address comments

cherry-pick

Author: Vahid Tavanashad

dpnp/dpnp_array.py

@@ -1033,31 +1033,6 @@ def prod(
 
         return dpnp.prod(self, axis, dtype, out, keepdims, initial, where)
 
-    def ptp(
-        self,
-        axis=None,
-        out=None,
-        keepdims=numpy._NoValue,
-        device=None,
-        usm_type=None,
-        sycl_queue=None,
-    ):
-        """
-        Range of values (maximum - minimum) along an axis.
-
-        For full documentation refer to :obj:`numpy.ptp`.
-        """
-
-        return dpnp.ptp(
-            self,
-            axis=axis,
-            out=out,
-            keepdims=keepdims,
-            device=device,
-            usm_type=usm_type,
-            sycl_queue=sycl_queue,
-        )
-
     def put(self, indices, vals, /, *, axis=None, mode="wrap"):
         """
         Puts values of an array into another array along a given axis.

dpnp/dpnp_iface_arraycreation.py

@@ -80,7 +80,6 @@
     "ogrid",
     "ones",
     "ones_like",
-    "ptp",
     "trace",
     "tri",
     "tril",
@@ -1602,69 +1601,6 @@ def ones_like(
     return call_origin(numpy.ones_like, x1, dtype, order, subok, shape)
 
 
-def ptp(
-    arr,
-    /,
-    axis=None,
-    out=None,
-    keepdims=numpy._NoValue,
-    *,
-    device=None,
-    usm_type=None,
-    sycl_queue=None,
-):
-    """
-    Range of values (maximum - minimum) along an axis.
-
-    For full documentation refer to :obj:`numpy.ptp`.
-
-    Returns
-    -------
-    ptp : dpnp.ndarray
-        The range of a given array.
-
-    Limitations
-    -----------
-    Input array is supported as :class:`dpnp.dpnp_array` or :class:`dpctl.tensor.usm_ndarray`.
-    Parameters `out` and `keepdims` are supported only with default values.
-    Otherwise the function will be executed sequentially on CPU.
-
-    Examples
-    --------
-    >>> import dpnp as np
-    >>> x = np.array([[4, 9, 2, 10],[6, 9, 7, 12]])
-    >>> np.ptp(x, axis=1)
-    array([8, 6])
-
-    >>> np.ptp(x, axis=0)
-    array([2, 0, 5, 2])
-
-    >>> np.ptp(x)
-    array(10)
-    """
-    if not isinstance(arr, (dpnp.ndarray, dpt.usm_ndarray)):
-        pass
-    elif axis is not None and not isinstance(axis, int):
-        pass
-    elif out is not None:
-        pass
-    elif keepdims is not numpy._NoValue:
-        pass
-    else:
-        max_array = dpnp.max(arr, axis=axis)
-        min_array = dpnp.min(arr, axis=axis)
-
-        _usm_type = arr.usm_type if usm_type is None else usm_type
-        _sycl_queue = dpnp.get_normalized_queue_device(
-            arr, sycl_queue=sycl_queue, device=device
-        )
-        return dpnp.array(
-            max_array - min_array, usm_type=_usm_type, sycl_queue=_sycl_queue
-        )
-
-    return call_origin(numpy.ptp, arr, axis, out, keepdims)
-
-
 def trace(x1, offset=0, axis1=0, axis2=1, dtype=None, out=None):
     """
     Return the sum along diagonals of the array.

dpnp/dpnp_iface_statistics.py

@@ -60,6 +60,7 @@
     "mean",
     "median",
     "min",
+    "ptp",
     "nanvar",
     "std",
     "var",
@@ -621,6 +622,55 @@ def min(a, axis=None, out=None, keepdims=False, initial=None, where=True):
     return call_origin(numpy.min, a, axis, out, keepdims, initial, where)
 
 
+def ptp(
+    a,
+    /,
+    axis=None,
+    out=None,
+    keepdims=False,
+):
+    """
+    Range of values (maximum - minimum) along an axis.
+
+    For full documentation refer to :obj:`numpy.ptp`.
+
+    Returns
+    -------
+    ptp : dpnp.ndarray
+        The range of a given array.
+
+    Limitations
+    -----------
+    Input array is supported as :class:`dpnp.dpnp_array` or :class:`dpctl.tensor.usm_ndarray`.
+    Otherwise the function will be executed sequentially on CPU.
+
+    Examples
+    --------
+    >>> import dpnp as np
+    >>> x = np.array([[4, 9, 2, 10],[6, 9, 7, 12]])
+    >>> np.ptp(x, axis=1)
+    array([8, 6])
+
+    >>> np.ptp(x, axis=0)
+    array([2, 0, 5, 2])
+
+    >>> np.ptp(x)
+    array(10)
+
+    """
+
+    if not isinstance(a, (dpnp.ndarray, dpt.usm_ndarray)):
+        pass
+    else:
+        return dpnp.subtract(
+            dpnp.max(a, axis=axis, keepdims=keepdims),
+            dpnp.min(a, axis=axis, keepdims=keepdims),
+            out=out,
+        )
+
+    return call_origin(numpy.ptp, a, axis, out, keepdims)
+
+
 def nanvar(x1, axis=None, dtype=None, out=None, ddof=0, keepdims=False):
     """
     Compute the variance along the specified axis, while ignoring NaNs.

