def _apply_mixture_(self, args: cirq.ApplyMixtureArgs):
zero_left = cirq.slice_for_qubits_equal_to(args.left_axes, 0)
one_left = cirq.slice_for_qubits_equal_to(args.left_axes, 1)
zero_right = cirq.slice_for_qubits_equal_to(
cast(Tuple[int], args.right_axes), 0)
one_right = cirq.slice_for_qubits_equal_to(
cast(Tuple[int], args.right_axes), 1)
args.out_buffer[:] = 0
np.copyto(dst=args.auxiliary_buffer0, src=args.target_tensor)
for kraus_op in [
np.sqrt(0.5) * np.eye(2, dtype=np.complex128),
np.sqrt(0.5) * x
]:
np.copyto(dst=args.target_tensor, src=args.auxiliary_buffer0)
cirq.apply_matrix_to_slices(args.target_tensor,
kraus_op, [zero_left, one_left],
out=args.auxiliary_buffer1)
cirq.apply_matrix_to_slices(
args.auxiliary_buffer1,
np.conjugate(kraus_op),
[zero_right, one_right],
out=args.target_tensor,
)
args.out_buffer += args.target_tensor
return args.out_buffer
def _apply_unitary_(self,
args: cirq.ApplyUnitaryArgs) -> Optional[np.ndarray]:
if cirq.is_parameterized(self):
return NotImplemented
am, bm, cm = (la.expm(-1j * self.exponent *
np.array([[0, w], [w.conjugate(), 0]]))
for w in self.weights)
a1 = args.subspace_index(0b1001)
b1 = args.subspace_index(0b0101)
c1 = args.subspace_index(0b0011)
a2 = args.subspace_index(0b0110)
b2 = args.subspace_index(0b1010)
c2 = args.subspace_index(0b1100)
cirq.apply_matrix_to_slices(args.target_tensor,
am,
slices=[a1, a2],
out=args.available_buffer)
cirq.apply_matrix_to_slices(args.available_buffer,
bm,
slices=[b1, b2],
out=args.target_tensor)
return cirq.apply_matrix_to_slices(args.target_tensor,
cm,
slices=[c1, c2],
out=args.available_buffer)
def _apply_unitary_(self, args: cirq.ApplyUnitaryArgs
) -> Optional[np.ndarray]:
if cirq.is_parameterized(self):
return None
am = cirq.unitary(cirq.Rx(-np.pi * self.exponent * self.weights[0]))
bm = cirq.unitary(cirq.Rx(-np.pi * self.exponent * self.weights[1]))
cm = cirq.unitary(cirq.Rx(-np.pi * self.exponent * self.weights[2]))
a1 = args.subspace_index(0b1001)
b1 = args.subspace_index(0b0101)
c1 = args.subspace_index(0b0011)
a2 = args.subspace_index(0b0110)
b2 = args.subspace_index(0b1010)
c2 = args.subspace_index(0b1100)
cirq.apply_matrix_to_slices(args.target_tensor,
am,
slices=[a1, a2],
out=args.available_buffer)
cirq.apply_matrix_to_slices(args.available_buffer,
bm,
slices=[b1, b2],
out=args.target_tensor)
return cirq.apply_matrix_to_slices(args.target_tensor,
cm,
slices=[c1, c2],
out=args.available_buffer)
def _apply_unitary_(self,
args: 'cirq.ApplyUnitaryArgs') -> Optional[np.ndarray]:
if cirq.is_parameterized(self):
return None
oi = args.subspace_index(0b01)
io = args.subspace_index(0b10)
ii = args.subspace_index(0b11)
if self.theta != 0 or self.zeta != 0 or self.chi != 0:
rx = protocols.unitary(cirq.rx(2 * self.theta))
rz1 = protocols.unitary(cirq.rz(-self.zeta + self.chi))
rz2 = protocols.unitary(cirq.rz(-self.zeta - self.chi))
inner_matrix = rz1 @ rx @ rz2
out = cirq.apply_matrix_to_slices(args.target_tensor,
inner_matrix,
slices=[oi, io],
out=args.available_buffer)
else:
out = args.target_tensor
if self.phi != 0:
out[ii] *= cmath.exp(-1j * self.phi)
if self.gamma != 0:
f = cmath.exp(-1j * self.gamma)
out[oi] *= f
out[io] *= f
out[ii] *= f * f
return out
def test_apply_matrix_to_slices():
