feat: Classical control flow with mid circuit measurements by speller26 · Pull Request #347 · amazon-braket/amazon-braket-default-simulator-python

speller26 · 2026-02-19T00:49:00Z

Issue #, if available:

Description of changes:

Testing done:

Merge Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your pull request.

General

I have read the CONTRIBUTING doc
I used the commit message format described in CONTRIBUTING
I have updated any necessary documentation, including READMEs and API docs (if appropriate)

Tests

I have added tests that prove my fix is effective or that my feature works (if appropriate)
I have checked that my tests are not configured for a specific region or account (if appropriate)

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

More to come

* Squeeze single-element density matrix before computing expectation * Pinned setuptools to fix doc build * Fixed all linter errors

codecov · 2026-02-23T18:56:54Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (e6bc3bc) to head (f289b82).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##              main      #347    +/-   ##
==========================================
  Coverage   100.00%   100.00%            
==========================================
  Files           48        49     +1     
  Lines         4256      4754   +498     
  Branches       446       533    +87     
==========================================
+ Hits          4256      4754   +498

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

yitchen-tim · 2026-03-02T14:06:17Z

Some product-level UX questions:

Several new feature for MCM triggers _uses_advanced_language_features=True. This causes warning "This program uses OpenQASM language features that may not be supported on QPUs or on-demand simulators." This is somewhat confusing to users, as if there is something wrong in their program/execution and they cannot trust the simulation results. Given that we are pushing for official support of MCM. Should we remove these warnings for MCM, and reserve _uses_advanced_language_features=True for something truly experimental?
We are lacking result schema and result class that supports MCM and general variable results. Right now, variables play no role in the result object. Result object always returns the end-of-circuit measurement outcome for all used qubits regardless whether there is an measurement instruction. Is it intentional and expected? Have you thought about the UX around the fact that we are not outputting any variables?

laurencap

Reviewed 1.5 files today. Should have more time tomorrow!

laurencap · 2026-03-02T22:40:47Z

-        condition = cast_to(BooleanLiteral, self.visit(node.condition))
-        for statement in node.if_block if condition.value else node.else_block:
-            self.visit(statement)
+        self.context.handle_branching_statement(node, self.visit)


Curious about this. Doesn't the visitor usually extract all the important fields, and then call context methods on extracted fields? Why not call visit from this visitor?

This is really annoying, and I'm totally open to better ways to do this. The problem is, the node itself can contain MCM conditionals, so calling visit right away will trigger the branch hander; the recursive nature of these calls means evaluation is completed "bottom-up," while the branch simulations have to be done "top-down."

laurencap · 2026-03-02T22:47:51Z

+            measurement_target (Identifier | IndexedIdentifier | None): The AST node
+                for the classical variable being assigned, e.g. ``b`` in
+                ``b = measure q[0]``. Used by the branched MCM path to update
+                per-path classical variables. None for end-of-circuit measurements.


b[0] = measure q[0] is a valid circuit-final measurement right? In those cases, this is None. Is there no reason to track b[0] for circuit-final measurements?

I guess, because we return results by qubit index order at the end, it's not needed for anything?

Renamed to classical_destination and clarified documentation; None guarantees that it's terminal because the result can't be used (it isn't assigned to anything); however, assignment to a target may be either terminal or feedforward, but we don't know until the result is actually used.

laurencap · 2026-03-02T22:48:42Z

    def add_verbatim_marker(self, marker) -> None:
        """Add verbatim markers"""

+    def handle_branching_statement(self, node: BranchingStatement, visit_block: Callable) -> None:


Cross commenting from the last file -- visit_block callback feels odd to me. Not that important though. Nice documentation here

laurencap · 2026-03-02T22:53:10Z

+                with self.enter_scope():
+                    self.declare_variable(node.identifier.name, node.type, i)
+                    visit_block(deepcopy(node.block))
+            except _BreakSignal:


I see, maybe these were moved for ease of integration with _Continue/BreakSignal? Is that basically what changed from interpreter.py, support for these signals?

Yeah, break and continue actually weren't supported at all before, even for static circuits

laurencap · 2026-03-02T22:57:50Z

+        but no control flow depended on them) are registered in the circuit
+        as normal end-of-circuit measurements.
+        """
+        if not self._is_branched and self._pending_mcm_targets:


is_branched and circuit properties won't be called until interpretation is complete?

