Feature Request: Allow qvector unpacking (*) or direct qvector input for custom operations (cudaq.register_operation)

[mmvandieren]

Required prerequisites

• Search the issue tracker to check if your feature has already been mentioned or rejected in other issues.

Describe the feature

When defining custom quantum operations from unitary matrices using cudaq.register_operation, the generated operation strictly requires discrete cudaq.qubit arguments.

When working with a cudaq.qvector, there is currently no way to dynamically unpack the register into the custom operation call. Python's standard argument unpacking operator (*) is not supported by @cudaq.kernel definitions, and passing the qvector directly throws an error.

import numpy as np

# 1. Define a unitary matrix for a custom operation (e.g., a 2-qubit operator)
custom_matrix = np.eye(4) # Using Identity for simplicity in this MWE
cudaq.register_operation("my_custom_op", custom_matrix)

@cudaq.kernel
def main_kernel():
    # Allocate a vector of 2 qubits
    qvec = cudaq.qvector(2)
    
    # ---------------------------------------------------------
    # DESIRED BEHAVIOR 1: Unpacking (Currently fails to parse/compile)
    # ---------------------------------------------------------
    # my_custom_op(*qvec) 
    
    # ---------------------------------------------------------
    # DESIRED BEHAVIOR 2: Passing the qvector directly 
    # ---------------------------------------------------------
    # my_custom_op(qvec)
    
    # ---------------------------------------------------------
    # CURRENT REQUIRED WORKAROUND: Hardcoding indices
    # ---------------------------------------------------------
    my_custom_op(qvec[0], qvec[1])

# Test the compilation
cudaq.sample(main_kernel)```

[khalatepradnya]

• Related to issue Support cudaq.qvector for the multi-qubit custom op #2484

[sacpis]

Will look into it this week.