guppy-estimand: fault-tolerant resource estimates for Guppy programs.
https://pypi.org/project/guppy-estimand/
Give it a compiled Guppy/HUGR program, get back physical qubit count, runtime and error budget under a surface-code model. Gate counts come from walking the HUGR; costing runs through Qualtran's Beverland et al. models via an adapter, not from-scratch QEC math.
Example, QFT on 6 qubits (beverland scheme, d=17): 30 rotations, 8,082 physical qubits, error budget 6.7e-3.
v0.1, experimental:
- surface code only
- bounded control flow (conditionals as upper bound, loops with stated trip counts); raises an error rather than guess on anything it can't bound
- guppylang 1.2.0, Python 3.12+
Building it surfaced two discrepancies between Qualtran's Beverland model and the paper (quantumlib/Qualtran #1943, #1944), both confirmed by the maintainer. Fix for #1943 in PR #1949.
pip install guppy-estimand (uv recommended, qualtran's deps are heavy)
Feedback welcome, especially on which QEC schemes or programs to try next.
