You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 7b1cabe
Browse filesBrowse the repository at this point in the historyBrowse files
Merge M3 ArSonComp live performance baseline (Windows verified)
Combine compact premium header/toggles and optimized compressor engine with immediate dry host bypass, Windows VST3/Standalone smoke and reproducible stress testing. GitHub Actions 41209432109: Windows and Linux PASS; CRC and executable SHA256 verified; four Windows screenshots inspected. Hardware XRUN and sonic parity remain physical test gates.
Copy file name to clipboardExpand all lines: AGENTS.md
+5Lines changed: 5 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -9,3 +9,8 @@
9
9
- Use RAII and bounded containers. Avoid fast-math, speculative approximation or SIMD unless profiling and numerical tests support it.
10
10
- Run meaningful DSP/host tests, inspect GUI renders at minimum/default size, and document unverified hardware behavior. Never claim proprietary Pro-C parity without matched-input audio comparisons.
11
11
- Group meaningful changes before CI. Do not dispatch repetitive builds or commit generated binaries/build directories.
12
+
13
+
## M3 integration release gate
14
+
- Always reconcile the GUI development branch and M2 realtime-performance branch before producing a Windows test binary. An artifact from one branch alone is not a combined release.
15
+
- Host bypass must pass the first block bit-exactly and must not latch the parameter bypass state. User bypass remains a smoothed, separately attached parameter.
16
+
- Build M3 in one PR with both Windows JUCE editor smoke and Linux sanitizer, inspect four actual Windows screenshots and benchmark/stress files before declaring ready-to-test. Do not call CI throughput a hardware xrun proof.
# M3 — Integrated Windows VST3/Standalone hardware test gate
2
+
3
+
This milestone integrates the finished brushed-metal compact header/toggles with the M2 sparse-ramp real-time compressor optimizations, and makes **host bypass bit-transparent immediately**. The normal *user* bypass toggle remains a smoothed APVTS control; zero algorithmic samples are added.
4
+
5
+
## Scope and quality status
6
+
7
+
The plugin ships three independent control models (Clean, Feedback, Smooth), dual detector modes, internal/external sidechain, monitoring, wet/dry, stereo link, and program-dependent release. These are proprietary ArSon algorithms, not FabFilter emulations. [FabFilter Pro-C 3](https://www.fabfilter.com/help/pro-c/using/overview) provides an industry-level comparison target; sonic parity is **not** verified. For live use, keep any future lookahead/oversampling separate and opt-in because they can add latency.
8
+
9
+
## Install & validate — Windows 64-bit
10
+
11
+
1. Download the latest **ArSonComp-Windows-x64-ReadyToTest** ZIP from the matching M3 pull-request GitHub Actions run. Extract it. Check `BUILD-INFO.txt` commit SHA and `SHA256SUMS.txt`; do not use an old ZIP from M2.
12
+
2. For Standalone, open `Standalone/ArSonComp.exe`. For a DAW/VST3 host, copy the **entire**`Plugin/ArSonComp.vst3` directory into the host's VST3 scan folder (normally `C:\Program Files\Common Files\VST3\`); do not copy only the .vst3 inner DLL.
13
+
3. Inspect `Screenshots/` (full, minimum, MINI, sidechain) and the real GUI: embossed brand readable, premium compact header, preset dropdown capped at 176px, switches correct and never overlapped, no tooltip bubble.
14
+
4. Start at 48 kHz / 128 sample buffer, then 64, then 32 if stable. Record host, interface/driver version, CPU load, DPC spikes, **actual** host XRUN/dropout counter, and reported plugin latency. Repeat with 1, 8, 16 active plugin instances.
15
+
5. Listen on vocals, drums, bass and program bus at matched output loudness; adjust threshold for similar reduction when comparing with a commercial compressor. Assess attacks, recovery, pumping, breath/sibilance, aliasing and dry/wet coherence. Never equate same settings to same tone.
16
+
6. Verify host bypass on the **first** processing block with non-unity input/makeup; it should be exactly dry. Toggle the built-in Active/Bypass control separately to assess its intentional short fade; check restart, mono/stereo and external sidechain routing and Listen SC. Test preset save/load and host project restore.
17
+
7. Run a 60-minute session; open/close GUI repeatedly, automate threshold/knee, toggle modes and log private-byte trend plus latency/XRUN evidence. Stable memory use and no XRUN on your machine must be observed, not assumed from CI.
18
+
19
+
## Build evidence
20
+
21
+
GitHub PR checks gate Linux address/undefined sanitizers and engine contract tests, Windows VST3/Standalone build plus JUCE plugin/editor smoke and screenshot capture, 48kHz engine benchmark, and 32/64/128 × 1/8/24 engine-only realtime stress statistics (mean, P99, P99.9, max). Those stress numbers include hosted CI scheduler noise, **not** a physical audio dropout test.
22
+
23
+
No worker, mutex, allocation, filesystem access, GUI rendering or telemetry consumer on the audio callback; synchronous DSP state stays audio-thread-owned.
0 commit comments