Mid-circuit measurement feedforward support by shah-rushil · Pull Request #63 · amazon-braket/BraketSimulator.jl

shah-rushil · 2025-06-26T21:08:27Z

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.

shah-rushil · 2025-06-26T22:45:36Z

The relevant files for the buffer method are branched_evolve_operators.jl, branched_simulator_operators.jl, and somewhat of gate_kernels/sv_simulator/dm_simulator but those ones have a less new stuff and mostly just have some helper functions. Most of the circuit evolution with the branching struct is in the first two files.

…te of functionality is.

shah-rushil · 2025-06-27T22:31:33Z

I realized I made a major oversight in the classical assignment operation. I assumed that the right hand side is equal for all paths, when in reality it is not. So, I am trying to update that and will commit as soon as I get it fixed.

…case bugs

rmshaffer · 2025-07-01T21:34:44Z

@shah-rushil for the CI failures, you can merge the latest from https://github.com/amazon-braket/BraketSimulator.jl/tree/mcm-experimental into your fork - it should update the CI so it's only testing against supported Julia versions.

shah-rushil · 2025-07-07T17:43:53Z

There are still issues regarding variable scoping and return statements that I am trying to fix. For return statements, it works if the return statement is in the outer body of the function call but if it is in an if statement, it won't exit out of the function. For scoping, there are some nuances regarding which variables to store and remove. I will implement this by keeping a frame counter that indicates which frame level the variables were created in.

…operations along with tests

… gate modifiers

… increase coverage

speller26

More to come, just starting to review

speller26 · 2025-07-17T04:54:07Z

 """
-    FreeParameter
-    FreeParameter(name::Symbol) -> FreeParameter
+	FreeParameter


nit: We actually use spaces rather than tabs as convention; it keeps our line lengths consistent (the IDE can take care of auto-inserting the right number of spaces). This should also significantly reduce the number of lines in the PR

Just a bump on this; this way, the reader can focus on the parts that changed in the file.

speller26 · 2025-07-17T07:55:58Z

+	println(expr)
+	println(sim.states)
+	println(sim.variables)
+	println(sim.qubit_mapping)
+	println(sim.measurements)


We can remove all the print statements

speller26 · 2025-07-29T07:56:47Z

+	# Process the expression to get qubit indices
+	expr_type = head(qubit_expr)
+
+	if expr_type == :identifier


Rather than going down a long chain of if branches, is it possible to dispatch using a dictionary? Same goes for inner if statements that check types.

speller26 · 2025-07-29T07:59:38Z

 """
-    FreeParameter
-    FreeParameter(name::Symbol) -> FreeParameter
+	FreeParameter


Just a bump on this; this way, the reader can focus on the parts that changed in the file.

speller26 · 2025-07-29T08:04:53Z

@@ -0,0 +1,51 @@
+# All of the functions that take in a Program type (JAQCD) since it is no longer used


Just making sure, is this just something you're moving out of BraketSimulator.jl or something like that? These paths are still used by the old non-branched simulator.

speller26 · 2025-07-29T08:05:52Z

+end
+
+"""
+	BranchedSimulatorOperators


If this struct is the simulator itself, then it should be named so (rather than ...Operators)

speller26 · 2025-07-29T08:16:35Z

+		qubit_name = Quasar.name(qubit_expr)
+
+		# Get the index
+		qubit_ix = _evolve_branched_ast_operators(sim, qubit_expr.args[2])


Instead of working with general QasmExpressions, is there a better type we can work with? So instead of calling .args[2], which is hard to understand, the type will have a property that tells us what it is. This would also make the code far more robust against bugs that could be caused by indexing errors (or even worse, potential changes to QasmExpression structure).

speller26 · 2025-08-04T04:22:44Z

+        for i in 0:(length(state)÷2-1)
+            lower_ix = pad_bit(i, endian_target) + 1
+            higher_ix = flip_bit(lower_ix - 1, endian_target) + 1
+
+            # Swap amplitudes
+            temp = state[lower_ix]
+            state[lower_ix] = state[higher_ix]
+            state[higher_ix] = temp


Is it possible to reuse the X-gate logic here? Alternatively, change the existing X-gate logic to do this.

speller26 · 2025-08-04T04:23:41Z

+Reset a qubit to |0⟩ state in a state vector simulator.
+This projects the state to the |0⟩ state of the target qubit and then normalizes.
+"""
+function _apply_reset(simulator::StateVectorSimulator, target::Int)


Do we need both this and the preceding function?

