Real-Time Data Streaming

SkigenPlot is designed for live data visualization — EEG telemetry, sensor streams, simulation output. The dynamic vertex buffer architecture supports per-frame data updates at 60 Hz+.

Pattern

Call plot() or scatter() each time new data arrives. The widget automatically:

1. Re-interleaves x/y into the GPU vertex buffer
2. Recomputes the bounding box for auto-scaling
3. Uploads and renders in the next frame

// Timer-driven update
QTimer timer;
QObject::connect(&timer, &QTimer::timeout, [&]() {
    // Shift window and append new sample
    x = x.tail(n - 1).eval();
    y = y.tail(n - 1).eval();
    x.conservativeResize(n);
    y.conservativeResize(n);
    x(n - 1) = currentTime();
    y(n - 1) = readSensor();

    view.plot(x, y);
});
timer.start(16);  // ~60 Hz

Buffer Management

Dynamic vertex buffers grow automatically:

Initial capacity	Growth strategy	Behaviour
128 K floats (64 K vertices)	2× when exceeded	No pipeline recreation needed

For fixed-size scrolling windows, the buffer never grows — only contents are updated.

Performance Considerations

• Zero-copy path: When the source is a contiguous Eigen::VectorXf, the template evaluates to a direct memcpy into the interleave buffer.
• Expression evaluation: Non-contiguous expressions (blocks, casts) are evaluated once into a temporary before upload.
• GPU upload: Uses QRhiBuffer::Dynamic with updateDynamicBuffer() — CPU-visible memory, no staging copies.

Use Cases

Application	Update rate	Data size
EEG telemetry (mne-cpp)	60 Hz	1–10 K samples/channel
Sensor dashboard	30 Hz	100–1 K samples
Simulation live view	60 Hz	10 K–1 M points
Training loss curve	1–10 Hz	100–10 K epochs