[macem-release]

🤖 I have created a release beep boop

0.10.0 (2026-11-01)

⚠ BREAKING CHANGES

• CuboPtycho, CuboStrain and CuboEmbedding (the enums) and their deprecated scope are removed; nothing replaces them. Module-qualified names such as CuboStrain.LatticeStrain now resolve. mac4DSTEM uses none of the three enums or scope, so it needs no call-site change.

Features

• Bragg disk detection .single runs on every detector size (#61) (f60ef41)
• brightest-disk seed for the iterated-window origin at low dose (#65) (d4231bc)
• deprecate the placeholder scope constants and drop dead code (#41) (f33b33e)
• deprecate the placeholder scope constants; drop dead code (f33b33e)
• remove the enums named like their modules (2e4c0f6)
• single-slice ptychography .single takes every detector size (#59) (5d40072)
• SyntheticProbe centre, public FloatFFT2D.supports, consensus diagnostics for a supplied basis (#46) (98f5330)

Bug Fixes

• Metal sums give +-Inf, not NaN, for infinite pixels (#51) (5642ee1), closes #12

Performance Improvements

• CPU mean and max pattern 2x faster, TwoSum state in registers (#53) (a8b268e)
• CPU mean pattern 2x faster, sum held in registers over four positions (a8b268e)
• Float32 parallax shifter for every size, Bluestein on the vDSP Float DFT (#52) (60ddbea)
• masked and weighted CPU mean about 1.85x faster, sum in registers (#55) (1e4a716)
• Metal reducers 2x to 5x faster on a cube that is not resident (#57) (117a1f9)
• multicorr Bragg refinement 2x faster, loops ordered for contiguous reads (#63) (b8e85f9)
• multicorr refinement 2x faster, loops ordered for contiguous reads (b8e85f9)
• multicorr refinement as matrix products on Accelerate, 3.3x faster per cube (#66) (1130f2f)

---

This PR was generated with Release Please. See documentation.

[macem-release]

🤖 Created releases:

• 0.10.0

🌻