Problem Statement
There are oppurtunities to reduce the number of 1Q gates after we compile to the QSystem native gateset. These are fairly simple to add in.
This would involve calling the pytket SquashRzPhasedX pass in QSystemPass after the conversion to the hardware native gate set is performed. This should clean up some redundant single qubit gates.
We did this optimisation for pytket-quantinuum compilation. It shouldn't impact fault-tolerance of circuits as 2Q gates are preserved... Therefore errors will not spread differently in the transformed circuit.
Proposed Solution
It should be easy to make a call to Tket1Pass from the tket1-passes project in the implementation of QSystemPass.
Something like
let squash_pass_json = serde_json::to_string(&tket_json_rs::pass::BasePass::StandardPass {
pass: tket_json_rs::pass::standard::StandardPass::SquashRzPhasedX,
})
Tket1Pass::run_from_json(&squash_pass_json, &mut circuit_ptr).unwrap();
// Add appropriate decoding -> HUGRComponent
tket-qsystem
Effort Estimate
low
Additional Context
Here is a concrete example using pytket where such optimisations reduce the number of 1q gates
from pytket import Circuit, OpType
from pytket.passes import AutoRebase, SquashRzPhasedX
from pytket.circuit.display import render_circuit_jupyter as draw
# Define a circuit
circ = Circuit(3).H(0).H(1).H(2).CX(0, 1).CX(1, 2).Rz(1/4, 2)
draw(circ)
We can now convert the circuit to {ZZPhase, Rz, PhasedX} with AutoRebase
AutoRebase({OpType.ZZPhase, OpType.Rz, OpType.PhasedX}).apply(circ)
print("#Rz gates:", circ.n_gates_of_type(OpType.Rz))
print("#PhasedX gates:", circ.n_gates_of_type(OpType.PhasedX))
draw(circ)#Rz gates: 8
#PhasedX gates: 7
Now if we apply 1Q squashing via SquashRzPhasedX
SquashRzPhasedX().apply(circ)
print("#Rz gates:", circ.n_gates_of_type(OpType.Rz))
print("#PhasedX gates:", circ.n_gates_of_type(OpType.PhasedX))
draw(circ)#Rz gates: 3
#PhasedX gates: 5
Problem Statement
There are oppurtunities to reduce the number of 1Q gates after we compile to the QSystem native gateset. These are fairly simple to add in.
This would involve calling the pytket SquashRzPhasedX pass in
QSystemPassafter the conversion to the hardware native gate set is performed. This should clean up some redundant single qubit gates.We did this optimisation for pytket-quantinuum compilation. It shouldn't impact fault-tolerance of circuits as 2Q gates are preserved... Therefore errors will not spread differently in the transformed circuit.
Proposed Solution
It should be easy to make a call to
Tket1Passfrom the tket1-passes project in the implementation ofQSystemPass.Something like
Component
tket-qsystem
Effort Estimate
low
Additional Context
Here is a concrete example using pytket where such optimisations reduce the number of 1q gates
We can now convert the circuit to
{ZZPhase, Rz, PhasedX}withAutoRebaseNow if we apply 1Q squashing via
SquashRzPhasedX