Use CuPy array in pip_bitmap_column_to_binary_array (#1418)

The performance regression in #1413 is due to numba's `DeviceNDArray`
is slow in slicing. Recent cudf's DataFrame construction has simplified the construction and delegated construction
to similar logic that handles __cuda_array_interface__. Since the construction involves slicing the array, we need
this operation to be fast. In that sense, we should cast the use of DeviceNDArray to cupy array to support fast
slicing.

closes #1413

Authors:
  - Michael Wang (https://github.com/isVoid)

Approvers:
  - Paul Taylor (https://github.com/trxcllnt)
  - Mark Harris (https://github.com/harrism)
  - https://github.com/jakirkham

URL: #1418

python/cuspatial/cuspatial/tests/spatial/join/test_point_in_polygon.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2019, NVIDIA CORPORATION.
+# Copyright (c) 2019-2024, NVIDIA CORPORATION.
 
 import numpy as np
 
@@ -245,17 +245,17 @@ def test_pip_bitmap_column_to_binary_array():
     expected = np.array(
         [[0, 0, 0, 0], [1, 1, 0, 1], [0, 0, 1, 1], [1, 0, 0, 1]], dtype="int8"
     )
-    np.testing.assert_array_equal(got.copy_to_host(), expected)
+    np.testing.assert_array_equal(got.get(), expected)
 
     col = cudf.Series([], dtype="i8")._column
     got = pip_bitmap_column_to_binary_array(col, width=0)
     expected = np.array([], dtype="int8").reshape(0, 0)
-    np.testing.assert_array_equal(got.copy_to_host(), expected)
+    np.testing.assert_array_equal(got.get(), expected)
 
     col = cudf.Series([None, None], dtype="float64")._column
     got = pip_bitmap_column_to_binary_array(col, width=0)
     expected = np.array([], dtype="int8").reshape(2, 0)
-    np.testing.assert_array_equal(got.copy_to_host(), expected)
+    np.testing.assert_array_equal(got.get(), expected)
 
     col = cudf.Series(
         [
@@ -273,9 +273,9 @@ def test_pip_bitmap_column_to_binary_array():
         ],
         dtype="int8",
     )
-    np.testing.assert_array_equal(got.copy_to_host(), expected)
+    np.testing.assert_array_equal(got.get(), expected)
 
     col = cudf.Series([0, 0, 0])._column
     got = pip_bitmap_column_to_binary_array(col, width=3)
     expected = np.array([[0, 0, 0], [0, 0, 0], [0, 0, 0]], dtype="int8")
-    np.testing.assert_array_equal(got.copy_to_host(), expected)
+    np.testing.assert_array_equal(got.get(), expected)

python/cuspatial/cuspatial/utils/join_utils.py

@@ -1,7 +1,8 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 import operator
 
+import cupy as cp
 from numba import cuda
 
 import rmm
@@ -27,7 +28,7 @@ def binarize(in_col, out, width):
 
 def apply_binarize(in_col, width):
     buf = rmm.DeviceBuffer(size=(in_col.size * width))
-    out = cuda.as_cuda_array(buf).view("int8").reshape((in_col.size, width))
+    out = cp.asarray(buf).view("int8").reshape((in_col.size, width))
     if out.size > 0:
         out[:] = 0
         binarize.forall(out.size)(in_col, out, width)