speller26 · 2025-08-04T04:24:08Z

+Apply measurement projection to a state vector simulator.
+This collapses the state to the subspace corresponding to the measurement outcome.
+"""
+function apply_projection(simulator::StateVectorSimulator, qubit::Int, outcome::Int)


speller26 · 2025-08-04T04:24:47Z

+Reset a qubit to |0⟩ state in a density matrix simulator.
+This projects the state to the |0⟩ state of the target qubit.
+"""
+function _apply_reset(simulator::DensityMatrixSimulator, target::Int)


speller26 · 2025-08-04T04:26:13Z

+    return
+end
+
+function apply_gate!(::Reset, state_vec::AbstractStateVector{T}, qubit::Int) where {T<:Complex}


There's already an implementation for reset implemented in the simulator files; why have another here?

speller26 · 2025-08-04T04:28:15Z

+As of now, this only works with registers and classical bits. Functionality for any classical variable
+still needs to be implemented.
+"""
+function evaluate_qubits(sim::BranchedSimulator, qubit_expr::QasmExpression)


This is a really long function; it would be easier to reason about if it were split into multiple helpers

speller26 · 2025-08-04T04:36:42Z

+
+Evaluates gate modifiers. Always returns a dictionary mapping path indices to tuples of (modifier_expr, inner_expr).
+"""
+function evaluate_modifiers(sim::BranchedSimulator, expr::QasmExpression)


This can be private too, since it's called by a private function

speller26 · 2025-08-04T04:38:09Z

+end
+
+"""
+    _apply_modifiers(gate_op::QuantumOperator, modifiers::Vector, target_indices::Vector{Int}) -> QuantumOperator


Would be good to distinguish the purpose of this from evaluate_modifiers; why do why have both, since the names suggest that they do the same thing?

speller26 · 2025-08-04T04:44:01Z

+
+Handle a classical variable assignment in the branched simulation model.
+"""
+function _handle_classical_assignment(sim::BranchedSimulator, expr::QasmExpression)


Another very complex function that could be split up (much easier to reason about sub-functions than multiple nested conditions)

speller26 · 2025-08-04T04:47:14Z

+    current_paths = copy(sim.active_paths)
+
+    # This is for assigning to a bit register or any classical variable
+    if head(lhs) == :identifier


Another good place to use a dictionary rather than conditionals, since we're dispatching on types

… functions

speller26

Looks good! Just a few more minor comments.

speller26 · 2025-08-12T22:16:29Z

+        target_indices = target_indices_dict[path_idx]
+
+        # User-defined gate - create a new scope and execute the body
+        if haskey(sim.gate_defs, gate_name)


The two branches of this conditional definitely deserve dedicated functions; user-defined gates for sure deserve a handler considering their complexity, and a docstring explaining why it's so complex.

speller26 · 2025-08-12T22:40:22Z

+
+        # Convert the parameter results to a dictionary mapping path indices to parameter vectors
+        for path_idx in sim.active_paths
+            if param_results isa Dict && haskey(param_results, path_idx)


Rather than doing the isa Dict check each time, you can probably short-circuit that before entering the loop, since it'll evaluate to the same thing each time.

speller26 · 2025-08-12T22:44:03Z

+
+        for loop_value in loop_variable_vals
+            # Set the loop variable for each active path
+            for path_idx in sim.active_paths


A bit confusing to see path_idx as variable here shadowing the outer path_idx variable. In fact, does this even need to be a loop, since

sim.active_paths = [path_idx]

above?

speller26 · 2025-08-12T22:46:49Z

+function _handle_casting(sim::BranchedSimulator, expr::QasmExpression)
+    casting_to = expr.args[1].args[1]
+    value = _evolve_branched_ast(sim, expr.args[2])


Would probably be good to comment the type of node that's expected here, because it's hard to tell what value actually is

speller26 · 2025-08-12T22:47:35Z

+function _handle_indexed_identifier(sim::BranchedSimulator, expr::QasmExpression)
+
+    identifier_name = Quasar.name(expr)
+    indices = _evolve_branched_ast(sim, expr.args[2])


Would probably be good to comment the type of node that's expected here, because it's hard to tell what indices actually is

speller26 · 2025-08-12T22:59:23Z

+            if isa(gate_op, GPhase)
+                # Create a controlled version of the gate
+                # For phase gates, the control is the only qubit
+                bitvals = tuple(1)
+
+                # Create the controlled gate
+                gate_op = Control(gate_op, bitvals)
+            else
+                # Create a controlled version of the gate
+                # The first qubit is the control, the rest are targets for the gate
+                bitvals = tuple(1)
+
+                # Create the controlled gate
+                gate_op = Control(gate_op, bitvals)


