Add has_stabilizer_effect protocol and support native Clifford simulator gates #2760
Conversation
…rted_gates list and delete the old method
|
Only adds support for gates that can be run by CliffordState. Does not, yet, try to decompose. |
|
Ping! |
| def test_y_stabilizer(): | ||
| gate = cirq.Y | ||
| assert cirq.has_stabilizer_effect(gate) | ||
| assert not cirq.has_stabilizer_effect(gate**0.5) |
There was a problem hiding this comment.
Square root of Y is a stabilizer gate.
There was a problem hiding this comment.
As I mentioned in #2760 (comment), I'm just adding support for the operations that CliffordState supports right now (just like it was before this) and does not try to decompose. Y**0.5 is a stabilizer gate and so is iSwap but just restricting myself to the original 7 gates before #2803 adds all of them to CliffordState since right now the simulator would just balk at Y**0.5.
If you feel we should add all these gates to the protocol at once, I can wait for the #2803 to go in. I can then have the protocol just do what is_supported_gate doing now as described in https://github.com/quantumlib/Cirq/pull/2803/files#r390152850. It would save on redundant work and definitely seems like a better option than having to define _has_stabilizer_effect_ on all gates.
| assert cirq.has_stabilizer_effect(gate) | ||
| assert not cirq.has_stabilizer_effect(gate**0.5) | ||
| assert cirq.has_stabilizer_effect(gate**0) | ||
| assert not cirq.has_stabilizer_effect(gate**-0.5) |
There was a problem hiding this comment.
Should be true not false.
| def test_x_stabilizer(): | ||
| gate = cirq.X | ||
| assert cirq.has_stabilizer_effect(gate) | ||
| assert not cirq.has_stabilizer_effect(gate**0.5) |
There was a problem hiding this comment.
Should be true not false.
| assert cirq.has_stabilizer_effect(gate) | ||
| assert not cirq.has_stabilizer_effect(gate**0.5) | ||
| assert cirq.has_stabilizer_effect(gate**0) | ||
| assert not cirq.has_stabilizer_effect(gate**-0.5) |
There was a problem hiding this comment.
Should be true not false.
| def test_z_stabilizer(): | ||
| gate = cirq.Z | ||
| assert cirq.has_stabilizer_effect(gate) | ||
| assert cirq.has_stabilizer_effect(gate**0.5) |
There was a problem hiding this comment.
Should be true not false.
| assert cirq.has_stabilizer_effect(gate) | ||
| assert cirq.has_stabilizer_effect(gate**0.5) | ||
| assert cirq.has_stabilizer_effect(gate**0) | ||
| assert cirq.has_stabilizer_effect(gate**-0.5) |
There was a problem hiding this comment.
Should be true not false.
| foo = sympy.Symbol('foo') | ||
| assert not cirq.has_stabilizer_effect(gate**foo) | ||
|
|
||
|
|
There was a problem hiding this comment.
You should also test that measurement is a stabilizer effect, and that gates like CCZ and the depolarizing channel are not.
Strilanc
added
the
automerge
CirqBot
added
the
front_of_queue_automerge
CirqBot
removed
automerge
Follow up to #2600 and part of #2423