Files
prometeu-runtime/crates/tools/pbxgen-stress/music/README.md
T
bquarkz c2bf8d055c feat(stress): add a deterministic particle demo with sequenced music
Move Storm Eagle playback and source assets to Stress Console and keep Dummy Boy silent. Exercise VM math, seeded Random and persistent typed particles with a projected tunnel, bounded 512-sprite heavy workload, reproducible replays and measured costs.
2026-11-20 22:15:47 +00:00

57 lines
3.0 KiB
Markdown

# Stress Console music fixture
This fixture uses the Storm Eagle transcription supplied by the owner, sourced
from [Hooktheory](https://www.hooktheory.com/theorytab/view/setsuo-yamamoto/mega-man-x---storm-eagle).
`transcription.json` retains the musical fields: 90 melody notes, 20 chords,
relative degrees/accidentals, inversions and exact 480-PPQ source ticks. The
arranger multiplies ticks by two for native PPQ960, preserving the triplets and
the 72-beat loop. The nearest integer tempo to 168 BPM is 357143 us/quarter.
The eight source WAVs are an unchanged PCM subset extracted from the supplied
`MegaManX-SF-3.0.sf2`, whose INFO author is CliffyJiffy. Original bank fingerprint,
root keys, rates, pitch corrections and raw instrument generators are retained
in `soundfont-source.json`. The complete extracted banks remain at the sibling
port's `games/ports/snes/megaman-x/audio/`; playback and regeneration of this
selected fixture do not depend on that directory.
`tools/arrange_music.py` creates the offline authoring manifest `song.json` and
`instrument-mapping.json`. The latter records source-to-target pitch, loop,
gain, ADSR, pan and modulation decisions, including omitted SF2 hold/modulation
envelopes. Melodic negative SF2 sustain attenuation is clamped conceptually to
full sustain; target gain/articulation are explicit musical mix choices.
Drum Kit root overrides are applied. Source sample rates are preserved; root
and pitch correction are applied by native instruments. No full SF2 conversion,
chorus or SNES DSP emulation is claimed.
To rebuild the MusicBank from the existing WAVs/JSON, from the runtime root:
```sh
python3 crates/tools/pbxgen-stress/tools/regenerate_music.py
cargo run -p pbxgen-stress
```
This requires Python 3, Java 21+ and the sibling Studio checkout. A temporary
Gradle init task resolves the existing packer's runtime classpath. `PackMusic.java`
calls `AudioBankPacker.pack` twice, checks identical output, and writes the bank
and its size metrics. It contains no codec or MusicBank serializer. Studio source
and build scripts are not changed. Use `--studio PATH` for another checkout.
`storm-eagle.mbnk` and the usual `assets.pa` envelope are generated assets.
`bank-metrics.json` distinguishes cartridge payload, PCM bytes and canonical
decoded size; those values do not include all allocator/host overhead.
Run conformance and the optional WAV export with:
```sh
cargo test -p pbxgen-stress
PROMETEU_STRESS_WAV=/tmp/stress-console-storm-eagle.wav \
cargo test -p pbxgen-stress --release --test music_cartridge \
full_sequence_loops_and_frame_controls_preserve_audio -- --nocapture
```
The WAV is rendered by the real cartridge/headless/native mixer path at stereo
48 kHz and contains more than two loops. The test also compares playback with
and without bounded, representative graphics frames across both loops, checks
pause/resume and reports signal metrics and observed voices. It does not measure
sustained frame rate. Listening is still required to judge the mix.