OPT-11 recursive split-and-merge hull reduction

Native record opt11-8cfbc0c65f420dfd; measured source commit 488d4ca8888d91c44b64d5d2f0baf564d1c28ae7. Schema csl.opt11-recursive-hull-report/1.

Three fixed two-dimensional signed, duplicate and affine-degenerate fixtures. All 1274 emitted deletion certificates passed exact dyadic validation; literal monotone-chain hulls and retained sets were checked independently where reducer output was available. 0 of 42 calls were incomplete or diagnostic, and remain visible in the trial table and raw rows. Exact dyadic deletion certificates and exact two-dimensional hull equivalence on three fixed finite binary64 point clouds.

Paired observations

OPT-11 paired median call durationsDescriptive medians of six alternating paired calls per fixture. No speedup claim.Median duration (nanoseconds)0904300180860027129003617200signed-grid baselinesigned-grid recursiveduplicate-collinear baselineduplicate-collinear recursiverank-one-offset baselinerank-one-offset recursive
Median public-call duration in nanoseconds. Six alternating paired observations per fixture; each axis starts at zero. The table retains individual observations.
fixturepointsmeasured_pairsbaseline_median_nsbaseline_min_nsbaseline_max_nsrecursive_median_nsrecursive_min_nsrecursive_max_nsmedian_paired_delta_recursive_minus_baseline_nsfailed_pairs
signed-grid3362704100.0264970028037002786750.026144003028500117950.00
duplicate-collinear3362778150.0259460028794002822650.021183003141500-4900.00
rank-one-offset3362685650.0259340028444003196350.024395003617200470550.00

All paired trials

fixtureroundwarmuprecursive_firstbaseline_nsrecursive_nspaired_delta_recursive_minus_baseline_nsbaseline_lp_callsrecursive_lp_callsleaf_lp_callsmerge_lp_callspeak_merge_pointsconcurrent_child_subtreesparallel_overhead_microsecondsbaseline_statusrecursive_statusstatus
signed-grid0TrueFalse39456003557100-3885003345331272816completecompletecomplete
signed-grid1FalseTrue271530028496001343003345331272650completecompletecomplete
signed-grid2FalseFalse264970028098001601003345331272652completecompletecomplete
signed-grid3FalseTrue2712200271710049003345331272644completecompletecomplete
signed-grid4FalseFalse28037002614400-1893003345331272613completecompletecomplete
signed-grid5FalseTrue269600030285003325003345331272975completecompletecomplete
signed-grid6FalseFalse266210027637001016003345331272642completecompletecomplete
duplicate-collinear0TrueTrue27428002402200-340600333932742590completecompletecomplete
duplicate-collinear1FalseFalse28347002118300-716400333932742406completecompletecomplete
duplicate-collinear2FalseTrue27609002311700-449200333932742557completecompletecomplete
duplicate-collinear3FalseFalse27480003141500393500333932742397completecompletecomplete
duplicate-collinear4FalseTrue27954002601900-193500333932742436completecompletecomplete
duplicate-collinear5FalseFalse25946003043400448800333932742385completecompletecomplete
duplicate-collinear6FalseTrue28794003063100183700333932742424completecompletecomplete
rank-one-offset0TrueFalse27005003096100395600333932742375completecompletecomplete
rank-one-offset1FalseTrue27041003164600460500333932742417completecompletecomplete
rank-one-offset2FalseFalse26672003077800410600333932742372completecompletecomplete
rank-one-offset3FalseTrue27475003228100480600333932742407completecompletecomplete
rank-one-offset4FalseFalse259340036172001023800333932742459completecompletecomplete
rank-one-offset5FalseTrue28444003356600512200333932742428completecompletecomplete
rank-one-offset6FalseFalse26399002439500-200400333932742575completecompletecomplete

Performance interpretation: Six alternating paired attempts per fixture; 0 of 42 reducer calls were not complete. The timing summary includes every pair, including incomplete outcomes; interpretation is inconclusive with no universal speedup or scalability claim.. Timing is one public registered run_with call. Fixture parsing, reducer/injected-workspace creation, exact validation, serialization and rendering are excluded; recursive child LP-workspace creation and root task scheduling are included. Backend phase times, native thread count, CPU time, peak memory and source-mapped limiter are unavailable.

