Uniform LES corrected regression — PASS
Exact clean candidate: eec4677db1308e2a9ce4e5f195c994356c11de13; parent: 2cf9853952e52f59bea63cf5c1d3a1cd76e53eed; production unchanged: a135252ae78b47b57d7e77d19001700b3bedac8b
Source fingerprint: a5dbfd804befce868067d79a23e5fd47dce7f505b3f9b1bb2145fb57cb5f2178
Explicit source profile a135252;101 tracked sources hash matched.
ManifestSHA256:f0136b8d435d0c96e5054c3a4c19130097141fd14b2c6676e3edac08c6ab6237
Historical LES profiles byte-identical:{"source_pins.json": "87ca033619da1fc02bee822bfc5681b518a92a155c3c3d1dbbe5265c96066b39", "source_pins_dev8403956.json": "58187908485f2af7c78677d53de98015878372444265ce9e667683b8ef5d410f"}

12 existing host runner/oracle tests PASS,0failures/errors/skips,no PyCUDA/context.
One actual existing runner entrypoint;--experiment a135252 --variant corrected
--run-gpu --device 0;absolute existing Python,explicit candidate import paths.
Physical GPU0,V100,existing startup/compiler lock held;private cache/scratch.
Unchanged8x34,H32,periodicx,halfwayywalls,shiftedFP32 SRT,nu.02,U.04,Delta1.
Cs0/.16/1,checkpoints519/1038/2075;3READY records with one advancing warmup each,
both buffers/flags/count reset as original runner;reported trajectory6228launches.
DONEpassed=true;exit0;active process10.535915s/600s.
Case gate statistics:[{"cs": 0.0, "initial_relative_l2": 1.425896962799328e-08, "warmup_steps": 1, "initial_state_restored": true, "checkpoints": [519, 1038, 2075], "maximum_relative_l2": 0.0010708805559761557, "maximum_oracle_grid_relative": 3.388697083921683e-05, "maximum_mass_drift": 3.1936497890683313e-10, "maximum_transverse": 1.2259988513085908e-08, "maximum_x_invariance": 0.0, "all_observations_passed": true}, {"cs": 0.16, "initial_relative_l2": 1.425896962799328e-08, "warmup_steps": 1, "initial_state_restored": true, "checkpoints": [519, 1038, 2075], "maximum_relative_l2": 0.0010600693163657967, "maximum_oracle_grid_relative": 3.395644410079756e-05, "maximum_mass_drift": 3.3935609877744355e-10, "maximum_transverse": 9.691575434626714e-09, "maximum_x_invariance": 0.0, "all_observations_passed": true}, {"cs": 1.0, "initial_relative_l2": 1.425896962799328e-08, "warmup_steps": 1, "initial_state_restored": true, "checkpoints": [519, 1038, 2075], "maximum_relative_l2": 0.0011344394035840875, "maximum_oracle_grid_relative": 3.6517062854463756e-05, "maximum_mass_drift": 5.818217019282201e-10, "maximum_transverse": 1.0273652040895028e-08, "maximum_x_invariance": 0.0, "all_observations_passed": true}]
SGS increments:[{"elapsed_steps": 519, "measured": 0.0005643205025357007, "expected": 0.0005635139210398016, "relative_error": 0.0014313426266573152}, {"elapsed_steps": 1038, "measured": 0.0009559652045421085, "expected": 0.0009553655762693306, "relative_error": 0.0006276427450102538}, {"elapsed_steps": 2075, "measured": 0.0013869542451979272, "expected": 0.001386869319500092, "relative_error": 6.123554443172746e-05}];existing last-two-time gate retained.
Frozen limits:{"linear_relative_l2": 0.003, "les_relative_l2": 0.005, "sgs_increment_relative_error": 0.15, "oracle_grid_relative": 8e-05, "mass_drift": 5e-05, "max_transverse": 1e-05, "x_invariance": 1e-06, "initial_relative_l2": 1e-05}
Cs1 is only an amplified diagnostic.No synthetic reverse-header GPU run,sweep,
threshold change or scientific rerun.No projected-LES eligibility expansion.

Actual compiler sidecars:3;existing reviewed
capture helper checks PTX bytes before module load and retains metadata before
Simulation cleanup. Independent reused CPU audit verifies exact source/config
digest,unchanged kernels/generatedheaders;no extra compilation or GPU process.
Runtime:{"device": {"name": "Tesla V100-SXM2-16GB", "pci_bus_id": "0000:84:00.0", "capability": [7, 0], "driver": 12040, "runtime": [12, 4, 0]}, "python": "3.10.13 (main, Sep 11 2023, 13:44:35) [GCC 11.2.0]", "executable": "/home/frank14f/anaconda3/envs/pycuda_3_10/bin/python", "numpy": "1.26.4", "pycuda": "2026.1", "pycuda_path": "/home/frank14f/anaconda3/envs/pycuda_3_10/lib/python3.10/site-packages/pycuda/__init__.py", "nvcc": "/usr/local/cuda/bin/nvcc", "nvcc_version": "nvcc: NVIDIA (R) Cuda compiler driver\nCopyright (c) 2005-2024 NVIDIA Corporation\nBuilt on Thu_Mar_28_02:18:24_PDT_2024\nCuda compilation tools, release 12.4, V12.4.131\nBuild cuda_12.4.r12.4/compiler.34097967_0"}
Resources:{"peak_rss_bytes": 288641024, "peak_rss_plus_supervisor_bytes": 314179584, "min_ram_bytes": 63256449024, "min_vram_mib": 15826, "peak_campaign_bytes": 10461184, "peak_total_bytes": 1926033408}
Protection:{"all_worktrees_branches_remote_index_unchanged": true, "candidate_files_unchanged": true, "capture_helper_unchanged": true, "prior_files_unchanged": true, "scratch_clean": true}
All38 worktrees retained;only authorized clean detached candidate re-pinned.
26616 prior output files/source evidence preserved. Original root
index/status/HEAD,other sources,all branches/tags and remote source heads unchanged.
Scratch removed;GPU idle. Staging32MiB headroom verified;cleanup:[].
No environment/dependency/DynamisLab changes.No main/dev merge/default switch.
