The new parallel syntax does not interop properly with the (so far unreleased) QIR dynamic qubit allocation feature. Specifically, any calls to __quantum__rt__qubit_release should be defferred using the same logic as the static ID tracking during code gen. The pattern used during normal eval where the simulator handles qubit management can be mimicked, since parallel already forces inlining and unrolling within it's scope.
The new
parallelsyntax does not interop properly with the (so far unreleased) QIR dynamic qubit allocation feature. Specifically, any calls to__quantum__rt__qubit_releaseshould be defferred using the same logic as the static ID tracking during code gen. The pattern used during normal eval where the simulator handles qubit management can be mimicked, sinceparallelalready forces inlining and unrolling within it's scope.