isvr2020

I am making my first MSP object with more than one signal output. In my design, the number of signal input channels will always match the number of signal output channels. For practice, I want to make a simple pass-through that copies the input for each channel to the associated output channel, i.e. input 0 -> output 0, input 1 -> output 1, etc.

However, I am encountering unexpected behaviour (based on my understanding). For example, with 4 i/o channels, input 0 (leftmost) is copied to output 0 and output 3 (rightmost), and input 1 is copied to output 1 and output 2. Inputs 2 and 3 go no where seemingly.

This is my entire perform routine as of now (num_chans == 4 in the case I described above):

short num_chans = x->num_chans; // number of i/o channels (not total)

short i, j;
for (i = 0; i < num_chans; i++)
{
   for (j = 0; j < sampleframes; j++)
   { 
      outs[i][j] = ins[i][j];
   }
}

If anyone can help steer me in the right direction, that would be greatly appreciated.

However, I am encountering unexpected behaviour (based on my understanding). For example, with 4 i/o channels, input 0 (leftmost) is copied to output 0 and output 3 (rightmost), and input 1 is copied to output 1 and output 2.  Inputs 2 and 3 go no where seemingly.

Hi Everyone, 
I am making my first MSP object with more than one signal output. In my design, the number of signal input channels will alway

trouble-creating-multichannel-audio-external

I am reviving this old thread because I have exactly the same issue. I have an external with a number of audio outlets set by argument.

This code in the perform64 method doesn't work:

void pan_offset_perform64(t_pan *x, t_object *dsp64, double **ins, long numins, double **outs, long numouts, long sampleframes, long flags, void *userparam)

        for (int i = 0; i < x->x_outlets; i++) {

            *out++ = *in1++ * x->x_weights[i];

            if (x->x_test == 1) { // debugging

                object_post((t_object *)x, "j : %ld", j);

                object_post((t_object *)x, "outs[%ld][%ld] : %f", i, j, *out);

With the x->x_test flag set to 1 using another method, a configuration of 5 outlets and a signal vector size of 8 samples, I expected to get 5 x 8 = 40 values but got 5 only.

So, I guess it means I should iterate over the outlets in the dsp64 method instead, right? Or is there something wrong with my code?

This code in the perform64 method doesn't work:
void pan_offset_perform64(t_pan *x, t_object *dsp64, double **ins, long numins, double **outs, long numouts, long sampleframes, long flags, void *userparam)
{
	t_double	*in1 = ins[0];
	t_double	*out;
	
...(some calculations in between)...

	int j = 0;
	while (sampleframes--) {
		for (int i = 0; i < x->x_outlets; i++) {
			out = outs[i];
			*out++ = *in1++ * x->x_weights[i];
			if (x->x_test == 1) { // debugging
				object_post((t_object *)x, "j : %ld", j);
				object_post((t_object *)x, "outs[%ld][%ld] : %f", i, j, *out);
			}
		}
		j++;
	}	
	x->x_test = 0; //reset test flag
}

With the x->x_test flag set to 1 using another  method, a configuration of 5 outlets and a signal vector size of 8 samples, I expected to get 5 x 8 = 40 values but got 5 only.

try to nest loops in the other way, ie iterate over frames for each channel/outlet. In your code you set the pointer `out` to the first frame of the buffer for a channel/outlet each time you enter inner loop. Also you move `in1` ptr far beyond input buffer, so probably only the size of a struct and small vector size keep your external from crashing. 

OP had different problem, related to reusing signal buffers by DSP graph optimisation, which can be solved by setting Z_NO_INPLACE flag.

@Roald
try to nest loops in the other way, ie iterate over frames for each channel/outlet. In your code you set the pointer `out` to the first frame of the buffer for a channel/outlet each time you enter inner loop. Also you move `in1` ptr far beyond input buffer, so probably only the size of a struct and small vector size keep your external from crashing. 
OP had different problem, related to reusing signal buffers by DSP graph optimisation, which can be solved by setting Z_NO_INPLACE flag.


I think this alternative coding is what you suggests. Actually it's the first code I have tried but it doesn't provide better results.

        t_double    *out = outs[i];

        for (int j = 0; j < sampleframes; j++) {            

            *(out + j) = *(in1 + j) * (x->x_weights[i]);

                object_post((t_object *)x, "outs[%ld][%ld] : %f", i, j, *(out+j));

However it doesn't function better. If I activate the test flag, I have 5 lines with j = 0 but no iteration over j.

Also, I don't understand which part of the code makes you write "Also you move `in1` ptr far beyond input buffer (...)". Maybe there's still something I don't understand completely about pointers...

for (int i = 0; i < x->x_outlets; i++) {
		t_double	*out = outs[i];
		for (int j = 0; j < sampleframes; j++) {			
			*(out + j) = *(in1 + j) * (x->x_weights[i]);
			if (x->x_test == 1) { // debugging
				object_post((t_object *)x, "j : %ld", j);
				object_post((t_object *)x, "outs[%ld][%ld] : %f", i, j, *(out+j));
			}
		}
	}
	x->x_test = 0; //reset test flag

Also, I don't understand which part of the code makes you write "Also you move `in1` ptr far beyond input buffer (...)".  Maybe there's still something I don't understand completely about pointers...

 must be equal 1 or you haven't loaded new version of your external. Now code looks ok. Regarding 

 in your previous post: you increased the pointer

From what you wrote sampleframes must be equal 1 or you haven't loaded new version of your external. Now code looks ok. Regarding in1 in your previous post: you increased the pointerx->x_outlets*sampleframestimes in a row, so it went far beyond ins[0] buffer (of sampleframes size). 

Ahah. It was actually a bad implementation of the 

Ahah. It was actually a bad implementation of the Z_NO_INPLACE flag in my code.

I've added this line in the new method:
x->x_obj.z_misc = Z_NO_INPLACE;

Trouble creating multichannel audio external