Filip LeonarskiandClaude Opus 5.5 ba3f754752 RESULT-CHANGING (owner approved, ulp level): CPU profile fit in the GPU engine's arithmetic ...

Pass B of BraggIntegrationEngineCPU now does what BraggIntegrationEngineGPU's fit does: profiles in
float, the Gaussian from expf with the widths' reciprocals taken once and normalised by a reciprocal,
and the reweighted fit in float with one reciprocal of the variance for its three sums (the GPU's
__frcp_rn). Per-reflection geometry and the final sigma stay in double. The logic was already the
same on both; the CPU now differs from the GPU only in summation order.

Measured (CPU-only, Apple M6): Pass B loops 1.2x faster on P cores; cytochrome C 1.9%, myoglobin 1.5%
faster end to end, 4nwv integration steps 3%. Intensities move by ~1e-6 relative; merged statistics
in the fourth digit (error model b 1.3259e-3 -> 1.3258e-3, R_work 0.2440 -> 0.2439).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

2026-10-14 19:42:31 +02:00

..

azint

GPU engines wait for their stream before their buffers go; a lost context fails where it is seen

2026-10-10 10:59:24 +02:00

beam_stop

rugnux pre-scan: projection and defective pixels per card, the sample read once, beam stop memoised

2026-10-12 08:11:05 +02:00

bragg_integration

RESULT-CHANGING (owner approved, ulp level): CPU profile fit in the GPU engine's arithmetic

2026-10-14 19:42:31 +02:00

bragg_prediction

Bragg prediction: rotation partiality in the tail from erfc, the same on CPU and GPU

2026-10-13 16:10:38 +02:00

dark_mask_analysis

v1.0.0-rc.148 (#58)

2026-06-08 08:30:35 +02:00

geom_refinement

rugnux hypothesis arms read their residual on the pre-pass frames

2026-10-11 07:12:05 +02:00

hot_pixels

rugnux pre-scan: projection and defective pixels per card, the sample read once, beam stop memoised

2026-10-12 08:11:05 +02:00

image_preprocessing

marCCD upload: stage the stored pixels instead of page-locking the frame in place

2026-10-13 05:21:33 +02:00

indexing

CudaFFTPlan: wait for the plan's NULL-stream setup before it is used, output unchanged

2026-10-14 16:40:34 +02:00

lattice_search

v1.0.0-rc.173 (#83)

2026-09-29 15:57:32 +02:00

roi

GPU engines wait for their stream before their buffers go; a lost context fails where it is seen

2026-10-10 10:59:24 +02:00

rotation_indexer

Indexer failures are thrown, not read as "no lattice"; GPU memory errors say what to do

2026-10-10 18:56:00 +02:00

scale_merge

tNCS: the modulation null's searches on several workers, output unchanged

2026-10-14 19:42:31 +02:00

spot_finding

AdaptiveSpotFinderCPU: ring histogram indexed [value][ring], output unchanged

2026-10-14 19:42:31 +02:00

structure_refinement

MR score-engine cache and merge buffers: no hang, no doubled wait; near-tie doc follows offline decide-late

2026-10-13 21:25:15 +02:00

CMakeLists.txt

Move the atomic-model code to image_analysis/structure_refinement/ and WriteModel to writer/

2026-10-07 14:05:37 +02:00

IndexAndRefine.cpp

Revert "PROTOTYPE (owner decision): ffbidx stills index each frame with one call on all spots"

2026-10-14 14:58:34 +02:00

IndexAndRefine.h

PROTOTYPE (owner decision): stills min-pix choice searches the lattice once per frame, ~2x faster

2026-10-14 10:34:04 +02:00

IntegrationOutcome.h

v1.0.0-rc.172 (#82)

2026-09-22 06:48:37 +02:00

LoadFCalcFromMtz.cpp

v1.0.0-rc.173 (#83)

2026-09-29 15:57:32 +02:00

LoadFCalcFromMtz.h

v1.0.0-rc.173 (#83)

2026-09-29 15:57:32 +02:00

MXAnalysisAfterFPGA.cpp

1.0.0-rc.174 (#84)

2026-10-06 14:03:18 +02:00

MXAnalysisAfterFPGA.h

v1.0.0-rc.156 (#66)

2026-07-03 19:18:56 +02:00

MXAnalysisWithoutFPGA.cpp

PROTOTYPE: the min-pix 1 search gate applies only to de-novo indexing

2026-10-14 14:58:34 +02:00

MXAnalysisWithoutFPGA.h

rugnux two-pass: the integration radius decided on the pre-pass frames, before the canonical pass

2026-10-11 13:01:23 +02:00

ReflectionArena.h

v1.0.0-rc.172 (#82)

2026-09-22 06:48:37 +02:00

SensorAbsorption.h

v1.0.0-rc.173 (#83)

2026-09-29 15:57:32 +02:00

UpdateReflectionResolution.cpp

v1.0.0-rc.167 (#77)

2026-09-09 07:25:13 +02:00

UpdateReflectionResolution.h

v1.0.0-rc.172 (#82)

2026-09-22 06:48:37 +02:00

WriteReflections.cpp

RESULT-CHANGING (owner approved): stills unmerged MTZ written only for the crystals on the run's cell

2026-10-14 13:15:13 +02:00

WriteReflections.h

Merged mmCIF: drop the scaled unmerged block

2026-10-09 18:51:56 +02:00