[qref 2.5] Subroutines take in extracted quantum.bit values during semantics conversion

[paul0403]

Context:
When converting subroutines from reference semantics to value semantics, currently we take in whole registers as-is:

func.func @subroutine(%r: !qref.reg<1>) {
    %q = qref.get %r[0] : !qref.reg<1> -> !qref.bit
    qref.custom "PauliX"() %q : !qref.bit
    return
}

func.func @main() attributes {quantum.node} {
    %r = qref.alloc(1) : !qref.reg<1>
    func.call @subroutine(%r) : (!qref.reg<1>) -> ()
    qref.dealloc %r : !qref.reg<1>
    return
}

  func.func @subroutine(%arg0: !quantum.reg) -> !quantum.reg {
    %0 = quantum.extract %arg0[ 0] : !quantum.reg -> !quantum.bit
    %out_qubits = quantum.custom "PauliX"() %0 : !quantum.bit
    %1 = quantum.insert %arg0[ 0], %out_qubits : !quantum.reg, !quantum.bit
    return %1 : !quantum.reg
  }
  func.func @main() attributes {quantum.node} {
    %0 = quantum.alloc( 1) : !quantum.reg
    %1 = call @subroutine(%0) : (!quantum.reg) -> !quantum.reg
    quantum.dealloc %1 : !quantum.reg
    return
  }

However, this is obviously not optimal, since the subroutine call, although only on a subset of qubits, will interject into the data flow of all qubit values in that register.

If we can statically know which qubits the subroutine needs, we can just extract before the call:

  func.func @subroutine(%arg0: !quantum.bit) -> !quantum.bit {
    %out_qubits = quantum.custom "PauliX"() %arg0 : !quantum.bit
    return %out_qubits : !quantum.bit
  }
  func.func @main() attributes {quantum.node} {
    %0 = quantum.alloc( 1) : !quantum.reg
    %1 = quantum.extract %0[ 0] : !quantum.reg -> !quantum.bit
    %2 = call @subroutine(%1) : (!quantum.bit) -> !quantum.bit
    %3 = quantum.insert %0[ 0], %2 : !quantum.reg, !quantum.bit
    quantum.dealloc %3 : !quantum.reg
    return
  }

Description of the Change:
Extract qubits from static indices before the callsites of the subroutine, and use the extracted qubits instead.

Benefits:
More optimized quantum dataflow; unified treatment of how we handle subroutines vs other regions (control flow and adjoint) during the qref-to-quantum semantics conversion.

Possible Drawbacks:
None

Related GitHub Issues:
[sc-115119]

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.12%. Comparing base (1959863) to head (e6128c2).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2642      +/-   ##
==========================================
- Coverage   97.15%   97.12%   -0.04%     
==========================================
  Files          36      157     +121     
  Lines        4295    17920   +13625     
  Branches        0     1709    +1709     
==========================================
+ Hits         4173    17405   +13232     
- Misses        122      378     +256     
- Partials        0      137     +137     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[rniczh]

🚀 @paul0403 , just some minor comments from my side. Overall looks good to me.

[mehrdad2m]

Thanks @paul0403, Just a few minor comments

[dime10]

I'm a bit confused about the organization of the code into the two headers and the source file (i.e. why code is one place or another), but C++ can be confusing that way I suppose 😅