float x0, x1, I0, I1, out, inGain, outGain;
int32_t old_in;
int32_t outMax, outMin;

x0 = x1 = I0 = I1 = out = 0;
old_in = 0;

inGain = param_InGain * (1.0f / (1 << 25)) * (1.0f / (1 << 27));
outGain = 2.0f * param_OutGain;
outMax = (int32_t)(outGain);
outMin = -outMax;

// precedent value of input
x1 = x0;
// current input value (float)
x0 = inlet_in * inGain;
// precedent value of Integral
I1 = I0;
// calc of current Integral
// sat(x) = x  for |x| <  1
// sat(x) = 1  for  x >=  1
// sat(x) = -1 for  x <= -1
// Int(sat(x))=
// 0.5*x^2 for |x|<1
// -0.5+x   for |x|>=1
I0 = fabs(x0);
if (I0 <= 1) {
  I0 = 0.5f * I0 * I0;
} else {
  I0 = I0 - 0.5f;
}
// if the precendent and current input are different enough
// we use the differenciation trick (ie mean on the x1 x0 interval
if ((inlet_in & 0xFFFFF000) != (old_in & 0xFFFFF000)) {
  outlet_out = (int32_t)(outGain * (I0 - I1) / (x0 - x1));
} else {
  // if the precedent and current values are close, no antialiasing
  // is necessary and we avoid a division by a small x0-x1:
  // we simply calculate the value of the saturation
  outlet_out = x0 >= 1 ? outMax : x0 <= -1 ? outMin : (int32_t)(outGain * x0);
}
old_in = inlet_in;