peter-swimm/nardwahl
Watch
1
0
Fork
You've already forked nardwahl
 0
Code Issues 4 Pull requests Projects Releases Packages Wiki Activity Actions

Regenerate the parity matrix and baseline status from plan v26 #15

Merged
peterswimm merged 1 commit from baseline-v26 into main 2026-10-22 14:40:45 +00:00
Conversation 0 Commits 1 Files changed 3 +115 -111
peterswimm commented 2026-10-22 04:00:25 +00:00
Owner
Copy link

Scope

This is the post-merge refresh after #10–#14, with no code changes. It regenerates three files from plan v26, which is cut from main e736978:

  • docs/G1_PARITY_MATRIX.json
  • docs/G1_PARITY_MATRIX.md
  • docs/G1_BASELINE_STATUS.json

The plan was cut with --control-observer-ready --physical-output-source-ready --lfoa-modulation-ready. All four baseline worker shards and the full probe sweep were run against it.

Testing

  • python3 -m unittest discover -s tests: 150 tests OK.
  • tools/export_public.py into an empty directory verifies clean.
  • Fixtures are unchanged by #14: every v26 fixture hash matches a plan cut from 23ba636.
  • #14 left non-AudioIn Faust byte-identical: across all 1,511 fixtures, only AudioIn's three exports differ between 23ba636 and e736978.
    • AudioIn was re-rendered on e736978, and its three probes still pass, with error near 0.
  • Module 17 (EventSeq) was re-rendered on its own and all 79 probes pass. Its first-pass target timeouts came from running the baseline workers alongside the sweep.

Probe verdicts

Verdict v24 v25 v26
pass 842 922 947
peak-only 7 4 4
level 192 194 193
pitch 298 245 240
shape-or-phase 100 74 83
silence-mismatch 14 14 5
both-silent 36 36 39
total 1,489 1,489 1,511

Parity stages

Stage v24 v25 v26
sweep-clean 17 21 24
near 22 23 23
partial 23 22 24
divergent 44 40 38
not-swept 3 3 0
blocked / unobserved 0 0 0

Stage moves from v25 to v26 are all upward, with no module dropping a stage:

Module v25 → v26
AudioIn divergent → sweep-clean
GainControl divergent → partial
CompareAB not-swept → partial
MasterOsc not-swept → sweep-clean
PolyAreaIn not-swept → sweep-clean

Pass → fail check

  • v24 → v26: 2 flips, both agreed vacuous probes (0 real). 86 probes go fail → pass.
    • 3 of the 5 agreed vacuous probes now pass outright after v25's OscA fix.
  • Scratch (#13) → v26: 0 flips; 82 probes go fail → pass.
  • v25 → v26: 4 flips. 2 are agreed vacuous probes; 2 are flagged:
    • 023-021 (Mod-Env retrig): v25 pass, rms 0.0048 → v26 shape-or-phase, rms 0.0697.
    • 020-016 (ADSR retrig): v25 pass, rms 0.006 → v26 pitch, rms 0.0929.
    • Both v25 passes were midpoint-equivalent: 023-001 was 0.0066 and 020-001 was 0.0074. A Keyboard note on retrig never fired.
    • Under #13's stimulus table the retrig now fires a ClkGen. ADSR retrig is unmodelled, and the Mod-Env retrig model still misses even with v25's ClkGen phase.

Baseline status: 109 of 109 modules completed (106 completed and 3 unobservable before). Midpoints with signal on both engines rise from 104 to 107.

