Refactor greedy pauli simp #1611
Conversation
This reverts commit 9e9a2fd.
…)`` doesn't work with wasm
tket/test/src/test_GreedyPauli.cpp
Outdated
| @@ -74,9 +44,9 @@ SCENARIO("Clifford synthesis") { | |||
| } | |||
| GIVEN("2Q Simple Clifford") { | |||
| Circuit circ(2); | |||
| circ.add_op<unsigned>(OpType::Y, {0}); | |||
| // circ.add_op<unsigned>(OpType::Y, {0}); | |||
There was a problem hiding this comment.
Can we add these back in to the tests?
| /** | ||
| * @brief Conditional Pauli rotation | ||
| */ | ||
| class ConditionalPauliRotation : public PauliRotation { |
There was a problem hiding this comment.
Do we only have ConditionalPauliRotation and not ConditionalClifford? My concern is that if a user provides a conditional Clifford operation, and not a conditional rotation, then each conditional Clifford adds new branching as a series of ConditionalPauliRotation? Is this how they would be handled currently?
| SpPauliStabiliser qpt = cliff_.get_row_product(SpPauliStabiliser(qpm)); | ||
| auto [pauli_dense, theta] = dense_pauli(qpt, n_qubits_, angle); | ||
| PauliNode_ptr node; | ||
| if (conditional) { |
There was a problem hiding this comment.
Ah Ok. So on conditional = true, regardless of whether it's a Clifford angle or not, we add it to the circuit as a Pauli Exponential (or potentially one with a Clifford angle), and they get resolved at synthesis. So if it happens to have a Clifford angle, it doesn't matter, we can still use the pre-existing logic, it just turns out eventually it will get synthesised with only Cliffords coincidentially.
That's a really nice solution!
| /** | ||
| * @brief Black box node for classical Ops | ||
| */ | ||
| class ClassicalNode : public PauliNode { |
There was a problem hiding this comment.
Would this cover all the ops for pytket-quantinuum? i.e.:
_ADDITIONAL_GATES = {
OpType.Reset,
OpType.Measure,
OpType.Barrier,
OpType.RangePredicate,
OpType.MultiBit,
OpType.ExplicitPredicate,
OpType.ExplicitModifier,
OpType.SetBits,
OpType.CopyBits,
OpType.ClassicalExpBox,
OpType.WASM,
}
|
|
||
| unsigned ACPairNode::tqe_cost() const { | ||
| return static_cast<unsigned>( | ||
| 1.5 * (n_anti_commute_entries_ - 1) + n_commute_entries_); |
There was a problem hiding this comment.
Please could you add a comment explaining the rationale behind this cost
Description
Supports mid-circuit measurements, resets, conditionals and classical gates.
A modified version of the existing Pauli graph is needed for this feature.
Related issues
Please mention any github issues addressed by this PR.
Checklist