Add new multithreaded TwoQubitPeepholeOptimization pass

crates/accelerate/src/consolidate_blocks.rs

@@ -117,7 +117,7 @@ pub(crate) fn consolidate_blocks(
                 dag.get_qargs(inst.qubits),
             ) {
                 all_block_gates.insert(inst_node);
-                let matrix = match get_matrix_from_inst(py, inst) {
+                let matrix = match get_matrix_from_inst(inst) {
                     Ok(mat) => mat,
                     Err(_) => continue,
                 };
@@ -215,7 +215,7 @@ pub(crate) fn consolidate_blocks(
                 *block_qargs.iter().min().unwrap(),
                 *block_qargs.iter().max().unwrap(),
             ];
-            let matrix = blocks_to_matrix(py, dag, &block, block_index_map).ok();
+            let matrix = blocks_to_matrix(dag, &block, block_index_map).ok();
             if let Some(matrix) = matrix {
                 let num_basis_gates = match decomposer {
                     DecomposerType::TwoQubitBasis(ref decomp) => {
@@ -279,7 +279,7 @@ pub(crate) fn consolidate_blocks(
                     first_qubits,
                 )
             {
-                let matrix = match get_matrix_from_inst(py, first_inst) {
+                let matrix = match get_matrix_from_inst(first_inst) {
                     Ok(mat) => mat,
                     Err(_) => continue,
                 };
@@ -303,7 +303,7 @@ pub(crate) fn consolidate_blocks(
                     already_in_block = true;
                 }
                 let gate = dag[*node].unwrap_operation();
-                let operator = match get_matrix_from_inst(py, gate) {
+                let operator = match get_matrix_from_inst(gate) {
                     Ok(mat) => mat,
                     Err(_) => {
                         // Set this to skip this run because we can't compute the matrix of the

crates/accelerate/src/convert_2q_block_matrix.rs

@@ -31,21 +31,23 @@ use crate::euler_one_qubit_decomposer::matmul_1q;
 use crate::QiskitError;
 
 #[inline]
-pub fn get_matrix_from_inst(py: Python, inst: &PackedInstruction) -> PyResult<Array2<Complex64>> {
+pub fn get_matrix_from_inst(inst: &PackedInstruction) -> PyResult<Array2<Complex64>> {
     if let Some(mat) = inst.op.matrix(inst.params_view()) {
         Ok(mat)
     } else if inst.op.try_standard_gate().is_some() {
         Err(QiskitError::new_err(
             "Parameterized gates can't be consolidated",
         ))
     } else if let OperationRef::Gate(gate) = inst.op.view() {
-        Ok(QI_OPERATOR
-            .get_bound(py)
-            .call1((gate.gate.clone_ref(py),))?
-            .getattr(intern!(py, "data"))?
-            .extract::<PyReadonlyArray2<Complex64>>()?
-            .as_array()
-            .to_owned())
+        Python::with_gil(|py| {
+            Ok(QI_OPERATOR
+                .get_bound(py)
+                .call1((gate.gate.clone_ref(py),))?
+                .getattr(intern!(py, "data"))?
+                .extract::<PyReadonlyArray2<Complex64>>()?
+                .as_array()
+                .to_owned())
+        })
     } else {
         Err(QiskitError::new_err(
             "Can't compute matrix of non-unitary op",
@@ -55,7 +57,6 @@ pub fn get_matrix_from_inst(py: Python, inst: &PackedInstruction) -> PyResult<Ar
 
 /// Return the matrix Operator resulting from a block of Instructions.
 pub fn blocks_to_matrix(
-    py: Python,
     dag: &DAGCircuit,
     op_list: &[NodeIndex],
     block_index_map: [Qubit; 2],
@@ -73,7 +74,7 @@ pub fn blocks_to_matrix(
     let mut output_matrix: Option<Array2<Complex64>> = None;
     for node in op_list {
         let inst = dag[*node].unwrap_operation();
-        let op_matrix = get_matrix_from_inst(py, inst)?;
+        let op_matrix = get_matrix_from_inst(inst)?;
         match dag
             .get_qargs(inst.qubits)
             .iter()

