def test_sum_of_squares(dtype, nelem):
dtype = cudf.dtype(dtype).type
data = gen_rand(dtype, nelem)
sr = Series(data)
df = cudf.DataFrame(sr)
got = sr.sum_of_squares()
got_df = df.sum_of_squares()
expect = (data ** 2).sum()
if cudf.dtype(dtype).kind in {"u", "i"}:
if 0 <= expect <= np.iinfo(dtype).max:
np.testing.assert_array_almost_equal(expect, got)
np.testing.assert_array_almost_equal(expect, got_df.iloc[0])
else:
print("overflow, passing")
else:
np.testing.assert_approx_equal(
expect, got, significant=accuracy_for_dtype[dtype]
)
np.testing.assert_approx_equal(
expect, got_df.iloc[0], significant=accuracy_for_dtype[dtype]
)
