Skip to content

SonKu P7.2: realtime XYZ native JUCE ACK + backwards-compatible experimental v3 state - #685

Closed
masarray wants to merge 4 commits into
feat/sonku-p71-xyz-direct-and-fir-20261114from
feat/sonku-p72-native-xyz-ack-state-20261114
Closed

masarray wants to merge 4 commits into
feat/sonku-p71-xyz-direct-and-fir-20261114from
feat/sonku-p72-native-xyz-ack-state-20261114

Conversation

@masarray

Copy link
Copy Markdown
Owner

Concrete milestone stacked on #681 → #679 → #678 → #674

Brings P7.1's measured KEMAR direct 3D source to the existing real JUCE SPSC→audio callback→completion ACK contract, rather than creating a second FX graph.

Implementation: One extra Settings::directElevationDegrees field exists only when ARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT=ON. Existing asynchronous native queue consumes joint ROOM/distance/elevation edits with validity and physically bounded source height, applies changes without reprepare or clearing audio, and reports one authoritative native completion. Invalid edit fails closed without shifting orbit. Real JUCE native FX persists research-only SONK v3 (50 bytes) and restores v1 (30) / v2 (46) exactly. Public default still LIGHT + v2 46 bytes, with P7.2 compiled OUT. No public GUI/.smix schema change, installer, VST3 catalog ID change, WebAudio path, Callback worker or Harmonice rename.

Evidence: Scoped Linux 44.1/48k moving-source PCM/phase, 64/257 chunk parity, invalid-state rejection, device reprepare/downgrade; scoped real Windows MSVC/JUCE 8.0.15 v3 ACK / v1-v2 restore and baseline v2 regression. Tests build only three small source/JUCE targets, not the entire application or installer.

Hard veto: Public release, Win ASIO 64/128/256 with concurrent WIRE/VST3, audible externalization listening against dearVR, public frontend UX, browser JSON→host state migration and installer license remain blocked in Issue #683. Keep draft, do not merge automatically. Design: docs/SONKU_P72_NATIVE_XYZ_ACK_V3_RESEARCH_20261114.md.

@cloudflare-workers-and-pages

cloudflare-workers-and-pages Bot commented Nov 14, 2026 •

Copy link
Copy Markdown

Deploying arsonmix with  Cloudflare Pages  Cloudflare Pages

Latest commit: 08596bf
Status: ✅  Deploy successful!
Preview URL: https://f058886f.arsonmix.pages.dev
Branch Preview URL: https://feat-sonku-p73-xyz-gui-bridg.arsonmix.pages.dev

View logs

Copy link
Copy Markdown
Owner Author

P7.2 native experimental engineering milestone — CLOSED 14 Nov 2026

Exact GitHub HEAD 08596bf44ab09da5ac07c1ee854a0359c36a14ba, PR #685 is intentionally DRAFT, based on P7.1 PR #681.

Implemented (actual source code, not only design)

  • Native JUCE SonKuSpatialProcessor prepares existing one-authority SixSurfaceKemarResearch only with ARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT=ON (CMake default OFF). This implies measured MIT KEMAR elevation data; public source builds keep Lightweight, old state v2, and no extra 3D HRIR tables.
  • Extended the existing Settings POD in the opt-in native build with directElevationDegrees; the existing native SPSC queue consumes ROOM/distance/elevation as one edit, applied by the AUDIO callback, acknowledged through the same completion/readback mechanism. No second render engine, new UI state authority, background worker, additional realtime convolution, callback allocation, or graph reset to change 3D position.
  • 3D pre-admission rejects impossible room/source height, bad distance, NaN, and out-of-range elevation. Failed edits do NOT rotate the already-running orbit and cannot mutate accepted GUI/snapshot state. Same native PCM instance slews KEMAR direct and six reflected image paths.
  • Experimental native FX SONK v3 = 50 bytes, extending v2 by exactly one 32-bit elevation, read/write and restart. v2 46-byte and v1 30-byte restore still accepted in the opt-in JUCE adapter. Public default retains 46-byte SONK v2 and rejects unsupported v3; no public .smix JSON state/schema or GUI controls have been added.

Exact-head automated evidence, ALL PASS

  • GitHub Actions #43560297883 SUCCESS at head 08596bf44ab09da5ac07c1ee854a0359c36a14ba, two focused jobs:
    • Linux C++20 strict -Werror PASS: exact source P6 and P7.1 negative controls + 44.1/48k P7.2 native realtime 0→+25→−25° edit, 64/257 frame partition parity, stereo anti-phase continuity/finite PCM, rejection/no orbit movement, sample rate/device restart and P6 downgrade.
    • Real Windows MSVC + JUCE 8.0.15 Release CTest 3/3 PASS (0 failures): NativeSonKuSessionV1V2AndAudioAppliedAck, NativeSonKuP72AtomicXyzLiveEditSource, NativeSonKuP72JuceV3XyzLiveAckV1V2Restore. Targeted 3 source/JUCE test targets, no full installer, no physical ASIO. This is genuine JUCE PCM→completion ACK and historical v1/v2/v3 state test, not a mocked JavaScript UI.
  • Ancestor focused P7, P7.1 MSVC Windows and P6 Windows JUCE workflows also PASS at the PR head. No unrelated Harmonice DSP implementation was changed.

Remains OPEN in production qualification Issue #683

  1. Approved GUI/bridge and truly published .smix v3 migration with accepted-state UX and preset recovery.
  2. Actual Windows ASIO 64/128/256 + WIRE / VST3 on 1/2/4 inserts with p95/p99, restart, XRUN and no clicks. Source-level MSVC tests do NOT establish hardware realtime.
  3. Loudness-matched blinded headphone comparisons with dearVR and real music; nonpersonalized KEMAR externalization and naturalness NOT proven.
  4. Production hardware admission policy, license/third-party notices, and conflict-safe reconciliation of stacked draft PRs against fresh main, without touching concurrent GUI/Player/Guitar FX changes.

Decision: P7.2 source integration and focused tests complete; this development thread CAN CLOSE. Keep this and all ancestor SonKu PRs DRAFT / unmerged and leave public 3D XYZ OFF until Issue #683 has real evidence.

Copy link
Copy Markdown
Owner Author

MILESTONE A — archived historical P6/P7 experimental SonKu stage (CLOSED UNMERGED). The original measured-KEMAR/XYZ study remains traceable and is not authorized as a product change. Listening-approved SonKu P9.9 was integrated to main through selective PR #733; older geometry research was never to be bulk-merged. Source branch feat/sonku-p72-native-xyz-ack-state-20261114 and exact SHA 08596bf44ab09da5ac07c1ee854a0359c36a14ba; original PR target feat/sonku-p71-xyz-direct-and-fir-20261114. Registry: docs/governance/MILESTONE_A_PR_BRANCH_RECONCILIATION_20261117.md on chore/milestone-a-pr-history-20261117. This closure does not delete source, alter checkpoint refs, merge code, retune audio, or change release. Reopen only for a separately approved verified feature.

@masarray masarray closed this Nov 17, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant