rvirmoors

Hello, I looked and I haven't seen this already addressed somewhere.

I'm writing an object that's supposed to play a sequence of audio pulses with critical timing accuracy.

So far I've been assuming that every perform() takes the same physical time, and so I use:

playhead += lib::math::samples_to_milliseconds(dsp_vec_size, samplerate())

The problem is, I also do some other computation in perform() (sending and receiving to/from a LibTorch model), so the resulting clicks are not equidistant.

(i've played with all combinations of Overdrive/Interrupt, same behaviour...)

Is there a way to create a parallel thread with perfect timing accuracy that I can anchor my playhead to?

I'm writing an object that's supposed to play a sequence of audio pulses with critical timing accuracy.
So far I've been assuming that every perform() takes the same physical time, and so I use:
playhead += lib::math::samples_to_milliseconds(dsp_vec_size, samplerate())to compute my current time.

The problem is, I also do some other computation in perform() (sending and receiving to/from a LibTorch model), so the resulting clicks are not equidistant.
(i've played with all combinations of Overdrive/Interrupt, same behaviour...)


perhaps using a timer<> like in min.note.make, would that be sample-accurate though?


Hello, I looked and I haven't seen this already addressed somewhere.
I'm writing an object that's supposed to play a sequence of audio ... 

precise-timing-in-min

HIGH STAR RANCH, is built around community, a colorful local heritage, a love of the outdoors, an appreciation for beautiful living spaces, and a devotion to adventure.

HIGH STAR RANCH,  is built around community, a colorful local heritage, a love of the outdoors, an appreciation for beautiful living spaces, and a devotion to adventure.
barn wedding venues

oops, scratch this, my code was wrong :) .. perform() 

in fact take the same physical time, as it should.

for reference, here's some code to trigger a click every second:

double buf_ms = lib::math::samples_to_milliseconds(vec_size, samplerate());

for (int c = 0; c < output.channel_count(); c++) {

for (int i = 0; i < output.frame_count(); i++) {

if (playhead >= seconds * 1000. - buf_ms) {

int ratio = (seconds * 1000. - playhead) * vec_size / buf_ms;

cout << "seconds: " << seconds << " ratio " << ratio << endl;

this keeps working even if I add back all the other fancy stuff I had in perform(). yay!

oops, scratch this, my code was wrong :) .. perform() does in fact take the same physical time, as it should.

for reference, here's some code to trigger a click every second:
auto vec_size = input.frame_count();
double buf_ms = lib::math::samples_to_milliseconds(vec_size, samplerate());
  playhead += buf_ms;
  for (int c = 0; c < output.channel_count(); c++) {
    auto out = output.samples(c);
    for (int i = 0; i < output.frame_count(); i++) {
      out[i] = 0.;
    }
    if (playhead >= seconds * 1000. - buf_ms) {
      int ratio = (seconds * 1000. - playhead) * vec_size / buf_ms;
      seconds++;
      out[ratio] = -1.;
      cout << "seconds: " << seconds << " ratio " << ratio << endl;
    }
  }
this keeps working even if I add back all the other fancy stuff I had in perform(). yay!


Precise timing in Min?