Direct RGBA receive: source socket pressure
Configuration and method
These measurements use a private diagnostic fixture with source socket-pressure counters.
Noisedeck A sends 1920 × 1080 RGBA packets to Noisedeck B through Sync.
Native camera publication remains active. B receives direct RGBA through an HTTP BYOB reader, not through the camera.
The receiver retains one credit. The source byte budget is one packet: 8,294,464 bytes.
The packet includes 8,294,400 pixel bytes and a 64-byte source header.
The fixture retains the receiver, uploader and ACK pooling changes from its parent.
Timing uses synth.testPattern with pattern 0 and gridSize 4. Pixel controls use pattern 5 and gridSize 16.
The compiled source and uniforms passed the existing before-and-after checks. Both configurations use GLSL.
Retained host metadata identifies an Apple M2, eight logical CPUs, 24 GiB of memory and Darwin 23.6.0.
This metadata is not an in-run thermal or frequency measurement.
Measurement results
Open full-size diagram
B clock. All 180 complete seconds. Vertical scale 0–66. Lines mark 50 and 60. Partial tails remain in the table.The table uses B’s full elapsed duration for unique pair rates. Input and final markers must match within the same render.
Each main run contains 180 complete seconds and one partial tail. The raw data retains both input and final counts.
All eight runs passed diagnostic, source, native and cleanup checks. Positive and poison-negative controls passed their full-pixel predicates.
The five-second and three-minute observations are measurements, not full-pixel control qualifications.
Source pressure
Open full-size diagram
A clock. Positive checks only. Dashed line: one-packet byte budget. Dotted line: zero remaining bytes. No continuous buffer trace.Open full-size diagram
A clock. Counts of actual drops in each second and the partial tail. Positive-check records are not counted as drops.The source checks its byte budget at admission, before encoding and after encoding.
A check rejects a packet when its additional bytes exceed the limit or its observed queued bytes exceed the remaining budget.
For these observations, packet size equals the limit. Thus, any positive bufferedAmount can reject the next packet.
The recorded positive checks include queued values below one complete packet.
The recorder stores positive checks and actual drops. It does not store timed bufferedAmount values for passed checks.
The buffer diagrams show discrete positive observations. Lines do not connect the observations.
The byte thresholds in the table are inclusive counts. They describe observed points and do not predict results with another budget.
A positive-check record and its drop record are separate events for one decision. The drop diagrams count only actual drops.
The diagrams use A’s start and end source cuts. B uses its own measurement clock in a separate diagram.
Sequence identity does not establish latency or simultaneous timing across processes.
Startup and cleanup
Startup events remain in the journal. The desktop startup prefix contains sink generations 1 and 2.
Warmup and cleanup events remain separate from the A measurement interval. Cleanup pressure is not part of the measured drop count.
There are no unread pressure-journal gaps. Ring overwrite counts remain explicit in the data.
Diagnostic work
Each context has 21 pressure cuts. Each cut has one 250 ms deadline for capture, pressure drains and pressure-journal writes.
The recorded durations include controller waiting and I/O. They are not CPU costs or a measured effect on rendering.
Existing source checks and the original source-journal writes occur outside this pressure-cut deadline.
The recorder retains 393,312 bytes of numeric storage. It adds clock reads for positive checks and drops.
There is no uninstrumented comparison with the same scheduling conditions.
Limitations
These sequential observations do not establish causation, repeatability, sustained 60 frames/s or hours-long operation.
They do not establish released-product performance or external camera delivery.
No in-run thermal or frequency measurements are available. This report does not identify garbage collection as the cause of pressure.
Native durations include setup and cleanup. Their monotonic clock is separate from both page clocks.
The earlier allocation report contains separate ordinary observations. This diagnostic does not revise those measurements.
Source references
The data includes result, control, journal, source-manifest and native-receipt digests.
SHA256SUMS identifies every public file. Private source receipts remain outside the report.
Numeric data · Text report · Checksums
Separate allocation measurements