_is_branched ia called frequently, but _pending_measurements will always be empty once the simulation is in branched mode (although that can be optimized).

MCM tests now all start from an OpenQASM program.

yitchen-tim · 2026-03-09T14:56:18Z

Opened #351 to fix and allow re-measuring qubits.

aniksd-braket · 2026-03-31T00:20:04Z

            )
-        self.context.add_measure(qubits, targets)
+        if node.target and self.context.supports_midcircuit_measurement:
+            self.context.add_measure(qubits, targets, classical_destination=node.target)


at this point in the visitor, there is no way to know whether the measurement is being used as a MCM right?

correct, supports_midcircuit_measurement is to activate the code path that the measurement is potentially a mcm, but does not commit to a branch simulation path.

But I do agree it's a bit confusing that the logic relys on the presence of classical_destination.

Code path that relies on the presence of classical_destination seems fragile and hard to maintain. Is there a more explicit way?

aniksd-braket · 2026-03-31T00:27:57Z

+            # so that expressions like ``y = x`` read from the correct path.
+            saved_active = list(self.context._active_path_indices)
+            for path_idx in saved_active:
+                self.context._active_path_indices = [path_idx]


It is unclear why the _active_path_indices variable is being overwritten here

I think this is to handle the edge case that I added: test_get_value_reads_correct_path_in_shared_if_body . The program context always get value from the first active branch which would cause silent error. So here it tries to populate correct identifiers' value for active branches.

It does feels hacky this way.

+1 to Aniket comment. We should avoid overwriting. Changing the logic that the program context always only get value from the first active branch may be the key to avoid that

aniksd-braket · 2026-03-31T00:36:40Z

+    def update_value(self, variable: Identifier | IndexedIdentifier, value: Any) -> None:
+        """Update variable value, operating per-path when branched.
+
+        When branched, updates the variable on all active paths. Indexed


When would a variable be shared across paths need to be updated? Wouldn't it be local to the path?

All active paths, namely the paths that are using this variable value.

aniksd-braket · 2026-03-31T00:40:41Z

+            dest_name = mcm_dest.name if isinstance(mcm_dest, Identifier) else mcm_dest.name.name
+            if dest_name == name:
+                if not self._is_branched and self._shots > 0:
+                    self._is_branched = True


why make this branched. The variable may never be used for branching right?

After fixing the bug Tim pointed out, this might actually be a misnomer; flushing used to be done only on encountering control flow. This can be optimized, but I consider it a low priority because I don't think we're gonna see much use of MCMs outside of control flow.

aniksd-braket · 2026-03-31T00:47:14Z

+
+        for path_idx in saved_active:
+            self._active_path_indices = [path_idx]
+            if cast_to(BooleanLiteral, self._visitor(node.condition)).value:


Here is the MCM used in the condition been flushed? Otherwise this casting will throw an exception right?

aniksd-braket · 2026-03-31T00:50:02Z

+        false_paths = []
+
+        for path_idx in saved_active:
+            self._active_path_indices = [path_idx]


I don't why understand _active_path_indices is being overwritten here, shouldn't it be extended?

We're evaluating each path separately, so at this point we start evaluating everything that can happen starting from this point; any new branches only happen within this path. We can probably do some work to merge paths, but that gets real hairy and goes beyond the scope of the PR.

aniksd-braket · 2026-03-31T00:57:04Z

+                if isinstance(index, RangeDefinition)
+                else index.values
+            )
+            for i in index_values:


Isn't this logic repeated from the interpreter?

Yeah, I can try extracting this

AbeCoull · 2026-04-10T16:06:52Z

-            )
-
        # Apply projection matrix
        projected_state = state.copy()


Can't this be done inplace?

