Repository navigation
v8.3.1
Reversible arithmetic (#337, #338, #339)
Basis permutations U|x⟩ = |f(x)⟩ on registers, checked against the reference circuits of Vedral–Barenco–Ekert (quant-ph/9511018), Cuccaro et al. (quant-ph/0410184), Draper (quant-ph/0008033) and Beauregard (quant-ph/0205095).
add_registers,subtract_registers,add_constant(c23f409).compare_registers,compare_constant(a8add5d).add_registers_mod,add_constant_mod(with controls),multiply_add_mod,multiply_mod,power_mod(dc7525b).- complex128 kept: x64 enabled before building JAX arrays (fc78b15); error paths and NumPy path covered by tests (e9b2717, 1d76a78, ab3e72b).
Postselection (#340)
postselect(sv, n_qubits, qubit, state): projects a qubit onto0,1,+,-,+ior-i, returns the renormalised state and its probability (Aaronson, quant-ph/0412187) (4e8f68f, f3f3e99).
Parser: user-defined gates (#341)
- OpenQASM 2
gatedefinitions are expanded, including nested and parametrised ones; Qiskit circuits that define their own gates (e.g.mcx) now parse and run (4da3423, 8aca908).
Mitigation
zne_shot_allocation(#344): splits a shot budget across ZNE noise factors in proportion to |Richardson coefficient| × σ (Temme–Bravyi–Gambetta, PRL 119, 180509, Eq. 3); about 14% lower variance than a uniform split at factors 1, 2, 3 (63ad3ba, 1853131).bloch_zne,bloch_zne_jit(#345): single-qubit Richardson on the Bloch vector, projected back onto the Bloch ball (Smolin–Gambetta–Smith, arXiv:1106.5458) (aa82fbc).bounded_exponential_extrapolate: documented limitation with three noise factors and shot noise (aa82fbc).- 64-bit precision enabled at every public mitigation entry point (
with_x64) (#343) (55d8ab4, 4d8f90c, fc54f6d).
Dashboard (#335)
- Streamlit Dashboard moved into
tools/dashboard/: English UI, circuit editor, engine explorer, installers for Windows, macOS and Linux,dashboard-installersworkflow (7a517ea, 979785e).
Citation
Full diff: v8.3.0...v8.3.1