Implement asynchronous fill method using dpctl kernels (#2055)

* Enhance `dpnp_array.fill` method

Leverages dpctl's strided fill and memset for setting contiguous memory to 0

* Fix missing disclaimer in dpnp_arraycreation.py

* Import `dpnp_array` directly

* Skip `test_fill_with_numpy_scalar_ndarray`

New fill implementation does not permit NumPy array values, consistent with fill_diagonal

* Add dependencies to zeros and full kernels in `dpnp_fill`

* Remove redundant validation of first `dpnp_fill` argument

* Improve `dpnp_fill` array/scalar path logic

* Disallow inputs to `dpnp_fill` on separate queues

* Adjust skip message for `test_fill_with_numpy_scalar_ndarray`

* Tweak error messages in `dpnp_fill`

* Add tests for new `fill` method

* Update docstring for `fill` method

* Fix pre-commit in cupy fill tests

* Change `asarray` to `astype` in `dpnp_fill`

NumPy arrays are no longer permitted and queue coercion does not occur in the `fill` method, so `astype` is sufficient

* Expand TEST_SCOPE to include `test_fill.py`

* Remove redundant check from `dpnp_fill`

* Use `_cast_fill_val` private function from `dpctl.tensor._ctors`

* Add tests per PR review by @antonwolfy

* Improve validation of `val` for `fill` method

* Add to permit NumPy bools as `dpnp_fill` scalar fill values

* Use `dpnp.bool` in `dpnp_fill` and make `isinstance` check more efficient

* Replace branching for `fill` scalar type with `_cast_fill_value`

* Add additional tests for `fill`

`test_fill_non_scalar` now checks for strings and `test_fill_bool` added to verify bools are properly cast to 1

---------

Co-authored-by: Anton <[email protected]>

Author: ndgrigorian

.github/workflows/conda-package.yml

@@ -29,6 +29,7 @@ env:
       test_copy.py
       test_counting.py
       test_fft.py
+      test_fill.py
       test_flat.py
       test_histogram.py
       test_indexing.py

dpnp/dpnp_algo/dpnp_arraycreation.py

@@ -1,3 +1,29 @@
+# -*- coding: utf-8 -*-
+# *****************************************************************************
+# Copyright (c) 2016-2024, 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.
+# *****************************************************************************
+
 import math
 import operator
 

dpnp/dpnp_algo/dpnp_fill.py

@@ -0,0 +1,78 @@
+# -*- coding: utf-8 -*-
+# *****************************************************************************
+# Copyright (c) 2016-2024, 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.
+# *****************************************************************************
+
+from numbers import Number
+
+import dpctl.tensor as dpt
+import dpctl.utils as dpu
+from dpctl.tensor._ctors import _cast_fill_val
+from dpctl.tensor._tensor_impl import (
+    _copy_usm_ndarray_into_usm_ndarray,
+    _full_usm_ndarray,
+    _zeros_usm_ndarray,
+)
+
+import dpnp
+
+
+def dpnp_fill(arr, val):
+    arr = dpnp.get_usm_ndarray(arr)
+    exec_q = arr.sycl_queue
+
+    # if val is an array, process it
+    if dpnp.is_supported_array_type(val):
+        val = dpnp.get_usm_ndarray(val)
+        if val.shape != ():
+            raise ValueError("`val` must be a scalar or 0D-array")
+        if dpu.get_execution_queue((exec_q, val.sycl_queue)) is None:
+            raise dpu.ExecutionPlacementError(
+                "Input arrays have incompatible queues."
+            )
+        a_val = dpt.astype(val, arr.dtype)
+        a_val = dpt.broadcast_to(a_val, arr.shape)
+        _manager = dpu.SequentialOrderManager[exec_q]
+        dep_evs = _manager.submitted_events
+        h_ev, c_ev = _copy_usm_ndarray_into_usm_ndarray(
+            src=a_val, dst=arr, sycl_queue=exec_q, depends=dep_evs
+        )
+        _manager.add_event_pair(h_ev, c_ev)
+        return
+    elif not isinstance(val, (Number, dpnp.bool)):
+        raise TypeError(
+            f"array cannot be filled with `val` of type {type(val)}"
+        )
+    val = _cast_fill_val(val, arr.dtype)
+
+    _manager = dpu.SequentialOrderManager[exec_q]
+    dep_evs = _manager.submitted_events
+
+    # can leverage efficient memset when val is 0
+    if arr.flags["FORC"] and val == 0:
+        h_ev, zeros_ev = _zeros_usm_ndarray(arr, exec_q, depends=dep_evs)
+        _manager.add_event_pair(h_ev, zeros_ev)
+    else:
+        h_ev, fill_ev = _full_usm_ndarray(val, arr, exec_q, depends=dep_evs)
+        _manager.add_event_pair(h_ev, fill_ev)

dpnp/dpnp_array.py

@@ -928,13 +928,16 @@ def fill(self, value):
         """
         Fill the array with a scalar value.
 
