Add numpy backend ccx implementation to improve performance

blueqat/backends/numpy_backend.py

@@ -253,6 +253,24 @@ def gate_s(self, gate, ctx):
             qubits[(i & (1 << target)) != 0] *= 1.j
         return ctx
 
+    def gate_ccz(self, gate, ctx):
+        c1, c2, t = gate.targets
+        qubits = ctx.qubits
+        n_qubits = ctx.n_qubits
+        i = ctx.indices
+        indices = (i & (1 << c1)) != 0
+        indices &= (i & (1 << c2)) != 0
+        indices &= (i & (1 << t)) != 0
+        qubits[indices] *= -1
+        return ctx
+
+    def gate_ccx(self, gate, ctx):
+        c1, c2, t = gate.targets
+        ctx = self.gate_h(HGate(t), ctx)
+        ctx = self.gate_ccz(CCZGate(gate.targets), ctx)
+        ctx = self.gate_h(HGate(t), ctx)
+        return ctx
+
     def gate_u1(self, gate, ctx):
         qubits = ctx.qubits
         n_qubits = ctx.n_qubits

blueqat/gate.py

@@ -227,6 +227,11 @@ def fallback(self, n_qubits):
         ]
 
 
+class CCZGate(Gate):
+    """2-Controlled Z gate"""
+    lowername = "ccz"
+
+
 class U1Gate(OneQubitGate):
     """U1 gate"""
     def __init__(self, targets, lambd, **kwargs):