# Output is input.
with pytest.raises(ValueError, match='out'):
target = np.eye(2)
_ = cirq.apply_matrix_to_slices(target=target,
matrix=np.eye(2),
out=target,
slices=[0, 1])
# Wrong matrix size.
with pytest.raises(ValueError, match='shape'):
target = np.eye(2)
_ = cirq.apply_matrix_to_slices(target=target,
matrix=np.eye(3),
slices=[0, 1])
# Empty case.
np.testing.assert_allclose(
cirq.apply_matrix_to_slices(target=np.array(range(5)),
matrix=np.eye(0),
slices=[]),
np.array(range(5)),
)
# Middle 2x2 of 4x4 case.
np.testing.assert_allclose(
cirq.apply_matrix_to_slices(target=np.eye(4),
matrix=np.array([[2, 3], [5, 7]]),
slices=[1, 2]),
np.array([[1, 0, 0, 0], [0, 2, 3, 0], [0, 5, 7, 0], [0, 0, 0, 1]]),
)
# Middle 2x2 of 4x4 with opposite order case.
np.testing.assert_allclose(
cirq.apply_matrix_to_slices(target=np.eye(4),
matrix=np.array([[2, 3], [5, 7]]),
slices=[2, 1]),
np.array([[1, 0, 0, 0], [0, 7, 5, 0], [0, 3, 2, 0], [0, 0, 0, 1]]),
)
# Complicated slices of tensor case.
np.testing.assert_allclose(
cirq.apply_matrix_to_slices(
target=np.array(range(8)).reshape((2, 2, 2)),
matrix=np.array([[0, 1], [1, 0]]),
slices=[(0, slice(None), 0), (1, slice(None), 0)],
).reshape((8, )),
[4, 1, 6, 3, 0, 5, 2, 7],
)
# Specified output case.
out = np.zeros(shape=(4, ))
actual = cirq.apply_matrix_to_slices(target=np.array([1, 2, 3, 4]),
matrix=np.array([[2, 3], [5, 7]]),
slices=[1, 2],
out=out)
assert actual is out
np.testing.assert_allclose(actual, np.array([1, 13, 31, 4]))
def _apply_unitary_(self, args: cirq.ApplyUnitaryArgs
) -> Optional[np.ndarray]:
if cirq.is_parameterized(self):
return None
inner_matrix = cirq.unitary(cirq.rx(-2 * np.pi * self.exponent))
a = args.subspace_index(0b0011)
b = args.subspace_index(0b1100)
return cirq.apply_matrix_to_slices(args.target_tensor,
inner_matrix,
slices=[a, b],
out=args.available_buffer)
def _apply_unitary_(self,
args: cirq.ApplyUnitaryArgs) -> Optional[np.ndarray]:
if cirq.is_parameterized(self):
return None
inner_matrix = cirq.unitary(cirq.Ry(-self.exponent * np.pi))
oi = args.subspace_index(0b01)
io = args.subspace_index(0b10)
return cirq.apply_matrix_to_slices(args.target_tensor,
inner_matrix,
slices=[oi, io],
out=args.available_buffer)
def _apply_unitary_(self, args: 'cirq.ApplyUnitaryArgs') -> Optional[np.ndarray]:
if cirq.is_parameterized(self):
return None
if self.theta != 0:
inner_matrix = protocols.unitary(cirq.rx(2 * self.theta))
oi = args.subspace_index(0b01)
io = args.subspace_index(0b10)
out = cirq.apply_matrix_to_slices(
args.target_tensor, inner_matrix, slices=[oi, io], out=args.available_buffer
)
else:
out = args.target_tensor
if self.phi != 0:
ii = args.subspace_index(0b11)
out[ii] *= cmath.exp(-1j * self.phi)
return out
