v2026.9.11.1

PyAutoArray v2026.9.11.1

What's New

Breaking Changes

• feat: per-call arcsec_after_decimal override for arcsecond tick labels (#547)
• feat: numba CPU interferometer curvature path — direct_conv, geometry-gated (#543) (#545)
  • Additive only — nothing removed, renamed or re-signatured.
  • • New class InversionInterferometerSparseNumba and its kernel module (a new
• perf: NumPy/scipy application path for InterferometerSparseOperator (xp=np never imports JAX) (#544)
  • Additive only. Six InterferometerSparseOperator methods gain a keyword-only xp=np
  • argument, defaulting to the NumPy branch. The dataclass gains a
  • nufft_precision_operator field (the raw preload, previously discarded after the FFT) and
• perf: build nufft_precision_operator_from as a type-1 NUFFT (#541)
• perf(reg_split): compact the ConstantSplit split-stencil scatter (DelaunayNN 12x on A100) (#537)
  • No breaking changes. pixel_splitted_regularization_matrix_from gains two
  • optional keyword arguments, compact_width and wide_row_budget, defaulting to
  • the new module constants SPLIT_REG_COMPACT_WIDTH = 12 and
• perf: vectorise the over-sampled grid builder; skip the mask-edge overlay under fast plots (#534)
  • No public API change. grid_2d_slim_over_sampled_via_mask_from returns the same values in the same order; a private helper _sub_pixel_offsets_from is added beside it. plot_array behaves identically unless PYAUTO_FAST_PLOTS=1, where it no longer computes the (never-drawn) mask-edge overlay.
• sibson: cut DelaunayNN kernel launches — gated candidate unroll, single concatenated pass, chunk as memory guard (#533)

Internal

• fix(inversion): zero-signal adapt image gives NaN on the JAX path (#549)
• perf: apply the interferometer sparse operator with rfft2/irfft2 (#540)

---

Full changelog: 2026.9.8.1...2026.9.11.1