Add syntactic sugar for squin.two_qubit_pauli_channel

[jasonhan3]

Summary

Closes #341.

Adds syntax sugar for SQUIN two-qubit Pauli channels so users can pass Cirq-style
Pauli product probability dictionaries instead of manually constructing the full
15-entry probability list.

@squin.kernel
def main():
    q = squin.qalloc(2)
    squin.two_qubit_pauli_channel({"XX": 0.1, "YY": 0.2, "ZZ": 0.15}, q[0], q[1])

Omitted Pauli products default to 0.0.

The existing 15-entry probability list form and broadcast-style control/target lists continue to work.

Changes

• Add custom Python-call lowering for squin.noise.stmts.TwoQubitPauliChannel
• Support dict literals keyed by Pauli products such as "XX", "YY", and "ZZ"
• Preserve the existing 15-entry list/IList probability form
• Preserve broadcast controls/targets typed as IList[Qubit, N]
• Allow single-qubit controls/targets in the simple public syntax by wrapping them into one-element ILists during lowering
• Preserve Python list literals for controls/targets as ILists for broadcast use
• Centralize PAULI_PRODUCT_ORDER and reuse it from Cirq lowering/emission and fidelity analysis
• Correct the two-qubit Pauli probability operand type to IList[float, Literal[15]]

Verification

uv run --no-sync pytest test/squin/test_two_qubit_pauli_channel_sugar.py test/stim/passes/test_squin_noise_to_stim.py test/cirq_utils/test_squin_noise_to_cirq.py test/analysis/fidelity/test_fidelity.py
uv run --no-sync ruff check src/bloqade/squin/noise/stmts.py src/bloqade/squin/noise/_interface.py src/bloqade/squin/noise/__init__.py src/bloqade/squin/stdlib/simple/noise.py src/bloqade/cirq_utils/emit/noise.py src/bloqade/cirq_utils/lowering.py src/bloqade/squin/analysis/fidelity/impls.py test/squin/test_two_qubit_pauli_channel_sugar.py

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[github-actions]

☂️ Python Coverage

current status: ✅

Overall Coverage

Lines	Covered	Coverage	Threshold	Status
12290	10980	89%	0%	🟢

New Files

No new covered files...

Modified Files

File	Coverage	Status
src/bloqade/cirq_utils/emit/noise.py	83%	🟢
src/bloqade/cirq_utils/lowering.py	92%	🟢
src/bloqade/squin/analysis/fidelity/impls.py	86%	🟢
src/bloqade/squin/noise/_init_.py	100%	🟢
src/bloqade/squin/noise/_interface.py	100%	🟢
src/bloqade/squin/noise/stmts.py	100%	🟢
src/bloqade/squin/stdlib/simple/noise.py	77%	🟢
TOTAL	91%	🟢

updated for commit: cd9e0f5 by action🐍

[david-pl]

I'm not so sure about this change. It's a little weird because it lowers an ast.Dict, when kirin actually doesn't support it.

Something like this errors

@squin.kernel
def main():
    q = squin.qalloc(2)
    probs = {"XX": 0.1, "YY": 0.2, "ZZ": 0.15}
    squin.two_qubit_pauli_channel(probs, q[0], q[1])

with BuildError: Cannot lower Dict node: <ast.Dict object at 0x10ae5a4d0>

What's worse, however, is this silent footgun:

from bloqade import squin

probs = {"XX": 0.1, "YY": 0.2, "ZZ": 0.15}

@squin.kernel
def main():
    q = squin.qalloc(2)
    squin.two_qubit_pauli_channel(probs, q[0], q[1])

This seemingly works fine, but inspecting the IR, you see that it silently inserts the actual dict object and uses this as argument

func.func @main() -> !py.NoneType {
  ^0(%main_self):
  │ %0 = py.constant.constant 2 : !py.int
  │ %q = func.invoke qalloc(%0) : !py.IList[!py.Qubit, !Any] maybe_pure=False
  │ %1 = py.constant.constant {'XX': 0.1, 'YY': 0.2, 'ZZ': 0.15} : !Bottom
  │ %2 = py.constant.constant 0 : !py.int
  │ %3 = py.indexing.getitem(%q, %2) : !py.Qubit
  │ %4 = py.ilist.new(values=(%3)){elem_type=!py.Qubit} : !py.IList[!py.Qubit, Literal(1,int)]
  │ %5 = py.constant.constant 1 : !py.int
  │ %6 = py.indexing.getitem(%q, %5) : !py.Qubit
  │ %7 = py.ilist.new(values=(%6)){elem_type=!py.Qubit} : !py.IList[!py.Qubit, Literal(1,int)]
  │      squin.noise.two_qubit_pauli_channel(probabilities=%1, controls=%4, targets=%7)
  │ %8 = func.const.none() : !py.NoneType
  │      func.return %8

This will only fail later with spurious rewrite failures or an unhelpful error message.

[david-pl]

Er, what's this? This just aliases the wrapper right? That shouldn't be the case.

[david-pl]

Oh, I see, you can't support this signature in a kernel, right? Hm, well but this makes two_qubit_pauli_channel subtly different from all other channels and gates. Not sure if that's so great...