Skip to content

Output Matrix 0.1.1-experimental: always-on LEV label and smaller core-0 DSP, on the USB base (from #687) - #705

Merged
repeat98 merged 22 commits into
mainfrom
merge/output-matrix-0.1.1-687
Oct 19, 2026
Merged

Output Matrix 0.1.1-experimental: always-on LEV label and smaller core-0 DSP, on the USB base (from #687)#705
repeat98 merged 22 commits into
mainfrom
merge/output-matrix-0.1.1-687

Conversation

@repeat98

@repeat98 repeat98 commented Oct 19, 2026

Copy link
Copy Markdown
Owner

This is @npp1993's #687 with current main merged in, including E-Verb 0.1.2 (#703). Their commits are unchanged; I couldn't push to their fork. The changes themselves are theirs: the LEV box always shows the track's destination, the core-0 DSP is smaller (657 + 14 words, same output), and the beta base and controller package are rebuilt for it. Full description and evidence are in #687.

What the merge adds

  • Compiled packages. E-Verb 0.1.2 and this PR changed different rows of the same package files (plus the moduleVersions map).
    • Each merged file is exactly main's file with this PR's own changes applied, checked path by path, so it is what the compiler writes for the merged tree.
    • module-build.json and the Analog BD proofs carry the merged tree's fingerprint (moduleSourceFingerprint) and file hashes.
    • The Tape Echo and Euclid standalone recipes hash the four merged asset files.
    • The same method went through E-Verb 0.1.2: the ring stops clear of Sidechain Compressor's key window (from #697) #703's publish today: CI rebuilt the packages from source, they matched, and the site deployed.
  • CHANGELOG.md: keeps this PR's base entry.
  • Corrected release notes (src/community/module-changelogs/output-matrix.json):
    • The DSP code has run on hardware over USB, inside the base; on a card it has run only in the emulator.
    • This release publishes base 3e3e3de1 and package e2df2fbf, so a beta unit needs the base update before its next sync.
    • Untested on the unit for 0.1.1: the card image, Part/project reload, MASTER TRACK and STUDIO.

What the owner accepts by merging

  • 0.1.1 becomes the public card download, although only the USB/RAM-booted run happened on hardware.
  • A new beta base identity, so every beta unit updates its base.
  • Untested Part/project reload.
  • After a power cycle the LEV box shows stock's split bars until CUE is pressed (display only).

Checked: npm run check -- --base origin/main (2,367 tests) and npm run pr:ready (Ready) on the merged tree.

Supersedes #687.

Generated by Claude Code

npp1993 and others added 22 commits October 16, 2026 08:31
@npp1993 @claude
Fix: native builder places a floating cave in the overflow run before…
842d22f 
… refusing

When a floating ColdFire cave (Recorder Loop Fix's) did not fit the menu
cave window, build_bus.py asserted "past the stock zero run" before its own
overflow fallback ran. The browser composer (requestedCaves) puts the cave
in the overflow run at 0x400d24d0 and goes on to the real limit, so the two
refused the same selection for different reasons and module:verify counted
a mismatch (two EVERB selections on main, LOFI2's 13-module ones). Native
now places the cave by the browser's rule first (floating_cave_at): the
window at the next 128 bytes if it fits whole, else the overflow at the
next 4, else refuse.

- sdk/infrastructure-verification/floating-cave-overflow-builder.json:
  128 native selections built in the pinned image with main's and this
  build_bus.py (66 ledger-refused ones left out). All 100 images are
  identical, 10 refusals change only their reason, nothing refused now
  builds. Chained into the infrastructure tests, with unit tests for the
  placement rule and its parity with the browser's overflow constants.
- Analog BD composition proofs re-exported natively: all 212 identical,
  only builderSources and moduleSourceTreeSha256 change;
  verify-analog-bd-native.mjs passes (162 comparisons, 50 refusals, 6
  round trips).
- Module packages rebuilt in the isolated image and imported
  (--development): source-commit stamps only. The Tape Echo and Euclid
  standalone recipes' hashes of the four rebuilt shared asset rows are
  refreshed (provenance only, as in #634).

