Skip to content

rqm-compiler

rqm-compiler is the internal optimization and rewriting engine of the RQM stack.

It sits after the public rqm-circuits boundary and owns the internal circuit model, canonical optimization pipeline, and explicit backend-targeted lowering stages.

What rqm-compiler owns now

  • the internal compiler circuit model
  • optimization and rewriting passes
  • canonical internal single-qubit IR: u1q
  • proof-gated optimization commit semantics
  • explicit backend-targeted lowering via lower_circuit_for_backend(...)
  • explicit compile-and-lower flows via compile_for_backend(...)

Current optimization semantics

The current optimization pipeline includes:

  • normalize
  • canonicalize
  • flatten
  • to_u1q
  • merge_u1q
  • sign_canon
  • cancel_2q

Named single-qubit gates are lowered to u1q; backend-targeted lowering comes later.

The trust contract

rqm-compiler optimization is proof-gated and fail-closed.

That means:

  • a candidate optimized circuit may be built internally
  • the candidate is verified before commit
  • only verified candidates are committed
  • if verification is not established, the original circuit is returned unchanged

This is the key public semantic to preserve when documenting the compiler.

What rqm-compiler does not own

  • the canonical public circuit schema or API payload contract
  • auth, billing, or execution policy
  • Studio workflows
  • the quaternion / SU(2) math primitives themselves

Those responsibilities belong to rqm-circuits, rqm-api, rqm-studio, and rqm-core respectively.