tests/skipped_tests_gpu.tbl

@@ -244,9 +244,7 @@ tests/third_party/cupy/core_tests/test_ndarray_copy_and_view.py::TestArrayFlatte
 tests/third_party/cupy/core_tests/test_ndarray_copy_and_view.py::TestArrayFlatten::test_flatten_order_copied
 tests/third_party/cupy/core_tests/test_ndarray_copy_and_view.py::TestArrayFlatten::test_flatten_order_transposed
 
-tests/third_party/cupy/core_tests/test_ndarray_reduction.py::TestArrayReduction::test_ptp_nan
-tests/third_party/cupy/core_tests/test_ndarray_reduction.py::TestArrayReduction::test_ptp_nan_imag
-tests/third_party/cupy/core_tests/test_ndarray_reduction.py::TestArrayReduction::test_ptp_nan_real
+
 
 tests/third_party/cupy/core_tests/test_ndarray_reduction.py::TestCubReduction_param_0_{order='C', shape=(10,)}::test_cub_max
 tests/third_party/cupy/core_tests/test_ndarray_reduction.py::TestCubReduction_param_0_{order='C', shape=(10,)}::test_cub_min

tests/third_party/cupy/core_tests/test_ndarray_reduction.py

@@ -9,7 +9,6 @@
 from tests.third_party.cupy import testing
 
 
-@pytest.mark.usefixtures("allow_fall_back_on_numpy")
 @testing.gpu
 class TestArrayReduction(unittest.TestCase):
     @testing.for_all_dtypes()
@@ -149,70 +148,70 @@ def test_min_nan_imag(self, xp, dtype):
     @testing.numpy_cupy_allclose()
     def test_ptp_all(self, xp, dtype):
         a = testing.shaped_random((2, 3), xp, dtype)
-        return a.ptp()
+        return xp.ptp(a)
 
     @testing.with_requires("numpy>=1.15")
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_allclose()
     def test_ptp_all_keepdims(self, xp, dtype):
         a = testing.shaped_random((2, 3), xp, dtype)
-        return a.ptp(keepdims=True)
+        return xp.ptp(a, keepdims=True)
 
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_allclose()
     def test_ptp_axis_large(self, xp, dtype):
         a = testing.shaped_random((3, 1000), xp, dtype)
-        return a.ptp(axis=0)
+        return xp.ptp(a, axis=0)
 
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_allclose()
     def test_ptp_axis0(self, xp, dtype):
         a = testing.shaped_random((2, 3, 4), xp, dtype)
-        return a.ptp(axis=0)
+        return xp.ptp(a, axis=0)
 
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_allclose()
     def test_ptp_axis1(self, xp, dtype):
         a = testing.shaped_random((2, 3, 4), xp, dtype)
-        return a.ptp(axis=1)
+        return xp.ptp(a, axis=1)
 
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_allclose()
     def test_ptp_axis2(self, xp, dtype):
         a = testing.shaped_random((2, 3, 4), xp, dtype)
-        return a.ptp(axis=2)
+        return xp.ptp(a, axis=2)
 
     @testing.with_requires("numpy>=1.15")
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_allclose()
     def test_ptp_multiple_axes(self, xp, dtype):
         a = testing.shaped_random((2, 3, 4), xp, dtype)
-        return a.ptp(axis=(1, 2))
+        return xp.ptp(a, axis=(1, 2))
 
     @testing.with_requires("numpy>=1.15")
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_allclose()
     def test_ptp_multiple_axes_keepdims(self, xp, dtype):
         a = testing.shaped_random((2, 3, 4), xp, dtype)
-        return a.ptp(axis=(1, 2), keepdims=True)
+        return xp.ptp(a, axis=(1, 2), keepdims=True)
 
     @testing.for_float_dtypes()
     @testing.numpy_cupy_allclose()
     def test_ptp_nan(self, xp, dtype):
         a = xp.array([float("nan"), 1, -1], dtype)
-        return a.ptp()
+        return xp.ptp(a)
 
     @testing.for_complex_dtypes()
     @testing.numpy_cupy_allclose()
     def test_ptp_nan_real(self, xp, dtype):
         a = xp.array([float("nan"), 1, -1], dtype)
-        return a.ptp()
+        return xp.ptp(a)
 
     @testing.for_complex_dtypes()
     @testing.numpy_cupy_allclose()
     def test_ptp_nan_imag(self, xp, dtype):
         a = xp.array([float("nan") * 1.0j, 1.0j, -1.0j], dtype)
-        return a.ptp()
+        return xp.ptp(a)
 
 
 # This class compares CUB results against NumPy's