With LOFI2's branch, module:verify lofi2 reports 0 mismatches of 130.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix: label the LEV box with the track's destination
0b8f9ca 
In MATRIX the LEV box now always shows where the track plays (MN, PHN, ...)
instead of LEV, not only while CUE is held. A detour at 0x4004dcfa replaces
the plain LEV label push; the level value while turning, XLV, FUNC's MAIN
view, NORMAL and STUDIO stay stock. Gate 117/117; LCD captures in TESTING.md.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix gate: compare runs over the same samples from transport…
a9d7171 
… start

Run.rms took each capture's own last 2,000 samples, so two runs that
captured a different number of frames after transport start were compared
one sample apart (DIR AB 127 failed at +0.057% in combined image OMXATS27;
lined up from transport start the two match to 26 LSB). Checks that compare
runs (MATRIX MAIN vs stock, the level law, mono vs stereo, DIR) now trim
every capture to the shortest before taking the window. Tolerances unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix: core-0 code 727+42 -> 657+14 words, same output
ac94037 
One bus-ramp routine through Y pointers (the first frame runs it twice),
pointer-based per-sample ramps and end copy, OR-ed input gain tests, direct
address-register moves, a parallel move per bus level, and the destination
table packed three list numbers to a word. Sample-identical to the previous
build in 26 ot_emu scenarios; worst-case mixdown 4,240 -> 4,194 executed
instructions per frame; gate 117/117.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental on the USB base: controller re-port,…
d0f2866 
… re-pins, base and packages rebuilt

The LEV box label (new detour lev_box_lev at 0x4004dcfa, shared
dest_label) and the smaller core-0 DSP (657 + 14 words instead of
727 + 42) change sources the USB base pins.

Controller (stock_hook_controllers.py): the label moves the CUE-held
path's stock-equal span from offset 58 to 38 of lev_box_cue, and
lev_box_lev repeats the same 14 bytes (stock 0x4003ffd8, same SHA-256), so
STOCK_COPIES gets both entries (six copies). controller_sites picks up the
new detour: 20 sites. The package's stock-free scan passes (10 shared
non-zero bytes, fails at 12).

Re-pinned manifest.py, matrix_mix.asm and matrix.s in resident_dsp.py,
stock_hook_controllers.py and test_analog_bd_part_fence.py. The resident
Matrix unit is 675 words (was 773).

Base rebuilt from main a59677e with the beta flags (--write; main's
committed base reproduced byte for byte here first): package a9b5fe45,
identity 3e3e3de1, browser .bin = build_core.py .bin ca8b457f. The lease
entries did not move, so Analog BD's and Sidechain's package bytes are
unchanged; packages relinked, the other base-bound packages only change
requiredBase. New trust rows for Output Matrix's package e2df2fbf.

Version 0.1.1-experimental (catalog, manifest), changelog entries. The
qualification record stays 0.1.0's until the USB hardware report exists.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: source-built packages (isolated bui…
44738ea 
…ld, development import)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: MKII hardware record over USB
6646ff8 
The author's panel checks on a RAM-booted beta base built from this
source (MAIN.raw 5ca0afbf, package e2df2fbf): MATRIX on and off, the
always-on LEV label, routing to MAIN, CUE, PHONES, a mono jack and OFF,
and MIX as the PHONES level, all passed. Qualification moves to 0.1.1
with the current DSP figures (275 instructions/sample; 657 + 14 P words)
and evidence/usb-hardware.md.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: OT UI captures from the 0.1.1 image
ebd59d7 
The capture plan rerun on a static 0.1.1 image (mainos a9d5e3b8): the four
screenshots are pixel-identical to 0.1.0's (the always-on label equals the
CUE-held one); records and captions updated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: regenerate module documents
d88e529 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: test pins, recipe provenance and pe…
0cf1845 
…rformance record