Looks like this can just be collapsed, since both branches are the same (unless Julia typing doesn't allow it)

speller26 · 2025-08-12T22:59:37Z

+            if isa(gate_op, GPhase)
+
+                # Create a controlled version of the gate with control on |0⟩ state
+                bitvals = tuple(0)
+
+                # Create the negatively controlled gate
+                gate_op = Control(gate_op, bitvals)
+            else
+                # Create a controlled version of the gate with control on |0⟩ state
+                bitvals = tuple(0)
+
+                # Create the negatively controlled gate
+                gate_op = Control(gate_op, bitvals)
+            end


speller26 · 2025-08-12T23:11:49Z

+        if is_measurement
+            # Process measurement outcomes for this path
+            if haskey(rhs_value.path_outcomes, current_path_idx)


For easier readability (fewer layers), combine these conditions, and make the else statement an elseif !is_measurement.

speller26 · 2025-08-12T23:12:48Z

+        if is_measurement
+            # Process measurement for this path
+            if haskey(rhs_value.path_outcomes, current_path_idx)


Same here with the conditional combining

speller26

Some readability suggestions; take or leave as you see fit

speller26 · 2025-08-14T22:48:23Z

+                        if haskey(sim.qubit_mapping, indexed_name)
+                            qubit_idx = sim.qubit_mapping[indexed_name]
+                            # Store the qubit index directly
+                            sim.variables[path_idx][param_name] = FramedVariable(param_name, :qubit_declaration, qubit_idx, false, sim.curr_frame+1)
+                        else
+                            error("Qubit $indexed_name not found in qubit mapping")
+                        end


Suggested change

if haskey(sim.qubit_mapping, indexed_name)

qubit_idx = sim.qubit_mapping[indexed_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = FramedVariable(param_name, :qubit_declaration, qubit_idx, false, sim.curr_frame+1)

else

error("Qubit $indexed_name not found in qubit mapping")

end

haskey(sim.qubit_mapping, indexed_name) || error("Qubit $indexed_name not found in qubit mapping")

qubit_idx = sim.qubit_mapping[indexed_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = FramedVariable(param_name, :qubit_declaration, qubit_idx, false, sim.curr_frame+1)

speller26 · 2025-08-14T22:48:58Z

+                        if haskey(sim.qubit_mapping, qubit_name)
+                            qubit_idx = sim.qubit_mapping[qubit_name]
+                            # Store the qubit index directly
+                            sim.variables[path_idx][param_name] = FramedVariable(param_name, :qubit_declaration, qubit_idx, false, sim.curr_frame+1)
+                        else
+                            error("Qubit $qubit_name not found in qubit mapping")
+                        end


Suggested change

if haskey(sim.qubit_mapping, qubit_name)

qubit_idx = sim.qubit_mapping[qubit_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = FramedVariable(param_name, :qubit_declaration, qubit_idx, false, sim.curr_frame+1)

else

error("Qubit $qubit_name not found in qubit mapping")

end

haskey(sim.qubit_mapping, qubit_name) || error("Qubit $qubit_name not found in qubit mapping")

qubit_idx = sim.qubit_mapping[qubit_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = FramedVariable(param_name, :qubit_declaration, qubit_idx, false, sim.curr_frame+1)

speller26 · 2025-08-14T22:52:17Z

+    if shots <= 0
+        error("The number of inputted shots must be a positive integer")
+    end
+    if shots != simulator.shots[1]
+        error("The number of shots in the simulator must be equal to the number of shots passed in")
+    end


Suggested change

if shots <= 0

error("The number of inputted shots must be a positive integer")

end

if shots != simulator.shots[1]

error("The number of shots in the simulator must be equal to the number of shots passed in")

end

shots <= 0 || error("The number of inputted shots must be a positive integer")

shots != simulator.shots[1] || error("The number of shots in the simulator must be equal to the number of shots passed in")

speller26 · 2025-08-14T22:53:04Z

+            if !isempty(register_qubits)
+                for qubit_idx in register_qubits
+                    push!(all_qubits, qubit_idx)
+                end
+            else
+                error("Missing qubit '$qubit_name'")
+            end


Suggested change

if !isempty(register_qubits)

for qubit_idx in register_qubits

push!(all_qubits, qubit_idx)

end

else

error("Missing qubit '$qubit_name'")

end

!isempty(register_qubits) || error("Missing qubit '$qubit_name'")

for qubit_idx in register_qubits

push!(all_qubits, qubit_idx)

end

speller26 · 2025-08-14T22:54:59Z

+                    if ix >= 0 && ix < length(register_qubits)
+                        push!(all_qubits, register_qubits[ix+1])  # Convert to 1-based indexing
+                    else
+                        error("Index $ix out of bounds for register $qubit_name")
+                    end


Suggested change

if ix >= 0 && ix < length(register_qubits)

push!(all_qubits, register_qubits[ix+1]) # Convert to 1-based indexing

else

error("Index $ix out of bounds for register $qubit_name")

end

(ix >= 0 && ix < length(register_qubits)) || error("Index $ix out of bounds for register $qubit_name")

push!(all_qubits, register_qubits[ix+1]) # Convert to 1-based indexing

speller26 · 2025-08-14T22:55:22Z

+                if path_qubit_ix >= 0 && path_qubit_ix < length(register_qubits)
+                    push!(all_qubits, register_qubits[path_qubit_ix+1])  # Convert to 1-based indexing
+                else
+                    error("Index $path_qubit_ix out of bounds for register $qubit_name")
+                end


Suggested change

if path_qubit_ix >= 0 && path_qubit_ix < length(register_qubits)

push!(all_qubits, register_qubits[path_qubit_ix+1]) # Convert to 1-based indexing

else

error("Index $path_qubit_ix out of bounds for register $qubit_name")

end

path_qubit_ix >= 0 && path_qubit_ix < length(register_qubits) || error("Index $path_qubit_ix out of bounds for register $qubit_name")

push!(all_qubits, register_qubits[path_qubit_ix+1]) # Convert to 1-based indexing

speller26 · 2025-08-14T22:56:22Z

+                if julia_index <= length(var.val)
+                    # Access the bit directly from the boolean array
+                    results[path_idx] = var.val[julia_index]
+                else
+                    error("Index out of bounds error, index too large for bit vector")
+                end


Suggested change

if julia_index <= length(var.val)

# Access the bit directly from the boolean array

results[path_idx] = var.val[julia_index]

else

error("Index out of bounds error, index too large for bit vector")

end

julia_index <= length(var.val) || error("Index out of bounds error, index too large for bit vector")

# Access the bit directly from the boolean array

results[path_idx] = var.val[julia_index]

speller26 · 2025-08-14T22:57:24Z

+        if haskey(sim.qubit_mapping, indexed_name)
+            qubit_idx = sim.qubit_mapping[indexed_name]
+            # Store the qubit index directly
+            sim.variables[path_idx][param_name] = ClassicalVariable(param_name, :qubit_declaration, qubit_idx, true)
+        else
+            error("Qubit $indexed_name not found in qubit mapping")
+        end


Suggested change

if haskey(sim.qubit_mapping, indexed_name)

qubit_idx = sim.qubit_mapping[indexed_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = ClassicalVariable(param_name, :qubit_declaration, qubit_idx, true)

else

error("Qubit $indexed_name not found in qubit mapping")

end

haskey(sim.qubit_mapping, indexed_name) || error("Qubit $indexed_name not found in qubit mapping")

qubit_idx = sim.qubit_mapping[indexed_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = ClassicalVariable(param_name, :qubit_declaration, qubit_idx, true)

speller26 · 2025-08-14T22:57:46Z

+        if haskey(sim.qubit_mapping, qubit_name)
+            qubit_idx = sim.qubit_mapping[qubit_name]
+            # Store the qubit index directly
+            sim.variables[path_idx][param_name] = ClassicalVariable(param_name, :qubit_declaration, qubit_idx, true)
+        else
+            error("Qubit $qubit_name not found in qubit mapping")
+        end


Suggested change

if haskey(sim.qubit_mapping, qubit_name)

qubit_idx = sim.qubit_mapping[qubit_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = ClassicalVariable(param_name, :qubit_declaration, qubit_idx, true)

else

error("Qubit $qubit_name not found in qubit mapping")

end

haskey(sim.qubit_mapping, qubit_name) || error("Qubit $qubit_name not found in qubit mapping")

qubit_idx = sim.qubit_mapping[qubit_name]

# Store the qubit index directly

sim.variables[path_idx][param_name] = ClassicalVariable(param_name, :qubit_declaration, qubit_idx, true)

speller26 · 2025-08-14T22:58:24Z

+    if expr_type in keys(evolution_dispatch_nodes)
+        return evolution_dispatch_nodes[expr_type](sim, expr)
+    else
+        error("Cannot process expression of type $expr_type.")
+    end


Suggested change

if expr_type in keys(evolution_dispatch_nodes)

return evolution_dispatch_nodes[expr_type](sim, expr)

else

error("Cannot process expression of type $expr_type.")

end

if expr_type in keys(evolution_dispatch_nodes)

return evolution_dispatch_nodes[expr_type](sim, expr)

error("Cannot process expression of type $expr_type.")

