Controlled exp_pauli is not working

[psm-optimizes]

Required prerequisites

• Consult the security policy. If reporting a security vulnerability, do not report the bug using this form. Use the process described in the policy to report the issue.
• Make sure you've read the documentation. Your issue may be addressed there.
• Search the issue tracker to verify that this hasn't already been reported. +1 or comment there if it has.
• If possible, make a PR with a failing test to give us a starting point to work on!

Describe the bug

1. I want to use the controlled version of exp_pauli to apply the Hamiltonian simulation of the coefficient matrix of a linear system of equations as a part of a HHL-type algorithm.
2. Throughout the circuit, I call the controlled exp_pauli in a for loop to apply the Hamiltonian controlled by different cubits.
3. The error I get is as follows:
   error: 'quake.extract_ref' op invalid constant index value
   RuntimeError Traceback (most recent call last)
   Cell In[17], line 3
   1 # Draw the quantum circuit
   2 print("============================== Circuit structure ============================== \n")
   ----> 3 print(cudaq.draw(GQLSS, b, eigs, r_choice, 2*np.pi, coefficients, words))

RuntimeError: cudaq::builder failed to JIT compile the Quake representation.

Steps to reproduce the bug

The code is available here.

Expected behavior

The circuit should be drawn.

Is this a regression? If it is, put the last known working version (or commit) here.

Not a regression

Environment

• CUDA-Q version: 0.10.0
• Python version: 3.11.8
• C++ compiler:
• Operating system: Linux blue2 5.10.0-20-amd64

Suggestions

No response

[bmhowe23]

Thanks for reporting this issue, @psm-optimizes. Based on a quick look, it looks like there is bug in how CUDA-Q is handling this section of code:

    # Apply y rotations on Ancilla qubit
    for i in range(len(qReg)+1):
        if i != 0:
            ry.ctrl((2*np.pi)/eigs[i-1], [qReg[i-1]], qAnc)  

While we investigate the underlying issue a bit further, I figured it would be worth sharing that there is a workaround available. If you change your code to this, then I think it will work.

    # Apply y rotations on Ancilla qubit
    for i in range(1, len(qReg)+1):
        ry.ctrl((2*np.pi)/eigs[i-1], [qReg[i-1]], qAnc)  

After resolving this issue, the next error message that comes up is RuntimeError: Applying controlled global phase via exp_pauli of identity operator is not supported. That message is occurring because CUDA-Q does not support calling exp_pauli on terms with all Is, so I believe you will need to make this change:

# Drop the `II` term. Adjust downstream calculations as needed.
words = words[1:]
coefficients = coefficients[1:]

Hopefully this helps.