- module-source and dsp-package tests: Output Matrix's version and its
  671-word core-0 package.
- Tape Echo and Euclid standalone recipes: the four rebuilt shared asset
  rows' hashes (provenance only, as in #634 and #645).
- evidence/performance.json for 0.1.1: 275 executed instructions per
  sample; the 32 s stress run was made on 0.1.0 and is not repeated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Analog BD composition proofs: source-tree stamp after Output Matrix 0…
acf4de5 
….1.1

Re-exported natively (export-composition-proofs.py --suite analog-bd
--static-stock --vendored-sdk): all 212 proofs identical; only
moduleSourceTreeSha256 changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993
Merge remote-tracking branch 'fork/fix/builder-overflow-parity' into …
03e6a68 
…claude/output-matrix-lev-label-usb

# Conflicts:
#	CHANGELOG.md
#	sdk/machines/octatrack/module-adapters/standalone-recipes.json
#	src/engine/assets/analog-bd-composition-proofs.json
#	src/engine/assets/bootstrap-package.json
#	src/engine/assets/coldfire-packages.json
#	src/engine/assets/descriptor-recipes.json
#	src/engine/assets/dsp-packages.json
#	src/engine/assets/menu-recipes.json
#	src/engine/assets/module-build.json
#	src/engine/assets/platform-writes.json
#	src/engine/assets/requested-packages.json
#	src/engine/assets/resident-dsp.json
#	src/engine/assets/rom-packages.json
#	src/engine/assets/usb-audio-packages.json
#	src/engine/assets/utility-packages.json
@npp1993 @claude
Output Matrix 0.1.1-experimental on #682: source-built packages and r…
706ecb9 
…ecipe provenance

Packages rebuilt in the isolated image on the merged tree and imported
(--development); Tape Echo and Euclid standalone recipes name the four
rebuilt shared asset rows' hashes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Analog BD composition proofs: source-tree stamp on the merged tree
6d4db60 
Re-exported natively: all 212 proofs and builderSources identical; only
moduleSourceTreeSha256 changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: native comparison record
d1c3574 
module:verify output-matrix: 52 native module-owned images match, 74
matching refusals, 0 mismatches of 126 selections (with #682's builder
fix; without it the two 11-module selections refused for different
reasons).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: source identity after the performan…
2c49b75 
…ce record

Only evidence/performance.json changed since the tested commit; manifest.py,
matrix.s, matrix_mix.asm and verify.py are as tested.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: regenerate module documents
d5dd47d 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Keep main's chooser-metadata order
d3f3d64 
module:verify rewrote the file with POLY8's entry in another position (no
content change); main's order keeps the committed base, the one tested on
the MKII, unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: power cycle on the flashed base
ca39671 
The author flashed the tested base: a power cycle kept MATRIX, the
routing and all seven installed modules. After it the LEV box showed
stock's split bars until CUE redrew the page (display only), recorded as
a limitation.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@npp1993 @claude
Output Matrix 0.1.1-experimental: regenerate module documents
6247f77 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@repeat98 @claude
Output Matrix 0.1.1 notes: what has run on hardware, and that this pu…
a47cf2f 
…blishes the base

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SakBf5zGyZY61TAzvKCKHC
@repeat98 @claude
Merge main into Output Matrix 0.1.1 (#687)
6e5fb51 
E-Verb 0.1.2 (#703) and this PR changed different rows of the same
compiled package files, plus the moduleVersions map. Each merged file is
exactly main's file with this PR's changes applied (checked path by path),
so it is what the compiler writes for the merged tree; the sourceCommit
stamps are this PR's. module-build.json and the Analog BD proofs carry the
merged tree's fingerprint (moduleSourceFingerprint) and file hashes, and
the Tape Echo and Euclid standalone recipes hash the four merged asset
files. CI's publish rebuilds the packages from source and refuses them
unless they match. CHANGELOG.md keeps this PR's entry.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SakBf5zGyZY61TAzvKCKHC
@repeat98
repeat98 merged commit fd77b90 into main Oct 19, 2026
8 checks passed
@repeat98 repeat98 mentioned this pull request Oct 19, 2026
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@repeat98 @npp1993