Hybrid Workflow Patterns — VQE and SQD at HPC Scale
VQE and SQD are already covered in depth as algorithms — this note is about what HPC specifically contributes around them, not how they work internally.
VQE’s HPC footprint
HPC generates the molecular integrals and basis sets classically before any quantum circuit runs; the QPU only evaluates the parameterized ansatz. Key insight — VQE is not a single computation, it’s repeated classical-quantum loops, and HPC’s job is to batch, parallelize, and validate those loops across many nodes rather than run them one at a time. The same trainability/barren-plateau and shot-count challenges from the algorithm itself apply here — HPC scale doesn’t remove them, it just gives you more parallel attempts to work around them.
SQD’s HPC footprint
HPC orchestrates and preprocesses (one- and two-body electronic integrals, qubit mapping, circuit preparation) before sampling; after the QPU produces bitstrings, classical HPC — CPU/GPU, MPI-based, multi-node — constructs the effective subspace and performs the diagonalization. The iterative configuration recovery step scales the same way: more classical compute means more samples can be corrected and re-diagonalized per round.
The common pattern
Both are the same shape: a narrow, specialized quantum kernel embedded inside a much larger classical workflow. The QPU never sees molecular integrals, optimizer state, or subspace diagonalization directly — it only ever sees a circuit and returns bitstrings or expectation values. Everything else, at whatever scale, is HPC’s job.
Related
- Variational Quantum Eigensolver (VQE)
- Sample-Based Quantum Diagonalization (SQD)
- Why Quantum Needs HPC
- The Quantum+HPC Software Stack
Self-Check
- What does HPC actually do in a VQE run that the QPU itself never sees?
- Why does SQD’s configuration recovery step specifically benefit from more classical compute?
- Could you describe the “common pattern” both algorithms share, in one sentence?