# Phase 154 evidence

## 154.13 release validation

| Requirement | Direct evidence                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        |
| ----------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| 154.13.7    | `neith-raster-validation::material` links the Phase 154 raster material library to the Phase 138 `neith-pathtracer` crate. Its versioned interchange document carries exact catalogue metadata and the path tracer's native `StandardSurface`; conversion preserves base colour, metalness, roughness, Disney specular via its equivalent dielectric IOR, anisotropy, coat, coat roughness, sheen, emission and thin-wall state. Tests exercise raster export, JSON process handoff, offline decode/re-encode, raster import, metadata and bounded numeric parity, actual offline BSDF evaluation, schema rejection, range rejection and non-finite-payload rejection. |

Verification on 2026-07-15:

- `cargo test --manifest-path libs/neith/raster-validation/Cargo.toml`: 53 unit
  tests and doc tests pass.
- `cargo test --manifest-path libs/neith/raster-material/Cargo.toml`: 145 unit
  tests and doc tests pass.
- `cargo test --manifest-path libs/neith/weaver/Cargo.toml -p neith-pathtracer`:
  52 unit tests and doc tests pass.
- Warning-denied Clippy passes for all targets of all three affected crates.
- The validation workspace format check and targeted raster-material source
  format check pass.
- `cargo build --release --manifest-path libs/neith/raster-validation/Cargo.toml`
  passes.

## Remaining non-local gates audit — 2026-07-15

All locally executable Phase 154 rows are complete. Three rows remain
deliberately unchecked because their acceptance conditions require physical
reference hardware unavailable in this checkout. This Linux x86-64 host exposes
only a Red Hat Virtio GPU, `nvidia-smi` reports no device, and
`/sys/class/power_supply` exposes no battery or power meter.

| Open requirement | Non-local acceptance boundary                                                                                                                                                                                   |
| ---------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| 154.13.4         | The requirement is an observed 60 FPS at 1080p on a defined mid-tier physical GPU with the 8M-triangle acceptance scene. A virtual display adapter cannot establish that result.                                |
| 154.13.5         | The requirement is an observed 30 FPS at 4K on a defined high-tier physical GPU with the 30M-triangle acceptance scene. No qualifying GPU is attached.                                                          |
| 154.13.10        | Battery-life impact requires baseline and raster workload measurements on representative laptops under controlled battery/power-meter conditions. This host has neither a battery nor a power telemetry device. |

The tested `perf` and `power` validators remain the local ingestion and scoring
halves of those campaigns. They reject non-finite/zero timing evidence, missing
warm-up, mismatched resolution or triangle counts, the wrong GPU class, empty or
unpaired power samples, invalid baselines, AC-powered captures and unnamed
devices. This audit is a scope boundary, not a completion waiver: each hardware
row stays open until measurements from the named target class are recorded.