crates/accelerate/src/high_level_synthesis.rs

@@ -698,7 +698,7 @@ fn extract_definition(
                         "cx".to_string(),
                         aview2(&CX_GATE),
                         1.0,
-                        "U",
+                        EulerBasis::U,
                         None,
                     )?;
                     let two_qubit_sequence =

crates/accelerate/src/lib.rs

@@ -52,6 +52,7 @@ pub mod synthesis;
 pub mod target_transpiler;
 pub mod twirling;
 pub mod two_qubit_decompose;
+pub mod two_qubit_peephole;
 pub mod uc_gate;
 pub mod unitary_synthesis;
 pub mod vf2_layout;

crates/accelerate/src/two_qubit_decompose.rs

@@ -1273,9 +1273,9 @@ type TwoQubitSequenceVec = Vec<(Option<StandardGate>, SmallVec<[f64; 3]>, SmallV
 #[derive(Clone, Debug)]
 #[pyclass(sequence)]
 pub struct TwoQubitGateSequence {
-    gates: TwoQubitSequenceVec,
+    pub gates: TwoQubitSequenceVec,
     #[pyo3(get)]
-    global_phase: f64,
+    pub global_phase: f64,
 }
 
 impl TwoQubitGateSequence {
@@ -1383,6 +1383,12 @@ impl TwoQubitBasisDecomposer {
         __num_basis_gates(self.basis_decomposer.b, self.basis_fidelity, u)
     }
 
+    /// Is the gate super controlled
+    #[inline]
+    pub fn super_controlled(&self) -> bool {
+        self.super_controlled
+    }
+
     fn decomp1_inner(
         &self,
         target: &TwoQubitWeylDecomposition,
@@ -1743,7 +1749,7 @@ impl TwoQubitBasisDecomposer {
         gate: String,
         gate_matrix: ArrayView2<Complex64>,
         basis_fidelity: f64,
-        euler_basis: &str,
+        euler_basis: EulerBasis,
         pulse_optimize: Option<bool>,
     ) -> PyResult<Self> {
         let ipz: ArrayView2<Complex64> = aview2(&IPZ);
@@ -1851,7 +1857,7 @@ impl TwoQubitBasisDecomposer {
         Ok(TwoQubitBasisDecomposer {
             gate,
             basis_fidelity,
-            euler_basis: EulerBasis::__new__(euler_basis)?,
+            euler_basis,
             pulse_optimize,
             basis_decomposer,
             super_controlled,
@@ -2021,7 +2027,7 @@ impl TwoQubitBasisDecomposer {
             gate,
             gate_matrix.as_array(),
             basis_fidelity,
-            euler_basis,
+            EulerBasis::__new__(euler_basis)?,
             pulse_optimize,
         )
     }
@@ -2319,8 +2325,13 @@ fn two_qubit_decompose_up_to_diagonal(
     let (su4, phase) = u4_to_su4(mat_arr);
     let mut real_map = real_trace_transform(su4.view());
     let mapped_su4 = real_map.dot(&su4.view());
-    let decomp =
-        TwoQubitBasisDecomposer::new_inner("cx".to_string(), aview2(&CX_GATE), 1.0, "U", None)?;
+    let decomp = TwoQubitBasisDecomposer::new_inner(
+        "cx".to_string(),
+        aview2(&CX_GATE),
+        1.0,
+        EulerBasis::__new__("U")?,
+        None,
+    )?;
 
     let circ_seq = decomp.call_inner(mapped_su4.view(), None, true, None)?;
     let circ = CircuitData::from_standard_gates(
@@ -2505,7 +2516,7 @@ impl<'py> IntoPyObject<'py> for RXXEquivalent {
 #[derive(Clone, Debug)]
 #[pyclass(module = "qiskit._accelerate.two_qubit_decompose", subclass)]
 pub struct TwoQubitControlledUDecomposer {
-    rxx_equivalent_gate: RXXEquivalent,
+    pub rxx_equivalent_gate: RXXEquivalent,
     euler_basis: EulerBasis,
     #[pyo3(get)]
     scale: f64,