The computed duration of erfsquare waveform templates is incorrect

[antalsz]

ScheduledBasicBlock::get_waveform_duration_seconds assumes that the duration of all built-in waveforms is equal to their duration parameter. This isn't true for erfsquare, though; the duration is just that of the main pulse, and the padleft and padright parameters specify silent padding on either side, which contributes to the total duration.

To see this, consider the following program:

DEFFRAME 0 "a":
    SAMPLE-RATE: 1e9
PULSE 0 "a" erfsquare(duration: 1.0, padleft: 0.2, padright: 0.3)
PULSE 0 "a" erfsquare(duration: 0.1, padleft: 0.7, padright: 0.7)
PULSE 0 "a" erfsquare(duration: 0.5, padleft: 0.6, padright: 0.4)
FENCE

If provided as a test case to program::schedule::tests::schedule_seconds, it will produce instruction timings of [0.0, 1.0, 1.1, 1.6], when it should instead produce [0.0, 1.5, 3.0, 4.5].

[erichulburd]

Is the solution in quil-rs to add total_duration to the WaveformTemplate trait?

[antalsz]

@erichulburd That's part of a more in-depth solution, but it requires figuring out how WaveformTemplate integrates into quil-rs; as it stands, it's exported but never used internally. This means that deciding on how WaveformTemplate should be used is a bigger change, and I'd like to fix this specific issue before we resolve that.