Initial commit

6074a8f

shah-rushil requested a review from a team as a code owner June 26, 2025 21:08

Cleaned up branched_simulator_operators.jl

4b68796

Tidied up evolution functions to make more clear what the current sta…

647d725

…te of functionality is.

rmshaffer reviewed Jun 27, 2025

View reviewed changes

Comment thread src/branched_simulator_operators.jl Outdated

Comment thread src/branched_simulator_operators.jl Outdated

Reorganized tests to illustrate which components are being tested

9ad53c4

shah-rushil added 4 commits June 27, 2025 15:55

Fixed issue with updating RHS for each active path

6242d8b

fixed bug with assigning classical arrays

6804091

Added simulation pruning based on number of shots passed down.

b215916

Implemented for loops, while loops, break statements, and fixed test …

dfd250e

…case bugs

rmshaffer mentioned this pull request Jun 30, 2025

infra: limit CI to latest and LTS releases #64

Merged

5 tasks

shah-rushil added 2 commits June 30, 2025 14:26

implemented parametric gates

a797562

added casting between ints and floats

3bf3247

shah-rushil and others added 4 commits July 1, 2025 15:00

Merge branch 'amazon-braket:main' into main

b8a3806

add tests for variable scoping and aliasing

7b90df4

Merge branch 'main' of https://github.com/shah-rushil/BraketSimulator.jl

344fa24

Partial functionality for function calls and aliasing

f8e575f

shah-rushil added 9 commits July 7, 2025 16:27

implemented scoping and aliasing fix

2bcffd9

added tests and implementation for gate modifiers

30e4e88

implemented arbitrary user defined function calls

6c1a41f

fixed user defined parametric gate implementation

df795b2

fixed bugs with input and gate operations and added builtin function …

531e90b

…operations along with tests

implemented modified user defined gates

f908c58

added gphase functionality and fixed commutivity error with power/inv…

15c70e5

… gate modifiers

fixed error with custom gate calls

7b9a785

added coverage information along with some fixes to tests and code to…

9ecd110

… increase coverage

shah-rushil added 5 commits July 11, 2025 16:30

implemented break and continue statements for control sequences

bc27de3

added functionality with python sdk along with reset support

cdacada

fixed small bug with alias indexing and updated braket extension code

69be703

Implemented and added tests for the reset operator

0ce9bb1

Fixed error with initializing classical bit registers

b3746e2

speller26 reviewed Jul 17, 2025

View reviewed changes

cleaned up code and addressed binary_op comment

f87a08b

speller26 reviewed Jul 29, 2025

View reviewed changes

shah-rushil added 8 commits July 29, 2025 09:59

test of replacing tabs with spaces

2a6d415

fixed more spacing issues

7c557ea

a few more spacing issues fixed

13ec2aa

cleaned up repository, naming, and spacing

67828ff

modification to include compatibility with python interface

235d7ba

converted node evolution to dictionary for constant lookup

399a8e1

updated code to increase coverage

ec768d4

improved code coverage and change tabs to spaces

7b1964f

speller26 reviewed Aug 4, 2025

View reviewed changes

shah-rushil added 4 commits August 4, 2025 11:02

removed extra functions and split down functions into multiple helper…

ecc23d2

… functions

fixed small type error in conditional checking

67a51ff

replaced default variable declataion

83d17e0

optimized reset operation

95a3845

speller26 reviewed Aug 12, 2025

View reviewed changes

improved readability of evolution helper functions

4c43580

speller26 reviewed Aug 14, 2025

View reviewed changes

small updates to style and removing failed tests

3355ad3

speller26 approved these changes Aug 14, 2025

View reviewed changes

fixed some small errors with CI

301080f

speller26 merged commit 3c80892 into amazon-braket:mcm-experimental Aug 18, 2025
0 of 7 checks passed

		@@ -0,0 +1,51 @@
		# All of the functions that take in a Program type (JAQCD) since it is no longer used

Uh oh!

Conversation

shah-rushil commented Jun 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merge Checklist

General

Tests

Uh oh!

shah-rushil commented Jun 26, 2025

Uh oh!

Uh oh!

Uh oh!

shah-rushil commented Jun 27, 2025

Uh oh!

rmshaffer commented Jul 1, 2025

Uh oh!

shah-rushil commented Jul 7, 2025

Uh oh!

speller26 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!

Uh oh!

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!

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!

speller26 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

shah-rushil commented Jun 26, 2025 •

edited

Loading