[padix-key]

Both functions computed the centroid and plane normal of each aromatic ring in a Python loop (mean() + np.cross() per ring). The per-ring NumPy call overhead dominated the runtime for structures with many rings.

Fix: compute all centroids and normals at once in a new _get_ring_geometry() helper (replaces _get_ring_normal()): centroids via np.add.at() on the concatenated ring coordinates, normals via a single np.cross() on the first three atoms of each ring. np.add.at() was chosen over np.add.reduceat(), as the latter is not bitwise identical to mean().

Centroids and normals are bitwise identical to the previous implementation for all rings in tests/structure/data/pdb/*.bcif; the returned interactions are identical for these structures (also with all nitrogen atoms artificially charged to exercise the pi-cation path).

Input	Function	Before	After	Speedup
1dix (1748 atoms, 37 rings)	find_stacking_interactions	1.11 ms	0.41 ms	2.7x
1igy (12750 atoms, 174 rings)	find_stacking_interactions	5.34 ms	2.00 ms	2.7x
3wip (16917 atoms, 260 rings)	find_stacking_interactions	8.30 ms	3.36 ms	2.5x
3wip (16917 atoms, 260 rings, 1 cation)	find_pi_cation_interactions	6.97 ms	2.05 ms	3.4x

Tests run: tests/structure/test_rings.py, tests/test_doctest.py::test_doctest[biotite.structure].

Authored by Claude

[codspeed]

Merging this PR will regress 2 benchmarks

⚡ 16 improved benchmarks
❌ 2 regressed benchmarks
✅ 91 untouched benchmarks
🆕 2 new benchmarks
⏩ 14 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	benchmark_cell_list_compute_contacts[MAPPING]	22.9 ms	26.7 ms	-14.3%
❌	benchmark_cell_list_compute_contacts[PAIRS]	22.7 ms	25.8 ms	-12.09%
⚡	benchmark_find_rotatable_bonds	177.1 ms	26.7 ms	×6.6
⚡	benchmark_superimpose[superimpose]	3,198.9 µs	646.3 µs	×4.9
⚡	benchmark_hbond	29.7 ms	10.1 ms	×2.9
⚡	benchmark_similar_kmers	483 µs	197.7 µs	×2.4
⚡	benchmark_superimpose_multi_model[superimpose]	14.4 ms	6.4 ms	×2.3
⚡	benchmark_get_assembly	1,594.7 ms	835.8 ms	+90.81%
⚡	benchmark_superimpose[superimpose_homologs]	60.6 ms	33.6 ms	+80.15%
⚡	benchmark_score_threshold_rule[None]	3.3 s	1.9 s	+70.66%
⚡	benchmark_score_threshold_rule[11*11*1*1***111]	3.3 s	1.9 s	+70.64%
⚡	benchmark_superimpose_multi_model[superimpose_homologs]	74.2 ms	44.8 ms	+65.63%
⚡	benchmark_superimpose[superimpose_without_outliers]	31.9 ms	20.8 ms	+53.28%
⚡	benchmark_match_kmer_selection[KmerTable-None]	642.6 µs	527.7 µs	+21.77%
⚡	benchmark_set_structure[bcif-True]	39 ms	32.7 ms	+19.13%
⚡	benchmark_superimpose[superimpose_structural_homologs]	53.4 ms	44.9 ms	+18.99%
⚡	benchmark_match[KmerTable-None]	739.6 µs	625.5 µs	+18.23%
⚡	benchmark_set_structure[bcif-False]	17.7 ms	15.1 ms	+16.93%
🆕	benchmark_stack[10]	N/A	6.7 ms	N/A
🆕	benchmark_stack[2]	N/A	2 ms	N/A

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing padix-key:perf/info-ring-geometry (e27f830) with main (81df912)²}

Footnotes

1. 14 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (b0d32ca) during the generation of this report, so 81df912 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