10 Recorded test results
10.1 Read the evidence
This page reads actual JSON/TOML files under docs/data at render time. It does not execute the solver, infer a pass from source code, or assign historical campaign results to this checkout. The source revision of this site is 6e3bd67a5, rendered 2026-10-24 16:19 UTC.
Inspect each record’s source fingerprint, command, backend, hardware/platform, timestamp, scope, exit status, and log. NOT_RUN means not measured; PASS applies only to that command and source fingerprint. A later code change can leave an earlier record stale even when this page renders successfully.
validation-cpu.json
| Backend | Recorded status | Assertions passed | Wall seconds | Source identity |
|---|---|---|---|---|
| cpu | PASS | 99198 | 69.24915283400333 | STALE: source differs |
Download the original record. This is recorded evidence, not a fresh measurement by the documentation build.
Execution log: validation-cpu.log
Recorded metadata and measurements
{
"schema_version": 1,
"backend": "cpu",
"status": "PASS",
"assertions_passed": 99198,
"wall_seconds": 69.24915283400333,
"returncode": 0,
"created_utc": "2026-10-20T17:14:01.210799+00:00",
"machine": "PHY-FD69NP6NG2",
"platform": "macOS-27.0.1-arm64-arm-64bit",
"threads": 4,
"source_fingerprint_sha256": "d3871a0a7f907d9dd7c4f23b59924e0413acd65b3dd744b751f148ec81441f73",
"command": [
"/Users/tabel/.julia/juliaup/julia-1.12.6+0.aarch64.apple.darwin14/bin/julia",
"--project=/Users/tabel/Projects/pmf-glm",
"--threads=4",
"/Users/tabel/Projects/pmf-glm/test/runtests.jl"
],
"log": "validation-cpu.log",
"scope": "current standalone extraction; automated numerical/operator tests"
}
validation-metal.json
| Backend | Recorded status | Assertions passed | Wall seconds | Source identity |
|---|---|---|---|---|
| metal | PASS | 243 | 37.346034458023496 | STALE: source differs |
Download the original record. This is recorded evidence, not a fresh measurement by the documentation build.
Execution log: validation-metal.log
Recorded metadata and measurements
{
"schema_version": 1,
"backend": "metal",
"status": "PASS",
"assertions_passed": 243,
"wall_seconds": 37.346034458023496,
"returncode": 0,
"created_utc": "2026-10-20T17:13:30.039710+00:00",
"machine": "PHY-FD69NP6NG2",
"platform": "macOS-27.0.1-arm64-arm-64bit",
"threads": 4,
"source_fingerprint_sha256": "d3871a0a7f907d9dd7c4f23b59924e0413acd65b3dd744b751f148ec81441f73",
"command": [
"/Users/tabel/.julia/juliaup/julia-1.12.6+0.aarch64.apple.darwin14/bin/julia",
"--project=/Users/tabel/Projects/pmf-glm/environments/metal",
"--threads=4",
"/Users/tabel/Projects/pmf-glm/scripts/test_gpu.jl",
"--backend=metal"
],
"log": "validation-metal.log",
"scope": "current standalone extraction; automated numerical/operator tests"
}
validation.json
| Backend | Recorded status | Assertions passed | Wall seconds | Source identity |
|---|---|---|---|---|
| cpu | PASS | 99198 | 69.24915283400333 | STALE: source differs |
| metal | PASS | 243 | 37.346034458023496 | STALE: source differs |
| cuda | not yet measured (NOT_RUN) | not reported | not yet measured | not recorded |
Download the original record. This is recorded evidence, not a fresh measurement by the documentation build.
Execution log: validation-cpu.log · Execution log: validation-metal.log
Recorded metadata and measurements
{
"schema_version": 1,
"backends": {
"cpu": {
"schema_version": 1,
"backend": "cpu",
"status": "PASS",
"assertions_passed": 99198,
"wall_seconds": 69.24915283400333,
"returncode": 0,
"created_utc": "2026-10-20T17:14:01.210799+00:00",
"machine": "PHY-FD69NP6NG2",
"platform": "macOS-27.0.1-arm64-arm-64bit",
"threads": 4,
"source_fingerprint_sha256": "d3871a0a7f907d9dd7c4f23b59924e0413acd65b3dd744b751f148ec81441f73",
"command": [
"/Users/tabel/.julia/juliaup/julia-1.12.6+0.aarch64.apple.darwin14/bin/julia",
"--project=/Users/tabel/Projects/pmf-glm",
"--threads=4",
"/Users/tabel/Projects/pmf-glm/test/runtests.jl"
],
"log": "validation-cpu.log",
"scope": "current standalone extraction; automated numerical/operator tests"
},
"metal": {
"schema_version": 1,
"backend": "metal",
"status": "PASS",
"assertions_passed": 243,
"wall_seconds": 37.346034458023496,
"returncode": 0,
"created_utc": "2026-10-20T17:13:30.039710+00:00",
"machine": "PHY-FD69NP6NG2",
"platform": "macOS-27.0.1-arm64-arm-64bit",
"threads": 4,
"source_fingerprint_sha256": "d3871a0a7f907d9dd7c4f23b59924e0413acd65b3dd744b751f148ec81441f73",
"command": [
"/Users/tabel/.julia/juliaup/julia-1.12.6+0.aarch64.apple.darwin14/bin/julia",
"--project=/Users/tabel/Projects/pmf-glm/environments/metal",
"--threads=4",
"/Users/tabel/Projects/pmf-glm/scripts/test_gpu.jl",
"--backend=metal"
],
"log": "validation-metal.log",
"scope": "current standalone extraction; automated numerical/operator tests"
},
"cuda": {
"status": "NOT_RUN",
"scope": "hardware validation has not been performed in this extraction"
}
}
}
10.2 Production and restart checks
The end-to-end gate executes a full-physics cosmological smoke case and compares every checkpointed field after four continuous steps against two steps followed by a fresh-process restart to step four. This tests real launcher and checkpoint behavior in addition to the operator suites. These short tests establish installation and restart consistency; the science convergence matrix remains necessary for a physical result.
julia --project=. -t 4 scripts/validate_production.jl --backend=cpu
julia --project=environments/metal -t 4 scripts/validate_production.jl --backend=metalproduction-cpu.toml
Download the original record. This is recorded evidence, not a fresh measurement by the documentation build.