LP backend: csl.lp.highs version 1.0.0+highs.1.11.0, reported by the executable's registered LP provider. Configured threads: 1; presolve: off; parallel mode: off. Observed native thread count remains unavailable.

Host: Windows-11-10.0.22631-SP0; machine AMD64; logical processors 12. Compiler: MSVC 19.44.35215.0; C++ setting stdcpp20, optimization MaxSpeed, NDEBUG enabled. Concurrent work: No deliberate competing benchmark or build workload was started. A five-second host sample reported 4% then 7% total processor use across 12 logical processors; no observed process accumulated more than 1 CPU-second in that interval. Ordinary background processes were not stopped. The measurement collector will retain before/after host snapshots and matching-process CPU deltas grouped by process name.. Machine-load snapshots and aggregate CPU seconds for matching processes: {"after": {"cpu": "12th Gen Intel(R) Core(TM) i5-12400F", "load_percent": [8], "logical_processors": 12, "os": "Windows-11-10.0.22631-SP0", "python": "3.13.5", "ram_bytes": 34189107200, "utc": "2026-10-18T17:31:22Z"}, "before": {"cpu": "12th Gen Intel(R) Core(TM) i5-12400F", "load_percent": [5], "logical_processors": 12, "os": "Windows-11-10.0.22631-SP0", "python": "3.13.5", "ram_bytes": 34189107200, "utc": "2026-10-18T17:31:18Z"}, "process_cpu_deltas": {"available": true, "scope": "matching observed processes only; no arguments/paths, new/exited processes omitted", "unit": "process CPU seconds over collection interval", "values": {"AdobeIPCBroker": 0.0, "ApplicationFrameHost": 0.0, "AvastUI": 0.0, "CCXProcess": 0, "ChatGPT": 0.015625, "DTShellHlp": 0.0, "DataExchangeHost": 0.0, "FLPService": 0.015625, "MSBuild": 0.0, "OpenConsole": 0.0, "PhoneExperienceHost": 0.0, "RedShieldVPN": 0.0, "RuntimeBroker": 0.0, "SearchHost": 0.0, "SecurityHealthSystray": 0.0, "ShellExperienceHost": 0.0, "StartMenuExperienceHost": 0.0, "SystemSettings": 0.0, "SystemSettingsBroker": 0.0, "Telegram": 0.03125, "TextInputHost": 0.0, "WINWORD": 0.0, "WebKitNetworkProcess": 0.0, "WidgetService": 0.0, "Widgets": 0.0, "WindowsTerminal": 0.0, "WizTree64": 0.703125, "cbservice": 0.046875, "chrome": 0.0, "cmd": 0, "cnqtbapp": 0.0, "codex": 0.0, "codex-code-mode-host": 0.0, "codex-computer-use-swift": 0.0, "conhost": 0.0, "ctfmon": 0.0, "dllhost": 0.0, "explorer": 0.0, "msedge": 0.0, "node": 0.0, "node_repl": 0.0, "opencode": 0.0, "pdf24": 0.0, "pwsh": 0.0, "python": 0.0, "rhinomcp": 0, "sihost": 0.0, "sketchup-mcp": 0, "svchost": 0.0, "tailscale-ipn": 0.0, "taskhostw": 0.0, "uv": 0.0, "vctip": 0.0}}}. Full sanitized Release and compiler evidence are in the build log and CMake compiler identification.

Saved evidence

Performance is exploratory and fixture-limited. No universal speedup, scalability, or correctness claim for arbitrary point clouds follows.

Publication opt11-a43ae7725dd3ab72d2034473aeb6c5805f2d1820b2ccd70b8383ef300a4b9865; native report opt11-8cfbc0c65f420dfd. Versioned envelope ยท Publication manifest.