def test_sparseToDense(self, dtype):
if not cusparse.check_availability('sparseToDense'):
pytest.skip('sparseToDense is not available')
m, n = self.shape
x = scipy.sparse.random(m, n, density=self.density, format=self.format,
dtype=dtype)
if self.format == 'csr':
x = sparse.csr_matrix(x)
elif self.format == 'csc':
x = sparse.csc_matrix(x)
elif self.format == 'coo':
x = sparse.coo_matrix(x)
y = cusparse.sparseToDense(x)
testing.assert_array_equal(x.todense(), y)
def toarray(self, order=None, out=None):
"""Returns a dense matrix representing the same value.
Args:
order ({'C', 'F', None}): Whether to store data in C (row-major)
order or F (column-major) order. Default is C-order.
out: Not supported.
Returns:
cupy.ndarray: Dense array representing the same matrix.
.. seealso:: :meth:`scipy.sparse.csr_matrix.toarray`
"""
order = 'C' if order is None else order.upper()
if self.nnz == 0:
return cupy.zeros(shape=self.shape, dtype=self.dtype, order=order)
if self.dtype.char not in 'fdFD':
return csr2dense(self, order)
x = self.copy()
x.has_canonical_format = False # need to enforce sum_duplicates
x.sum_duplicates()
if (cusparse.check_availability('sparseToDense')
and (not runtime.is_hip or (x.nnz > 0))):
# On HIP, nnz=0 is problematic as of ROCm 4.2.0
y = cusparse.sparseToDense(x)
if order == 'F':
return y
elif order == 'C':
return cupy.ascontiguousarray(y)
else:
raise ValueError('order not understood')
else:
# csr2dense returns F-contiguous array.
if order == 'C':
# To return C-contiguous array, it uses transpose.
return cusparse.csc2dense(x.T).T
elif order == 'F':
return cusparse.csr2dense(x)
else:
raise ValueError('order not understood')