3 Commits

Include renames

Author	SHA1	Message	Date
leonarski_fandClaude Opus 5	4bdb229fb8	spot_finding: find connected components on the GPU ... The spot finder flagged strong pixels on the device and then labelled them on the host, so every frame sent the packed bitmask back - 2.26 MB on a large detector - and the host walked all of it to recover a few hundred pixels. Do the labelling on the device instead: compact the bitmask into a flat-index-sorted list, find each pixel's backward neighbours by binary search, union them lock-free with path halving, then label, accumulate and filter in one kernel. Only the spot list comes back, and only one stream synchronisation per frame. The gain in the ordinary case is modest - about a quarter off per-image spot finding - because the host algorithm is genuinely fast on a normal frame. What justifies it is the frame that is not ordinary. The host labels a sorted sparse list through a window spanning two detector lines, so its cost is quadratic in how many strong pixels share a line. A lit band of detector rows - a hot module, a panel edge - costs 33 ms at two rows and 377 ms at fifteen, all of it under the pixel cap that was supposed to bound this, and none of it maskable when the cause is a diffraction ring rather than a defect: a ring runs tangent to a row at its top and bottom, which is exactly the shape that hurts. The device version is flat at 0.05 to 0.64 ms across every geometry tried, so an online run no longer stalls a quarter of a second on an ice ring. Rejecting an over-cap frame is now free too, since the count is known before any pixel is written. Also label once and filter three times. The per-image minimum-pixel search runs the extraction at three settings, but that setting only decides which components are kept - it does not change the components - so the search itself need not be repeated. This helps the host path as much as the device one. The resolution mask moves to the device as a bit mask, uploaded when the limits change rather than per frame, since the compaction needs it there. Parity is asserted permanently rather than argued: five cases covering realistic frames, occupancy from a hundred pixels to past the cap, the pathological geometries including rings, the resolution mask, and a hundred-repeat determinism check - requiring the same partition, the same spot order, and identical counts. The centroid is a float sum and therefore order-dependent, so the device walks each component from its root in ascending order and fuses its multiply-add the way the host's does; note that whether the host fuses at all depends on the architecture flags, so exact centroid equality is asserted where the compiler fuses and a two-ulp bound otherwise. Making those accumulators integer would remove that dependence entirely and is worth doing separately. Regression set: all 37 crystals identical to the last printed digit. Unit suite passes with the new cases. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-02 13:34:43 +02:00
leonarski_fandClaude Opus 5	aed1a7a6d6	Adaptive spot finder: pin the threshold to the image, and the GPU to itself ... The existing cases plant blobs at 200 on a background of 8..12, so any threshold between 12 and 200 passes them - replacing RingThreshold with a constant leaves them all green. Two cases that do not: - the CPU threshold has to track the background: a frame and the same frame scaled ten times must give the same spots, with a pixel a few sigma above the background staying unfound in both. A constant threshold, or one that drops the sigma term, fails one scale or the other. - the GPU engine has to agree with itself across runs, which is what the ring sums being order-independent buys. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-07-30 11:09:17 +02:00
leonarski_fandClaude Opus 5	0162afd324	Azimuthal integration: report the dimensions of the image the mapping covers ... The mapping took its width and height from the CONVERTED geometry unconditionally, while pixel_to_bin is sized per mode: converted when the geometry is transformed, raw module layout when it is not. In raw mode the two disagreed - 2068x2162 reported against a 1024x4096 map on a JF4M. Only the adaptive spot finders read those dimensions, and they read them for exactly the thing that breaks: the CPU finder derives npix = w*h and then indexes the image, pixel_to_bin and the resolution mask with it, so it walked ~277k pixels past the end of all three; the GPU finder stays in bounds but decodes the strong-pixel bit index with the wrong row stride and reports spots at wrong coordinates. Nothing combines raw geometry with adaptive detection today, so this was latent rather than live. Take them from GetXPixelsNum()/GetYPixelsNum(), which already follow the geometry mode. The converted path is unchanged - it is the same number there - and every internal use is inside SetupConvGeom, which only runs when the geometry is transformed. Covered by two tests: the mapping's dimensions must match pixel_to_bin in both modes, and the CPU adaptive finder must return a spot planted on a raw-geometry image at that raw pixel. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-07-29 09:31:58 +02:00