feat: Stim improvements (more ops + state printing) by jake-arkinstall · Pull Request #115 · Quantinuum/selene

jake-arkinstall · 2026-01-26T17:14:57Z

Summary

closes Add state_result support for Stim #54
adds more ops as noted in Improve Clifford check for PhasedX (and NPhasedX) gates tket#1984
removes stim as a submodule, instead depending on it via a cmake dependency

Background

On more ops

Previously we assumed that all clifford operations have angles in multiples of pi/2. There are some cases where angles can be pi/4. To take account of this, we now check angles are approximately pi/4, and use lookups to determine the gates. This is derived from the now-provided gate_definitions.py, which we will use in future as more gates are added.

On state printing

With the QuEST plugin, we support state dumping. This involves writing the internal state of the simulator to a file that can be read by the python frontend. Users can extract reduced density matrices in the space of qubits they have listed, and can also extract statevector distributions and (in the case of pure states) single statevectors.

This PR adds this equivalent functionality for the Stim simulator. Stabilizers are printed to a file as strings, and can then be pulled in and interpreted through the Stim selene plugin. One can read the full generator list directly (not recommended), extract a traced-out generator list, and fetch the density matrix and statevectors with the same interface as in QuEST.

PabloAndresCQ

Thanks Jake, it all looks good. There's a few minor comments that would be good to address, and other than that a few nit-picks that you are welcome to ignore.

I have not tested this yet for my use case, but I'm starting on it now. Feel free to merge though, I'm satisfied with the tests.

PabloAndresCQ · 2026-02-09T18:00:01Z

+        // |   0   | [any] |           | identity regardless of phi
+        // +-------+-------+-----------+
+        // | pi/2  |   0   | V         | pi/2 theta section
+        // | pi/2  | pi/2  | H X       |


Please indicate in all of these rows that you are using gate application order, rather than multiplication order; i.e. it is H then X, rather than H*X. I suggest writing it as H; X.

PabloAndresCQ · 2026-02-09T18:24:21Z

All good. I've double checked all of the Clifford gates are implemented correctly.

PabloAndresCQ · 2026-02-10T13:13:00Z

+        return Phase((self._value - other._value) % 4)
+
+    def __neg__(self) -> "Phase":
+        return self + Phase.REAL_NEGATIVE


This looks very bugprone:

Doesn't do modulo addition

It would cause a self==REAL_NEGATIVE to become IMAGINARY_POSITIVE

AFAIK you are not using negation anywhere... maybe just remove it?

PabloAndresCQ · 2026-02-10T13:20:22Z

+                raise ValueError(f"Unhandled Pauli multiplication: {self}, {other}")
+
+
+class StabilizerGenerator:


Nit-pick: I would just call this a Stabilizer. Any set of stabilizers is a generator set for some stabilizer group.

PabloAndresCQ · 2026-02-10T13:29:06Z

+
+
+class StabilizerList:
+    entries: list[StabilizerGenerator]  # le n: specified_qubits


Nit-pick: since you asked for suggestions on naming... I think the most appropriate thing to do here would be to define this as:
generators: list[Stabilizer]

An entry of the tableau, in my mind, is an element on a chosen (stabilizer, qubit) coordinate, similarly to entries (row, column) in matrices.

And as mentioned in a previous comment, any stabilizer is a generator of some group. What we care is that this particular list is our current choice of generators for the group stabilising the state. This is why I think it's more appropriate to call the variable generators rather than the type.

PabloAndresCQ · 2026-02-10T13:41:55Z

+    entries: list[StabilizerGenerator]  # le n: specified_qubits
+
+    def __init__(self, stabilizer_strings: list[str]):
+        n_qubits_known = len(stabilizer_strings)


Nit-pick: this is a bit puzzling to me.

If you are assuming the input is a pure stabiliser state, then len(stabilizer_strings) must match with all len(stab.paulis) for each stabiliser stab. In this case, you can add that extra check and I'd just call this variable n_qubits .

Otherwise, if you are accepting mixed state tableaus, this is variable name is confusing... You could have stabilizer_strings = ["+ZZ"]. This is a valid mixed state 1/2 (|00><00| + |11><11|) on two qubits.