Impact Statement

  • Published parity and baseline figures now reflect main after the v25 calibration, the stimulus-by-role work and AudioIn's Disting mapping.
    • 947 of 1,511 probes pass.
    • 24 modules are sweep-clean.
    • Every runtime module is now observed.
  • No model, fixture or tolerance changed.
  • Follow-ups exposed by real retrig stimulus: 023-021 (Mod-Env retrig fit) and 020-016 (ADSR retrig, unmodelled).
## Scope This is the post-merge refresh after #10–#14, with no code changes. It regenerates three files from plan v26, which is cut from main `e736978`: - `docs/G1_PARITY_MATRIX.json` - `docs/G1_PARITY_MATRIX.md` - `docs/G1_BASELINE_STATUS.json` The plan was cut with `--control-observer-ready --physical-output-source-ready --lfoa-modulation-ready`. All four baseline worker shards and the full probe sweep were run against it. ## Testing - `python3 -m unittest discover -s tests`: 150 tests OK. - `tools/export_public.py` into an empty directory verifies clean. - Fixtures are unchanged by #14: every v26 fixture hash matches a plan cut from `23ba636`. - #14 left non-AudioIn Faust byte-identical: across all 1,511 fixtures, only AudioIn's three exports differ between `23ba636` and `e736978`. - AudioIn was re-rendered on `e736978`, and its three probes still pass, with error near 0. - Module 17 (EventSeq) was re-rendered on its own and all 79 probes pass. Its first-pass target timeouts came from running the baseline workers alongside the sweep. **Probe verdicts** | Verdict | v24 | v25 | v26 | |---|---|---|---| | pass | 842 | 922 | 947 | | peak-only | 7 | 4 | 4 | | level | 192 | 194 | 193 | | pitch | 298 | 245 | 240 | | shape-or-phase | 100 | 74 | 83 | | silence-mismatch | 14 | 14 | 5 | | both-silent | 36 | 36 | 39 | | total | 1,489 | 1,489 | 1,511 | **Parity stages** | Stage | v24 | v25 | v26 | |---|---|---|---| | sweep-clean | 17 | 21 | 24 | | near | 22 | 23 | 23 | | partial | 23 | 22 | 24 | | divergent | 44 | 40 | 38 | | not-swept | 3 | 3 | 0 | | blocked / unobserved | 0 | 0 | 0 | Stage moves from v25 to v26 are all upward, with no module dropping a stage: | Module | v25 → v26 | |---|---| | AudioIn | divergent → sweep-clean | | GainControl | divergent → partial | | CompareAB | not-swept → partial | | MasterOsc | not-swept → sweep-clean | | PolyAreaIn | not-swept → sweep-clean | **Pass → fail check** - **v24 → v26:** 2 flips, both agreed vacuous probes (0 real). 86 probes go fail → pass. - 3 of the 5 agreed vacuous probes now pass outright after v25's OscA fix. - **Scratch (#13) → v26:** 0 flips; 82 probes go fail → pass. - **v25 → v26:** 4 flips. 2 are agreed vacuous probes; **2 are flagged:** - **023-021** (Mod-Env retrig): v25 pass, rms 0.0048 → v26 shape-or-phase, rms 0.0697. - **020-016** (ADSR retrig): v25 pass, rms 0.006 → v26 pitch, rms 0.0929. - Both v25 passes were midpoint-equivalent: 023-001 was 0.0066 and 020-001 was 0.0074. A Keyboard note on retrig never fired. - Under #13's stimulus table the retrig now fires a ClkGen. ADSR retrig is unmodelled, and the Mod-Env retrig model still misses even with v25's ClkGen phase. **Baseline status:** 109 of 109 modules completed (106 completed and 3 unobservable before). Midpoints with signal on both engines rise from 104 to 107. ## Impact Statement - Published parity and baseline figures now reflect main after the v25 calibration, the stimulus-by-role work and AudioIn's Disting mapping. - 947 of 1,511 probes pass. - 24 modules are sweep-clean. - Every runtime module is now observed. - No model, fixture or tolerance changed. - Follow-ups exposed by real retrig stimulus: 023-021 (Mod-Env retrig fit) and 020-016 (ADSR retrig, unmodelled).
Regenerate the parity matrix and baseline status from plan v26
All checks were successful
PR policy check / policy-gate (pull_request) Successful in 0s
Details
a9b0db885e 
Probe sweep v26: 947 of 1,511 probes pass (922 in v25); sweep-clean 24,
near 23, partial 24, divergent 38, none unswept. All 109 baseline midpoints
complete; 107 carry signal on both engines.
peterswimm deleted branch baseline-v26 2026-10-22 14:40:45 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
peter-swimm/nardwahl!15
No description provided.