        assert not dpt.all(R)
+        R = dpt.not_equal(sc, X)
+        assert isinstance(R, dpt.usm_ndarray)
+        assert not dpt.all(R)
+
+
+class MockArray:
+    def __init__(self, arr):
+        self.data_ = arr
+
+    @property
+    def __sycl_usm_array_interface__(self):
+        return self.data_.__sycl_usm_array_interface__
+
+
+def test_not_equal_mock_array():
+    get_queue_or_skip()
+    a = dpt.arange(10)
+    b = dpt.ones(10)
+    c = MockArray(b)
+    r = dpt.not_equal(a, c)
+    assert isinstance(r, dpt.usm_ndarray)
+
+
+def test_not_equal_canary_mock_array():
+    get_queue_or_skip()
+    a = dpt.arange(10)
+
+    class Canary:
+        def __init__(self):
+            pass
+
+        @property
+        def __sycl_usm_array_interface__(self):
+            return None
+
+    c = Canary()
+    with pytest.raises(ValueError):
+        dpt.not_equal(a, c)