The capture scored three surfaces: the 2D point inversion and a 1D line mirror per detector axis
(the Friedel-mate part along the spindle). Only the point surface is used - its strongest candidate
seeds the background walk that measures the beam centre. The line mirrors and their margins were
wired report-only when the capture was added and a consumer for them never landed; their one reader
was the pre-scan's "Beam centre capture:" log line. Removed from the scorer, both engines (fftw3f and
cuFFT: the 1D plans, the gather/accumulate/pack kernels and the line phase of the device-memory
estimate), the tests and the docs; the log line now gives the point centre and its margin.
Output unchanged: p.hkl byte-identical on lysozyme, cytochrome C, myoglobin and 5reo; [BeamCenter]
and [BeamCenterFFTGPU] pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PXc1zD1yxHqaR1DGdCTHL4
leonarski_fandClaude Opus 5.5
f1ed8997d5
