Add more default trials to sabre layout

crates/accelerate/src/sabre/layout.rs

@@ -62,6 +62,70 @@ pub fn sabre_layout_and_routing(
         &target,
         run_in_parallel,
     ));
+    starting_layouts.push(
+        (0..target.neighbors.num_qubits() as u32)
+            .map(Some)
+            .collect(),
+    );
+    starting_layouts.push(
+        (0..target.neighbors.num_qubits() as u32)
+            .rev()
+            .map(Some)
+            .collect(),
+    );
+    // This layout targets the largest ring on an IBM eagle device. It has been
+    // shown to have good results on some circuits targeting these backends. In
+    // all other cases this is no different from an additional random trial,
+    // see: https://xkcd.com/221/
+    if target.neighbors.num_qubits() == 127 {
+        starting_layouts.push(
+            [
+                0, 1, 2, 3, 4, 5, 6, 15, 22, 23, 24, 25, 34, 43, 42, 41, 40, 53, 60, 59, 61, 62,
+                72, 81, 80, 79, 78, 91, 98, 99, 100, 101, 102, 103, 92, 83, 82, 84, 85, 86, 73, 66,
+                65, 64, 63, 54, 45, 44, 46, 47, 35, 28, 29, 27, 26, 16, 7, 8, 9, 10, 11, 12, 13,
+                17, 30, 31, 32, 36, 51, 50, 49, 48, 55, 68, 67, 69, 70, 74, 89, 88, 87, 93, 106,
+                105, 104, 107, 108, 112, 126, 125, 124, 123, 122, 111, 121, 120, 119, 118, 110,
+                117, 116, 115, 114, 113, 109, 96, 97, 95, 94, 90, 75, 76, 77, 71, 58, 57, 56, 52,
+                37, 38, 39, 33, 20, 21, 19, 18, 14,
+            ]
+            .into_iter()
+            .map(Some)
+            .collect(),
+        );
+    } else if target.neighbors.num_qubits() == 133 {
+        // Same for IBM Heron 133 qubit devices. This is the ring computed by using rustworkx's
+        // max(simple_cycles(graph), key=len) on the connectivity graph.
+        starting_layouts.push(
+            [
+                108, 107, 94, 88, 89, 90, 75, 71, 70, 69, 56, 50, 51, 52, 37, 33, 32, 31, 18, 12,
+                11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0, 15, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28,
+                29, 36, 48, 47, 46, 45, 44, 43, 42, 41, 40, 39, 38, 53, 57, 58, 59, 60, 61, 62, 63,
+                64, 65, 66, 67, 74, 86, 85, 84, 83, 82, 81, 80, 79, 78, 77, 76, 91, 95, 96, 97,
+                110, 116, 117, 118, 119, 120, 111, 101, 102, 103, 104, 105, 112, 124, 125, 126,
+                127, 128, 113, 109,
+            ]
+            .into_iter()
+            .map(Some)
+            .collect(),
+        );
+    } else if target.neighbors.num_qubits() == 156 {
+        // Same for IBM Heron 156 qubit devices. This is the ring computed by using rustworkx's
+        // max(simple_cycles(graph), key=len) on the connectivity graph.
+        starting_layouts.push(
+            [
+                136, 123, 122, 121, 116, 101, 102, 103, 96, 83, 82, 81, 76, 61, 62, 63, 56, 43, 42,
+                41, 36, 21, 22, 23, 16, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 19, 35, 34,
+                33, 32, 31, 30, 29, 28, 27, 26, 25, 37, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55,
+                59, 75, 74, 73, 72, 71, 70, 69, 68, 67, 66, 65, 77, 85, 86, 87, 88, 89, 90, 91, 92,
+                93, 94, 95, 99, 115, 114, 113, 112, 111, 110, 109, 108, 107, 106, 105, 117, 125,
+                126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 139, 155, 154, 153, 152, 151,
+                150, 149, 148, 147, 146, 145, 144, 143,
+            ]
+            .into_iter()
+            .map(Some)
+            .collect(),
+        );
+    }
     let outer_rng = match seed {
         Some(seed) => Pcg64Mcg::seed_from_u64(seed),
         None => Pcg64Mcg::from_entropy(),

qiskit/transpiler/passes/layout/sabre_layout.py

@@ -148,7 +148,10 @@ def __init__(
                 (and ``routing_pass`` is not set) then the number of local
                 physical CPUs will be used as the default value. This option is
                 mutually exclusive with the ``routing_pass`` argument and an error
-                will be raised if both are used.
+                will be raised if both are used. In addition to the specified random
+                trials an additional 3 or 4 trials depending on the ``coupling_map``
+                value are run with common layouts in addition to the random trial
+                count specified by this value.
             skip_routing (bool): If this is set ``True`` and ``routing_pass`` is not used
                 then routing will not be applied to the output circuit.  Only the layout
                 will be set in the property set. This is a tradeoff to run custom

releasenotes/notes/add-more-sabre-trials-9b421f05d2f48d18.yaml

@@ -0,0 +1,12 @@
+---
+features_transpiler:
+  - |
+    The :class:`.SabreLayout` transpiler pass has been updated to run an
+    additional 2 or 3 layout trials by default independently of the
+    ``layout_trials`` keyword argument's value. For all backends a trivial
+    layout and it's reverse are included just like the :class:`.DenseLayout`
+    trial that were added in 1.2.0. In addition if the backends are 127,
+    133, or 156 qubits the largest rings on an IBM backend heavy hex
+    connectivity graph are added. On those backends which are commonly run
+    this can provide a good starting point for some circuits on these
+    backends and for all others it's just an additional random trial.

test/python/transpiler/test_sabre_layout.py

@@ -207,7 +207,7 @@ def test_layout_with_classical_bits(self):
         self.assertIsInstance(res, QuantumCircuit)
         layout = res._layout.initial_layout
         self.assertEqual(
-            [layout[q] for q in qc.qubits], [11, 19, 18, 16, 26, 8, 21, 1, 5, 15, 3, 12, 14, 13]
+            [layout[q] for q in qc.qubits], [2, 0, 5, 1, 7, 3, 14, 6, 9, 8, 10, 11, 4, 12]
         )
 
     # pylint: disable=line-too-long
@@ -271,7 +271,7 @@ def test_layout_many_search_trials(self):
         self.assertIsInstance(res, QuantumCircuit)
         layout = res._layout.initial_layout
         self.assertEqual(
-            [layout[q] for q in qc.qubits], [22, 7, 2, 12, 1, 5, 14, 4, 11, 0, 16, 15, 3, 10]
+            [layout[q] for q in qc.qubits], [0, 12, 7, 3, 6, 11, 1, 10, 4, 9, 2, 5, 13, 8]
         )
 
     def test_support_var_with_rust_fastpath(self):