Ofc, it doesn't really matter since this variable is only used in this scope.

PabloAndresCQ · 2026-02-10T14:38:23Z

+                )
+        return result
+
+    def get_single_state(self) -> np.ndarray:


You've dropped zero_threshold from get_single_state, get_dirac_notation and get_single_dirac_notation. Was this intentional?

PabloAndresCQ · 2026-02-10T14:55:37Z

+        ],
+    ),
+    ReductionExample(
+        # |+++++.....+++> (N_QUBITS_LARGE_EXAMPLE qubits)


Suggested change

# |+++++.....+++> (N_QUBITS_LARGE_EXAMPLE qubits)

# |00...0> + |11...1> (N_QUBITS_LARGE_EXAMPLE qubits)

This comment is wrong. The full state is a GHZ state, not a tensor product of |+> states.

The rest in this test is correct

…ation independently

…or support

…C api for selene's usage. Avoid usage of TableauSimulator and just drive the inner inverse tableau directly. Generate definitions using pytket (gate definitions provided in a script). Fix conformance testing framework as the cnot definition was incorrect (although observed measurements were okay, statevector prints were not)

…y mypy-checking selene-sim and selene-ext instead of the project root. Fixup a mypy complaint in the stim gate definitions file.

…r a submodule

🤖 I have created a release *beep* *boop* --- ## [0.2.5](selene-core-v0.2.4...selene-core-v0.2.5) (2026-02-24) ### Features * Add __version__ attributes ([#137](#137)) ([379ae01](379ae01)) * Interactive use of Selene from python ([#135](#135)) ([db3028d](db3028d)) * Stim improvements (more ops + state printing) ([#115](#115)) ([867d5e5](867d5e5)) ### Bug Fixes * add `___barrier` to Helios QIS for QIR emulation ([#136](#136)) ([dfbc4c3](dfbc4c3)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).

🤖 I have created a release *beep* *boop* --- ## [0.2.11](selene-sim-v0.2.10...selene-sim-v0.2.11) (2026-02-24) ### Features * Add __version__ attributes ([#137](#137)) ([379ae01](379ae01)) * Interactive use of Selene from python ([#135](#135)) ([db3028d](db3028d)) * Quest performance improvements ([#133](#133)) ([af640e9](af640e9)) * Stim improvements (more ops + state printing) ([#115](#115)) ([867d5e5](867d5e5)) * Test on QIS snapshots ([#134](#134)) ([c4d58ab](c4d58ab)) ### Bug Fixes * add `___barrier` to Helios QIS for QIR emulation ([#136](#136)) ([dfbc4c3](dfbc4c3)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).

🤖 I have created a release *beep* *boop* --- ## [0.3.0-alpha.1](selene-core-v0.3.0-alpha.0...selene-core-v0.3.0-alpha.1) (2026-06-12) ### ⚠ BREAKING CHANGES * Use struct exports for all plugin types, separate error model and simulator ([#169](#169)) * Add handling for an additional gateset ([#119](#119)) ### Features * Add __version__ attributes ([#137](#137)) ([379ae01](379ae01)) * Add handling for an additional gateset ([#119](#119)) ([5180b80](5180b80)) * Add simulate_delay functionality ([#139](#139)) ([cca97fa](cca97fa)) * Add support for object files provided as bytes ([#94](#94)) ([c4cfac6](c4cfac6)) * add timing to builtin runtimes and batching options to softrz runtime ([#158](#158)) ([049e123](049e123)) * Build improvements ([#142](#142)) ([12f399b](12f399b)) * Interactive use of Selene from python ([#135](#135)) ([db3028d](db3028d)) * Make interfaces shared, add ArgReader ([#171](#171)) ([143e742](143e742)) * QIR support using QIR-QIS ([#114](#114)) ([70ab294](70ab294)) * Result stream handling refactor ([#93](#93)) ([607a55e](607a55e)) * Stim improvements (more ops + state printing) ([#115](#115)) ([867d5e5](867d5e5)) * test on QIS instead of relying on the upper stack ([#150](#150)) ([b80a9c4](b80a9c4)) * Traces for analytics ([#160](#160)) ([24b9978](24b9978)) * Use mingw instead of msvc for windows wheels ([#143](#143)) ([3d91514](3d91514)) * Use struct exports for all plugin types, separate error model and simulator ([#169](#169)) ([c44a5c9](c44a5c9)) ### Bug Fixes * add `___barrier` to Helios QIS for QIR emulation ([#136](#136)) ([dfbc4c3](dfbc4c3)) * avoid using qir_major_version string for QIR detection ([#123](#123)) ([deaa0dc](deaa0dc)) * classify lowered qir-qis bitcode as helios ([#157](#157)) ([67fca60](67fca60)) * make qir-qis optional and stabilize CI ([#164](#164)) ([e52c0b2](e52c0b2)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).