AbeCoull · 2026-04-10T16:07:30Z

-            if qubit_value != self.result:
-                projected_state[i] = 0
+            # Zero out amplitudes that don't match the measurement result
+            for i in range(len(projected_state)):


consider njitting this as a follow up

…mazon-braket-default-simulator-python into mcm-experimental

yitchen-tim · 2026-04-13T20:57:49Z

Keeping the branch alive to continue to iterate on it

shah-rushil and others added 15 commits September 12, 2025 13:56

feat: Mid Circuit Measurement (#293)

259b318

Merge branch 'main' into mcm-experimental

253a450

Merge branch 'main' into mcm-experimental

5d38025

Merge branch 'main' into mcm-experimental

9729a19

Merge branch 'main' into mcm-experimental

90cc443

minor fixes

79ccfe9

formatting

0df15af

Fixed bug and incorrect test case

eb03649

Update .gitignore

b4c2ac8

Delete .vscode/launch.json

4bf80a1

More tests

6f1ec22

Simplified visitor

18e1dc2

More to come

Update branched_interpreter.py

e3c8982

Consolidate branching logic into existing classes

115612f

Fix measurements

4456be9

yitchen-tim self-requested a review February 19, 2026 22:02

fix: Fix build failures

0ae913b

* Squeeze single-element density matrix before computing expectation * Pinned setuptools to fix doc build * Fixed all linter errors

speller26 changed the base branch from main to fix February 21, 2026 00:33

Merge branch 'fix' into mcm-experimental

61f5a46

Base automatically changed from fix to main February 23, 2026 17:29

speller26 added 2 commits February 23, 2026 09:32

Merge branch 'main' into mcm-experimental

4a859cb

Merge branch 'main' into mcm-experimental

96b9b2e

add unit test for an edge case

cabffb7

yitchen-tim reviewed Feb 26, 2026

View reviewed changes

Comment thread src/braket/default_simulator/openqasm/program_context.py Outdated

yitchen-tim reviewed Mar 2, 2026

View reviewed changes

add various edge case tests for reset and MCM

64ab4e9

laurencap reviewed Mar 2, 2026

View reviewed changes

better classical assignment handling

6b6d207

rmshaffer reviewed Mar 5, 2026

View reviewed changes

speller26 added 3 commits March 5, 2026 17:10

Even more simplifications

a68744d

MCM tests now all start from an OpenQASM program.

rename

08f7329

Update gate_operations.py

8f99f71

yitchen-tim reviewed Mar 9, 2026

View reviewed changes

Comment thread src/braket/default_simulator/openqasm/interpreter.py

yitchen-tim closed this Mar 9, 2026

yitchen-tim deleted the mcm-experimental branch March 9, 2026 14:50

yitchen-tim restored the mcm-experimental branch March 9, 2026 14:51

yitchen-tim reopened this Mar 9, 2026

fix: allow re-measurement of qubits when MCM is supported (#351)

77fe95b

rmshaffer mentioned this pull request Mar 12, 2026

fix: call ProgramContext constructor amazon-braket/autoqasm#93

Merged

5 tasks

speller26 added 3 commits March 25, 2026 16:07

Allow using measurements outside branching

2bf6c35

Fix qubit reuse corner case

26babd4

Update program_context.py

7ae10d1

aniksd-braket reviewed Mar 31, 2026

View reviewed changes

speller26 added 3 commits April 1, 2026 17:21

Merge branch 'main' into mcm-experimental

5a7810e

Merge branch 'main' into mcm-experimental

c2a2960

Merge branch 'main' into mcm-experimental

d8e88e4

AbeCoull reviewed Apr 10, 2026

View reviewed changes

speller26 and others added 5 commits April 10, 2026 16:00

Update test_gate_operations.py

56dab41

Merge branch 'mcm-experimental' of https://github.com/amazon-braket/a…

bfd2175

…mazon-braket-default-simulator-python into mcm-experimental

test fixes

fe12d6e

Simplify measurements

189855f

Merge branch 'main' into mcm-experimental

f289b82

rmshaffer approved these changes Apr 13, 2026

View reviewed changes

yitchen-tim merged commit d93266f into main Apr 13, 2026
20 checks passed

yitchen-tim deleted the mcm-experimental branch April 13, 2026 20:57

yitchen-tim restored the mcm-experimental branch April 13, 2026 20:57

Conversation

speller26 commented Feb 19, 2026

Merge Checklist

General

Tests

Uh oh!

codecov Bot commented Feb 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

yitchen-tim commented Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

laurencap left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

yitchen-tim commented Mar 9, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yitchen-tim Apr 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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!

yitchen-tim Apr 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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!

codecov Bot commented Feb 23, 2026 •

edited

Loading

yitchen-tim commented Mar 2, 2026 •

edited

Loading

yitchen-tim Apr 3, 2026 •

edited

Loading

yitchen-tim Apr 3, 2026 •

edited

Loading