+        For full documentation refer to :obj:`numpy.ndarray.fill`.
+
         Parameters
         ----------
-        value : scalar
+        value : {dpnp.ndarray, usm_ndarray, scalar}
             All elements of `a` will be assigned this value.
 
         Examples
         --------
+        >>> import dpnp as np
         >>> a = np.array([1, 2])
         >>> a.fill(0)
         >>> a
@@ -946,8 +949,10 @@ def fill(self, value):
 
         """
 
-        for i in range(self.size):
-            self.flat[i] = value
+        # lazy import avoids circular imports
+        from .dpnp_algo.dpnp_fill import dpnp_fill
+
+        dpnp_fill(self, value)
 
     @property
     def flags(self):

tests/test_fill.py

@@ -0,0 +1,87 @@
+import dpctl
+import numpy as np
+import pytest
+from dpctl.utils import ExecutionPlacementError
+from numpy.testing import assert_array_equal
+
+import dpnp as dnp
+
+
+@pytest.mark.parametrize(
+    "val, error",
+    [
+        pytest.param(dnp.ones(2, dtype="i4"), ValueError, id="array"),
+        pytest.param(dict(), TypeError, id="dictionary"),
+        pytest.param("0", TypeError, id="string"),
+    ],
+)
+def test_fill_non_scalar(val, error):
+    a = dnp.ones(5, dtype="i4")
+    with pytest.raises(error):
+        a.fill(val)
+
+
+def test_fill_compute_follows_data():
+    q1 = dpctl.SyclQueue()
+    q2 = dpctl.SyclQueue()
+
+    a = dnp.ones(5, dtype="i4", sycl_queue=q1)
+    val = dnp.ones((), dtype=a.dtype, sycl_queue=q2)
+
+    with pytest.raises(ExecutionPlacementError):
+        a.fill(val)
+
+
+def test_fill_strided_array():
+    a = dnp.zeros(100, dtype="i4")
+    b = a[::-2]
+
+    expected = dnp.tile(dnp.asarray([0, 1], dtype=a.dtype), 50)
+
+    b.fill(1)
+    assert_array_equal(b, 1)
+    assert_array_equal(a, expected)
+
+
+@pytest.mark.parametrize("order", ["C", "F"])
+def test_fill_strided_2d_array(order):
+    a = dnp.zeros((10, 10), dtype="i4", order=order)
+    b = a[::-2, ::2]
+
+    expected = dnp.copy(a)
+    expected[::-2, ::2] = 1
+
+    b.fill(1)
+    assert_array_equal(b, 1)
+    assert_array_equal(a, expected)
+
+
+@pytest.mark.parametrize("order", ["C", "F"])
+def test_fill_memset(order):
+    a = dnp.ones((10, 10), dtype="i4", order=order)
+    a.fill(0)
+
+    assert_array_equal(a, 0)
+
+
+def test_fill_float_complex_to_int():
+    a = dnp.ones((10, 10), dtype="i4")
+
+    a.fill(complex(2, 0))
+    assert_array_equal(a, 2)
+
+    a.fill(float(3))
+    assert_array_equal(a, 3)
+
+
+def test_fill_complex_to_float():
+    a = dnp.ones((10, 10), dtype="f4")
+
+    a.fill(complex(2, 0))
+    assert_array_equal(a, 2)
+
+
+def test_fill_bool():
+    a = dnp.full(5, fill_value=7, dtype="i4")
+    a.fill(True)
+    assert_array_equal(a, 1)

tests/third_party/cupy/core_tests/test_ndarray_copy_and_view.py

@@ -274,9 +274,7 @@ def test_fill(self, xp, dtype):
         a.fill(1)
         return a
 
-    @testing.with_requires("numpy>=1.24.0")
-    @testing.for_all_dtypes_combination(("dtype1", "dtype2"))
-    @testing.numpy_cupy_array_equal(accept_error=ComplexWarning)
+    @pytest.mark.skip("Numpy allows Numpy scalar arrays as fill value")
     def test_fill_with_numpy_scalar_ndarray(self, xp, dtype1, dtype2):
         a = testing.shaped_arange((2, 3, 4), xp, dtype1)
         a.fill(numpy.ones((), dtype=dtype2))