🤖 I have created a release *beep* *boop* --- ## [0.3.0-alpha.1](selene-sim-v0.3.0-alpha.0...selene-sim-v0.3.0-alpha.1) (2026-06-12) ### ⚠ BREAKING CHANGES * Use struct exports for all plugin types, separate error model and simulator ([#169](#169)) * Add handling for an additional gateset ([#119](#119)) ### Features * Add __version__ attributes ([#137](#137)) ([379ae01](379ae01)) * Add event hook which records measurement results ([#104](#104)) ([01300ee](01300ee)), closes [#103](#103) * Add handling for an additional gateset ([#119](#119)) ([5180b80](5180b80)) * Add simulate_delay functionality ([#139](#139)) ([cca97fa](cca97fa)) * Add support for object files provided as bytes ([#94](#94)) ([c4cfac6](c4cfac6)) * add timing to builtin runtimes and batching options to softrz runtime ([#158](#158)) ([049e123](049e123)) * Better exception handling for parse_shots=False ([#70](#70)) ([3caf530](3caf530)) * Build improvements ([#142](#142)) ([12f399b](12f399b)) * Bump selene-core dependency and relax guppylang testing dependency ([#120](#120)) ([6375791](6375791)) * Cleanup error'd processes before log collection ([#98](#98)) ([77e698e](77e698e)) * **compiler:** Bump tket version; add wasm + gpu to the hugr-qis registry ([c69155d](c69155d)) * correct shot end strategy and error processing ([#91](#91)) ([93eaeb0](93eaeb0)) * Emit a nicer error when trying to emulate unsupported pytket ops ([#72](#72)) ([d88a28a](d88a28a)) * Fine-grained timeout configuration ([#69](#69)) ([072842e](072842e)) * Interactive use of Selene from python ([#135](#135)) ([db3028d](db3028d)) * Make interfaces shared, add ArgReader ([#171](#171)) ([143e742](143e742)) * QIR support using QIR-QIS ([#114](#114)) ([70ab294](70ab294)) * Quest performance improvements ([#133](#133)) ([af640e9](af640e9)) * random_advance ([#55](#55)) ([974b496](974b496)) * Record simulated delays in event hooks ([#154](#154)) ([608884a](608884a)) * Result stream handling refactor ([#93](#93)) ([607a55e](607a55e)) * Stim improvements (more ops + state printing) ([#115](#115)) ([867d5e5](867d5e5)) * Support state-dump passthrough on quantum replay simulator ([#108](#108)) ([1b01a01](1b01a01)) * test on QIS instead of relying on the upper stack ([#150](#150)) ([b80a9c4](b80a9c4)) * Test on QIS snapshots ([#134](#134)) ([c4d58ab](c4d58ab)) * Traces for analytics ([#160](#160)) ([24b9978](24b9978)) * update to tket-qsystem 0.20 ([#66](#66)) ([7191b07](7191b07)) * Use mingw instead of msvc for windows wheels ([#143](#143)) ([3d91514](3d91514)) * Use SeleneStartupError for errors before shot_start ([#170](#170)) ([85c2907](85c2907)) * Use struct exports for all plugin types, separate error model and simulator ([#169](#169)) ([c44a5c9](c44a5c9)) ### Bug Fixes * add `___barrier` to Helios QIS for QIR emulation ([#136](#136)) ([dfbc4c3](dfbc4c3)) * Add ENDING to allowed shot state on receiving meta information ([#100](#100)) ([fc4d673](fc4d673)) * Add numpy dependency back ([#127](#127)) ([c9a01eb](c9a01eb)) * avoid using qir_major_version string for QIR detection ([#123](#123)) ([deaa0dc](deaa0dc)) * classify lowered qir-qis bitcode as helios ([#157](#157)) ([67fca60](67fca60)) * **compiler:** error when entrypoint has arguments ([#84](#84)) ([604b131](604b131)) * **compiler:** update tket-qystem to fix CZ bug ([#78](#78)) ([3991f11](3991f11)) * correct post_runtime duration metric ([#74](#74)) ([0bef66a](0bef66a)) * make qir-qis optional and stabilize CI ([#164](#164)) ([e52c0b2](e52c0b2)) ### Documentation * add repository citation guidance and metadata ([#146](#146)) ([880a58b](880a58b)), closes [#145](#145) * Correct URL to guppy repository ([#121](#121)) ([27ded00](27ded00)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). Co-authored-by: Jake Arkinstall <65358059+jake-arkinstall@users.noreply.github.com>

