Direct ptychography reports

• Exact broad-search density batching

Dated records retain their original workload, backend, source revision and acceptance limits. They are evidence, not a statement of current release speed. A record that describes since-retired behaviour (the phase-variance search, Nelder–Mead or finite-difference refinement, the mean_phase and complex_wave estimators) says so in a Historical line at its top.

• Native Direct Ptycho preparation and exact QEM column decoding

• Full-scan native Metal fitting on Apple M5

• Exact-derivative direct ptychography fit in Python on CUDA and MPS

• Native Metal direct ptychography redraw precision at the lowest scan frequencies

• Where the 9 s of regularized preparation goes

• Direct phase reconstruction and viewer handoff

• Which reconstruction to measure drift on

• Live registration notch beyond frame 0

• Redraw at slider speed: q, -q symmetry for direct ptychography and a drag window for regularized methods

• Native Swift/Metal direct ptychography redraw while dragging aberration controls

• Lower CUDA direct ptychography staging memory without changing numerical results

• Initial fitting before fifth-order refinement

• Reuse Fourier preparation in four-case direct ptychography review

• Faster broad direct ptychography search with identical results

• CUDA broad-fifth research workflow

• CUDA direct ptychography region latency — 2026-10-21

• Sign of the odd aberrations in direct ptychography reconstruction (2026-10-21)

• direct ptychography joint-fit benchmark, 2026-10-21

• Coma, trefoil and spherical aberration in the model-based direct ptychography fit (2026-10-21)

• The MPS standard direct ptychography fit uses the agreement objective (2026-10-20)

• Tilt-corrected direct ptychography on CUDA: fit and preview speed (2026-10-20)

• CUDA standard fit on the agreement objective, and one phase-variance search (2026-10-20)

• Tilt-corrected direct ptychography on Metal: preview and fit speed (2026-10-20)

• Separable C10/C12 coherent direct ptychography on Metal, 2026-10-20

• MPS performance: what got faster, why the rest is slow, how to profile (2026-10-20)

• MPS memory: where direct ptychography’s gigabytes go and what fits which Mac (2026-10-20)

• Reuse source addresses in MPS direct ptychography reconstruction

• Exact group support in CUDA direct ptychography

• MPS coherent direct ptychography dispatch

• MPS direct ptychography preview reuse

• MPS fine-grid optical calculations, 2026-10-17

• Full BF MPS redraw comparison, 2026-10-17

• MPS coherent direct ptychography redraw — 2026-10-17

• Exact aperture bounds for interactive MPS direct ptychography

• Reuse MPS loading buffers and decode detector regions directly

• MPS direct ptychography correction and reduction fusion — 17 October 2026

• MPS direct ptychography execution and coherent reconstruction, 2026-10-17

• MPS direct ptychography coherent Fourier support — 17 October 2026

• MPS slider correction and interruptible diagnostics

• Coherent direct ptychography reduction on Metal

• Coherent direct ptychography reconstruction and scan-mean aliases

• MPS direct ptychography sampling and search, 2026-10-15

• CUDA direct ptychography refinement batching

• Later direct ptychography performance checkpoints

CUDA fine-grid kernel experiments, 23-24 October 2026

These records measured in-kernel q, -q pairing, compact support and closed-sideband skips on a separate branch. Main took a different design (half-plane spectrum with the -conj mirror, bounded BF groups and drag windows), so the kernels they describe are not in the current source. The numbers and rejected ideas remain as evidence.

• CUDA direct ptychography: shared BF coordinates and inactive blocks

• CUDA direct ptychography compact support and boundary dispatch

• CUDA direct ptychography closed-sideband correction and host transfer

• Skip closed-aperture loads in fine-grid direct ptychography

• Full-field direct ptychography redraws, 2026-10-23

• Native 512 direct ptychography instruction-reduction experiments

• Native 512 direct ptychography hardware-counter diagnosis

• Native 512 direct ptychography frequency pairing

• CUDA direct ptychography fine-grid correction

• Classify fine-grid direct ptychography mirror reuse once per frequency

• Native 512 direct ptychography: fixed-optics caching and launch layout

• Higher-order direct ptychography: group alias-edge work on CUDA