ferrite-lithic811 tests · 21 designs

ferrite-lithic-wavephase A342 tests

Waveform data

Cycle-indexed values you assert on directly, plus a VCD rendering for when you want eyes.

This crate exists because a waveform viewer is a debugging tool and an assertion is a test. Both are useful and they answer different questions: the viewer tells you what happened, the assertion tells you whether it should have.

So the data model is a value per cycle per port, and the primary API is assert_series. The VCD rendering is a secondary function of the same data, not a separate capture path.

Step by step

  1. 01

    Register the ports, with their widths

    register and register_shape attach a port to the dataset; top names the scope and identifiers gives the names used in the output.

    use ferrite_lithic_wave::WaveData;

let mut wave = WaveData::new("top");
wave.register(&port, "count")?;
  2. 02

    Record a cycle

    record takes the cycle number and a lookup closure, so the data can be gathered from a snapshot without the wave crate knowing what a snapshot is.

    wave.record(cycle, |signal| {
    Ok(snapshot.value(signal))
})?;
  3. 03

    Assert the series, not the last value

    assert_series compares the whole cycle-indexed sequence and reports the first cycle that differs. This is the single most useful function in the crate: a decoder that retires 39 of 40 symbols produces a correct final answer and a failing series.

    wave.assert_series(&port, &[
    Bits::constant(0, 8)?,
    Bits::constant(1, 8)?,
    Bits::constant(3, 8)?,
])?;
  4. 04

    Read the changes when the series is long

    series_changes returns (cycle, value) pairs for the cycles where the value moved, which is what makes a 40-cycle run of mostly-zeros readable in a failure message.

  5. 05

    Write VCD when you want to look at it

    write consumes the dataset into a Write, and render gives a string. Same data either way, so a waveform can never disagree with the assertion that passed.

What bites

  • Series are indexed by cycle, not by position

    A gap in cycle numbers is a gap, not a shift. A design that stops clocking produces a short series and a failing assertion rather than a silently compressed one.