Fixes for wire management in Catalyst

[dime10]

This PR contains several improvements regarding the wire handling in Catalyst:

• the runtime will now raise an exception when a qubit is accessed out of bounds, which improves memory safety, and doesn't require the backend to raise an invalid qubit error
• the arbitrary restriction of requiring at least one qubit to run a quantum circuit is lifted (now follows PL behaviour)
• the QJITDevice classes in the frontend now perform stricter validation of existing assumptions on the wires property of devices

Related GitHub Issues: #780

[sc-65105]

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.09%. Comparing base (ec9c826) to head (664d31b).
Report is 188 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #784   +/-   ##
=======================================
  Coverage   98.08%   98.09%           
=======================================
  Files          70       70           
  Lines        9731     9740    +9     
  Branches      780      782    +2     
=======================================
+ Hits         9545     9554    +9     
  Misses        151      151           
  Partials       35       35           

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

[WrathfulSpatula]

@dime10 This does, in fact, allow allocating 0 qubits (at least trivally) and provides a helpful error message when wire count is too low the requested circuit.

Thank you! LGTM! (With that said, the issue about suspicious "extra" phase gates in tutorial_qft.py remains, as per the issue.)

[dime10]

@dime10 This does, in fact, allow allocating 0 qubits (at least trivally) and provides a helpful error message when wire count is too low the requested circuit.

Thank you! LGTM! (With that said, the issue about suspicious "extra" phase gates in tutorial_qft.py remains, as per the issue.)

Thank you! I will have another look at the original issue then :)

[erick-xanadu]

Do you know why it fails on kokkos?

[dime10]

Do you know why it fails on kokkos?

Hmm, will need to investigate :/

[erick-xanadu]

@dime10 it is an issue with kokkos:

>>> dev = qml.device("lightning.kokkos", 0)
Kokkos::OpenMP::initialize WARNING: OMP_PROC_BIND environment variable not set
  In general, for best performance with OpenMP 4.0 or better set OMP_PROC_BIND=spread and OMP_PLACES=threads
  For best performance with OpenMP 3.1 set OMP_PROC_BIND=true
  For unit testing set OMP_PROC_BIND=false

>>> qml.qnode(dev)(lambda : qml.state())()
Segmentation fault (core dumped)

[dime10]

@dime10 it is an issue with kokkos:

>>> dev = qml.device("lightning.kokkos", 0)
Kokkos::OpenMP::initialize WARNING: OMP_PROC_BIND environment variable not set
  In general, for best performance with OpenMP 4.0 or better set OMP_PROC_BIND=spread and OMP_PLACES=threads
  For best performance with OpenMP 3.1 set OMP_PROC_BIND=true
  For unit testing set OMP_PROC_BIND=false

>>> qml.qnode(dev)(lambda : qml.state())()
Segmentation fault (core dumped)

And it only happens with 0 qubits. Could be an upstream issue then, while default.qubit and lightning.qubit work fine in that regime, lightning.kokkos might not.

[dime10]

@erick-xanadu I've disabled those tests on kokkos, don't think it's particularly important that they run on all devices. It's mainly to see that Catalyst allows 0 qubits end to end.

[erick-xanadu]

Just that minor comment.