Proper Use of Optimized DSPInt Functions

[grinch]

Hi, I am experimenting with writing my own audio library objects. Look a build a harmonic distortion algorithm and thought I'd try and use the optimized integer math functions the audio library provides.

My harmonic distortion is a transfer function represented by:

y = x - x^2 * x2c - x^3 * x3c

For variable amounts of second and third order harmonic distortion.

My question is whether I'm using the dspint functions correctly for this purpose. Does this implementation look alright?

void Harmonic_Distortion::update(void)
{
  audio_block_t *block = receiveWritable(0);

  if(block){
    for (int i=0; i < AUDIO_BLOCK_SAMPLES; i++) {
      int32_t x2 = multiply_16bx16b(block->data[i], block->data[i]);
      int32_t x3 = multiply_32x16b(x2, block->data[i]);
      x2 = multiply_32x16b(x2c, x2);
      x3 = multiply_32x16b(x3c, x3);
      block->data[i] -= (x2 + x3);
      block->data[i] = signed_saturate_rshift(block->data[i], 16, 0);
    }
    transmit(block);
    release(block);
  }

}

 

[WMXZ]

I would say, that does not work as intended.
Not for using DSPInt functions that are really not necessary for 32-bit operations, but compiler should be able to handle that.
But for taking the 3rd power of a 16bit value. In particular x3, defined as 32 bit can easily be too small for a potential 48 bit value (3*16 bit)
scaling all multiplication results back to 16 bit could help