Dynamic lifetime issues and the node collection

[orottier]

Apologies for the complex PR. I will let it simmer for a while and put some more thought in it, but wanted to share already.

Introduce AudioProcess::has_side_effects

A processor with no side effect will be dropped when the control handle
is gone and there are not output ports connected.

This makes the cleanup of AudioParams easier and more robust.

Relates to #397 and #468

[orottier]

/bench

[github-actions]

Benchmark result:


bench_ctor
  Instructions:             2761321 (-1.472882%)
  L1 Accesses:              4178740 (-1.329129%)
  L2 Accesses:                27509 (+0.043641%)
  RAM Accesses:               61624 (-0.019469%)
  Estimated Cycles:         6473125 (-0.867549%)

bench_audio_buffer_decode
  Instructions:             7372803 (-0.000163%)
  L1 Accesses:             10167726 (-0.001465%)
  L2 Accesses:                28338 (+0.514312%)
  RAM Accesses:               32160 (-0.027977%)
  Estimated Cycles:        11435016 (+0.002283%)

bench_sine
  Instructions:            68776515 (-0.320959%)
  L1 Accesses:            100311675 (-0.302505%)
  L2 Accesses:               267026 (+1.560533%)
  RAM Accesses:               62476 (-0.024003%)
  Estimated Cycles:       103833465 (-0.273132%)

bench_sine_gain
  Instructions:            73027923 (-0.429922%)
  L1 Accesses:            106810892 (-0.394051%)
  L2 Accesses:               269252 (-0.599169%)
  RAM Accesses:               62572 (-0.019174%)
  Estimated Cycles:       110347172 (-0.389146%)

bench_sine_gain_delay
  Instructions:           105860309 (-0.384253%)
  L1 Accesses:            153172557 (-0.358576%)
  L2 Accesses:               504111 (-0.640178%)
  RAM Accesses:               64129 (-0.020268%)
  Estimated Cycles:       157937627 (-0.358291%)

bench_buffer_src
  Instructions:            18298022 (-1.075672%)
  L1 Accesses:             26870516 (-0.997920%)
  L2 Accesses:               138011 (+3.236737%)
  RAM Accesses:              100672 (-0.020855%)
  Estimated Cycles:        31084091 (-0.797692%)

bench_buffer_src_delay
  Instructions:            75266383 (-0.382703%)
  L1 Accesses:            106373406 (-0.375810%)
  L2 Accesses:               342416 (+3.239345%)
  RAM Accesses:              100821 (-0.019833%)
  Estimated Cycles:       111614221 (-0.311042%)

bench_buffer_src_iir
  Instructions:            80302608 (-0.275095%)
  L1 Accesses:            114705628 (-0.258431%)
  L2 Accesses:               133766 (+0.455846%)
  RAM Accesses:              100748 (-0.019848%)
  Estimated Cycles:       118900638 (-0.247382%)

bench_buffer_src_biquad
  Instructions:            57683971 (-0.845705%)
  L1 Accesses:             77882678 (-0.780861%)
  L2 Accesses:               182297 (-3.689243%)
  RAM Accesses:              100870 (-0.015859%)
  Estimated Cycles:        82324613 (-0.781479%)

bench_stereo_positional
  Instructions:            95544000 (-1.573564%)
  L1 Accesses:            139796232 (-1.473760%)
  L2 Accesses:               399463 (+4.077235%)
  RAM Accesses:              100998 (-0.014850%)
  Estimated Cycles:       145328477 (-1.366454%)

bench_stereo_panning_automation
  Instructions:            29485493 (-0.971009%)
  L1 Accesses:             44416339 (-0.905535%)
  L2 Accesses:               159628 (+10.69212%)
  RAM Accesses:              100786 (-0.021824%)
  Estimated Cycles:        48741989 (-0.671583%)

bench_analyser_node
  Instructions:            37441279 (-0.588119%)
  L1 Accesses:             52473369 (-0.567977%)
  L2 Accesses:               187114 (+1.717821%)
  RAM Accesses:              101280 (-0.015795%)
  Estimated Cycles:        56953739 (-0.497044%)

bench_hrtf_panners
  Instructions:           320109462 (-0.084709%)
  L1 Accesses:            562456729 (-0.064506%)
  L2 Accesses:             10853194 (-0.099714%)
  RAM Accesses:              120791 (-0.076107%)
  Estimated Cycles:       620950384 (-0.067663%)

bench_sine_gain_with_worklet
  Instructions:            73521626 (-0.396689%)
  L1 Accesses:            107450201 (-0.354992%)
  L2 Accesses:               266740 (-1.431924%)
  RAM Accesses:               62747 (-0.015934%)
  Estimated Cycles:       110980046 (-0.361389%)


running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

[orottier]

Right, I think we should probably merge this as it solves a panic in the audio param dynamic lifetimes.
@b-ma did you manage to reproduce the panic you experienced in your complex graph? I hope this PR would solve it.

[b-ma]

No I didn't have time to make a simpler test bed, but I can test my app against this branch right now, the panic was rather easy to get, I let you now in the afternoon

[b-ma]

Just tested and I didn't manager to make it crash!

[b-ma]

Didn't understood all the details, but I will check that when needed and it looks all good to me

[orottier]

Cool, good to hear!