jake-arkinstall requested a review from PabloAndresCQ January 26, 2026 17:14

jake-arkinstall requested a review from ss2165 as a code owner January 26, 2026 17:14

PabloAndresCQ approved these changes Feb 10, 2026

View reviewed changes

jake-arkinstall added 6 commits February 12, 2026 10:52

Stim: decompose RXY phi to octants and optimise each parameter combin…

7530098

…ation independently

Support dumping of stabilisers from stim

92723e5

WIP: adding python handling of stim states

7731701

Finalize demo of reduction and provide test cases

5f9cdf2

Finish demo of stim state printing, with density matrix and statevect…

a44c6ff

…or support

jake-arkinstall force-pushed the feat/stim-fix--theta-pi--phi-quarter-pi branch from 8f2c769 to dec800f Compare February 12, 2026 13:31

jake-arkinstall added 4 commits February 12, 2026 13:45

Don't mypy-check files in the target directory - avoided by explicitl…

8ad1986

…y mypy-checking selene-sim and selene-ext instead of the project root. Fixup a mypy complaint in the stim gate definitions file.

Document stim gate application more explicitly

0755083

Address review comments in state.py and tests thereof

cb43aaa

Fixup actions, docs, and tooling to acknowledge that stim is no longe…

0d2163f

…r a submodule

jake-arkinstall merged commit 867d5e5 into main Feb 12, 2026
10 checks passed

jake-arkinstall deleted the feat/stim-fix--theta-pi--phi-quarter-pi branch February 12, 2026 14:34

This was referenced Feb 12, 2026

chore: release selene-sim 0.2.11 #131

Merged

chore: release selene-core 0.2.5 #132

Merged

This was referenced Apr 21, 2026

chore: release selene-sim 0.3.0-alpha.1 #162

Merged

chore: release selene-core 0.3.0-alpha.1 #163

Merged

This was referenced Jul 6, 2026

chore: release selene-sim 1.0.0-alpha.1 #188

Draft

chore: release selene-core 1.0.0-alpha.1 #189

Draft

Navigation Menu

feat: Stim improvements (more ops + state printing) - #115

feat: Stim improvements (more ops + state printing)#115
jake-arkinstall merged 10 commits into
mainfrom
feat/stim-fix--theta-pi--phi-quarter-pi

jake-arkinstall commented Jan 26, 2026 •

edited

Loading

Uh oh!

PabloAndresCQ left a comment

Uh oh!

PabloAndresCQ Feb 9, 2026

Uh oh!

PabloAndresCQ Feb 9, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

PabloAndresCQ Feb 10, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		raise ValueError(f"Unhandled Pauli multiplication: {self}, {other}")


		class StabilizerGenerator:



		class StabilizerList:
		entries: list[StabilizerGenerator] # le n: specified_qubits

	# \|+++++.....+++> (N_QUBITS_LARGE_EXAMPLE qubits)
	# \|00...0> + \|11...1> (N_QUBITS_LARGE_EXAMPLE qubits)

Conversation

jake-arkinstall commented Jan 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Background

On more ops

On state printing

Uh oh!

PabloAndresCQ left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

jake-arkinstall commented Jan 26, 2026 •

edited

Loading