From eb6589f6473431d25b244c5cec045ffe4e249d00 Mon Sep 17 00:00:00 2001
From: Filip Leonarski <filip.leonarski@psi.ch>
Date: Sat, 19 Sep 2026 23:22:12 +0200
Subject: [PATCH] rugnux: sixteen-bit Miller indices in the ingest sort key and
 the post-refine partial

Both arrays are one record per integrated observation - tens of millions on a fine-sliced long axis,
gigabytes each - and both carry the raw hkl only to sort and group on. A Miller index needs sixteen
bits (|h| <= a / d_min, in the hundreds even on the longest axis at atomic resolution), which takes
the ingest sort key from 24 to 20 bytes and the post-refine partial from 32 to 28.

Same comparisons, same order; merged output byte-identical.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013nW6FNRP1bBJJ8pfHiByAT
---
 image_analysis/geom_refinement/PostRefine.cpp     | 4 +++-
 image_analysis/geom_refinement/PostRefine.h       | 5 ++++-
 image_analysis/scale_merge/RotationScaleMerge.cpp | 3 ++-
 image_analysis/scale_merge/RotationScaleMerge.h   | 8 +++++---
 4 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/image_analysis/geom_refinement/PostRefine.cpp b/image_analysis/geom_refinement/PostRefine.cpp
index 4ad21536e..3d80f578e 100644
--- a/image_analysis/geom_refinement/PostRefine.cpp
+++ b/image_analysis/geom_refinement/PostRefine.cpp
@@ -219,7 +219,9 @@ PostRefineObservations GatherPostRefineObservations(std::vector<IntegrationOutco
                     if (!std::isfinite(r.I) || !std::isfinite(r.sigma) || r.sigma <= 0.0f) continue;
                     const float ox = std::isfinite(r.observed_x) ? r.observed_x : NAN;
                     const float oy = std::isfinite(r.observed_y) ? r.observed_y : NAN;
-                    pts[fill[r.h - h_lo]++] = Partial{r.h, r.k, r.l, r.image_number, r.I, r.sigma, ox, oy};
+                    pts[fill[r.h - h_lo]++] = Partial{static_cast<int16_t>(r.h), static_cast<int16_t>(r.k),
+                                                      static_cast<int16_t>(r.l), r.image_number,
+                                                      r.I, r.sigma, ox, oy};
                 }
                 // The caller said it will never read this image's reflections again, so hand the
                 // vector back the moment it is consumed: the payload shrinks image by image as it
diff --git a/image_analysis/geom_refinement/PostRefine.h b/image_analysis/geom_refinement/PostRefine.h
index b6f20833c..15d1ff4ac 100644
--- a/image_analysis/geom_refinement/PostRefine.h
+++ b/image_analysis/geom_refinement/PostRefine.h
@@ -94,7 +94,10 @@ struct PostRefineSettings {
 // per-bucket sort move all of it. The goniometer angle is not stored - it is a function of the
 // image number alone, and is rebuilt from it where it is needed.
 struct PostRefinePartial {
-    int h, k, l;
+    // Sixteen bits are plenty for a Miller index - |h| <= a / d_min, in the hundreds even on the
+    // longest axis at atomic resolution - and here four bytes of every partial are four bytes of an
+    // array that is gigabytes on a large cell.
+    int16_t h, k, l;
     float img;
     float I, sigma;
     float obs_x, obs_y;  // observed spot centroid (pixels); NAN if the box sum found no centroid
diff --git a/image_analysis/scale_merge/RotationScaleMerge.cpp b/image_analysis/scale_merge/RotationScaleMerge.cpp
index 93c39fea9..13e526acf 100644
--- a/image_analysis/scale_merge/RotationScaleMerge.cpp
+++ b/image_analysis/scale_merge/RotationScaleMerge.cpp
@@ -432,7 +432,8 @@ void RotationScaleMerge::Ingest() {
             for (int o = lo; o < hi; ++o) {
                 int32_t at = src_start[o];
                 for (const auto &r : partials_out[o].reflections) {
-                    keys[fill[r.h - h_min]++] = SortKey{r.h, r.k, r.l, r.image_number, at, r.d};
+                    keys[fill[r.h - h_min]++] = SortKey{static_cast<int16_t>(r.h), static_cast<int16_t>(r.k),
+                                                        static_cast<int16_t>(r.l), r.image_number, at, r.d};
                     ++at;
                 }
             }
diff --git a/image_analysis/scale_merge/RotationScaleMerge.h b/image_analysis/scale_merge/RotationScaleMerge.h
index 5fd1e8ca8..5731f5c11 100644
--- a/image_analysis/scale_merge/RotationScaleMerge.h
+++ b/image_analysis/scale_merge/RotationScaleMerge.h
@@ -142,10 +142,12 @@ private:
 
     // The narrow per-observation record the ingest sort orders: the raw hkl the runs are cut on, the
     // frame position that breaks a tie inside one, the observation's own index (which makes the order
-    // total - see the .cpp), and the resolution the range test reads. Twenty-four bytes against the
-    // Obs's eighty, and it is all the ingest needs before it knows which observations survive.
+    // total - see the .cpp), and the resolution the range test reads. Twenty bytes against the Obs's
+    // eighty, and it is all the ingest needs before it knows which observations survive. Sixteen bits
+    // are plenty for a Miller index - |h| <= a / d_min, in the hundreds even on the longest axis at
+    // atomic resolution - and on a long axis this array sets the sort's memory high-water mark.
     struct SortKey {
-        int32_t h, k, l;
+        int16_t h, k, l;
         float image_number;
         int32_t idx;
         float d;