Recorded metadata and measurements
{
"backend": "cpu",
"cosmology_status": "COMPLETED",
"created_utc": "2026-10-21T19:38:50.641",
"diagnostic_cadence_max_scaled_error": 0.0,
"julia_threads": 4,
"julia_version": "1.12.6",
"machine": "PHY-FD69NP6NG2",
"planar_max_scaled_error": 0.0,
"restart_error_tolerance": 5e-06,
"restart_fields": 22,
"restart_max_scaled_error": 1.1073875150202639e-10,
"run_directory": "results/production-cpu-tH3U4i",
"schema_version": 1,
"scope": "full-physics endpoint smoke plus four-cycle planar streaming full-MHD/photon/H-He-chemistry/gravity restart and diagnostic-cadence parity",
"source_fingerprint_sha256": "402403196ca5f6d958ebba5dafd0ecf4145cecbe7e378f765dd6bb618de2259c",
"status": "PASS",
"streaming_initial_relative_tolerance": 2e-06,
"wall_seconds": 227.07926988601685,
"restart_field_errors": {
"chem_H2I": 0.0,
"chem_HII": 0.0,
"chem_HeII": 0.0,
"chem_eint": 0.0,
"energy_residual": 0.0,
"gas_delta": 1.1073875150202639e-10,
"mhd_U_1": 0.0,
"mhd_U_2": 0.0,
"mhd_U_3": 0.0,
"mhd_U_4": 0.0,
"mhd_U_5": 0.0,
"mhd_U_6": 0.0,
"mhd_U_7": 0.0,
"mhd_U_8": 0.0,
"mhd_U_9": 0.0,
"particle_px": 4.733165431326071e-30,
"particle_py": 4.733165431326071e-30,
"particle_pz": 2.9281213511468896e-30,
"particle_vx": 2.5849394142282115e-26,
"particle_vy": 1.9387045606711586e-26,
"particle_vz": 1.3153304185167845e-26,
"photon_hat_batch": 6.145375463387856e-17
}
}
production-metal.toml
Download the original record. This is recorded evidence, not a fresh measurement by the documentation build.
Recorded metadata and measurements
{
"backend": "metal",
"cosmology_status": "COMPLETED",
"created_utc": "2026-10-21T19:54:51.356",
"diagnostic_cadence_max_scaled_error": 0.0,
"julia_threads": 4,
"julia_version": "1.12.6",
"machine": "PHY-FD69NP6NG2",
"planar_max_scaled_error": 0.0,
"restart_error_tolerance": 5e-06,
"restart_fields": 22,
"restart_max_scaled_error": 2.7755575615628914e-17,
"run_directory": "results/production-metal-hCD4xG",
"schema_version": 1,
"scope": "full-physics endpoint smoke plus four-cycle planar streaming full-MHD/photon/H-He-chemistry/gravity restart and diagnostic-cadence parity",
"source_fingerprint_sha256": "402403196ca5f6d958ebba5dafd0ecf4145cecbe7e378f765dd6bb618de2259c",
"status": "PASS",
"streaming_initial_relative_tolerance": 2e-06,
"wall_seconds": 340.5331790447235,
"restart_field_errors": {
"chem_H2I": 0.0,
"chem_HII": 0.0,
"chem_HeII": 0.0,
"chem_eint": 0.0,
"energy_residual": 0.0,
"gas_delta": 2.7755575615628914e-17,
"mhd_U_1": 0.0,
"mhd_U_2": 0.0,
"mhd_U_3": 0.0,
"mhd_U_4": 0.0,
"mhd_U_5": 0.0,
"mhd_U_6": 0.0,
"mhd_U_7": 0.0,
"mhd_U_8": 0.0,
"mhd_U_9": 0.0,
"particle_px": 6.310887241768095e-30,
"particle_py": 4.733165431326071e-30,
"particle_pz": 1.3320250593687446e-30,
"particle_vx": 1.2924697071141057e-26,
"particle_vy": 1.6155871338926322e-26,
"particle_vz": 7.519110147915694e-27,
"photon_hat_batch": 8.271743017139338e-25
}
}
10.3 Reproduce or update
julia --project=. -t auto scripts/validate.jl --backend=cpu
# On a Metal host, after Metal bootstrap:
julia --project=environments/metal -t auto scripts/validate.jl --backend=metal
# On an NVIDIA host, after CUDA bootstrap:
julia --project=environments/cuda -t auto scripts/validate.jl --backend=cuda
quarto render docs --fail-if-warnings
python3 docs/scripts/check_links.py docs/_sitevalidation.json aggregates the backend records; validation-{backend}.json and validation-{backend}.log retain the individual run evidence. Optional TOML records are also displayed. Failed runs and skipped devices must not be relabeled as a successful extraction validation.
The documentation build accepts evidence without requiring a GPU, Julia, external campaign directory, or numerical rerun. A green Pages build establishes documentation/link integrity, not science certification.