int32_t pitchS;
uint32_t dpS, _dpS;
uint32_t pS;
float dpM, _dpM, pM;

#if attr_type > 1
uint32_t pS2;
float pM2;
#endif

pitchS = 0;
pS = 0;
pM = 0.0f;
#if attr_type > 1
// 0.142857f sounds better slightly off 1/7
pM2 = attr_type == 5 ? 0.14f : attr_type == 4 ? 0.3333333f : 0.5f;
pS2 = 0;
#endif

int32_t idpM;
#if attr_type == 2
MTOFEXTENDED(inlet_pitch + param_pitch - (12 << 21), idpM)
#else
MTOFEXTENDED(inlet_pitch + param_pitch, idpM)
#endif

dpM = arm::q_to_float(idpM, 32);
_dpM = 1.0f / dpM;

pitchS += (inlet_rate + param_rate) / (3000); // /s
if (pitchS >= (12 << 21)) {
  pitchS -= (12 << 21);
  pS /= 2;
#if attr_type > 1
  pS2 /= 2;
#endif
}
if (pitchS < 0) {
  pitchS += (12 << 21);
  pS *= 2;
#if attr_type > 1
  pS2 *= 2;
#endif
}

#if attr_type == 2
MTOFEXTENDED(pitchS - (12 << 21), dpS)
MTOFEXTENDED(-pitchS + (12 << 21), _dpS)
#else
MTOFEXTENDED(pitchS, dpS)
MTOFEXTENDED(-pitchS, _dpS)
#endif

dpS >>= 7;
uint32_t dpS_8 = dpS >> 3;

pM += dpM;
// hard sync
if (pM >= 1.0f) {
  pM -= 1.0f;
  pS = (int32_t)((pM * _dpM) * dpS);
}

uint32_t accu = 0;
for (int i = 0; i < 8; i++) {
  pS += dpS_8;
  int32_t m = ___SMMUL(pS, idpM);
  accu += (m ^ (m >> 1));
}
outlet_wave = (int32_t)((___SMMUL(accu << 5, _dpS << 3) - pM * 3500000.0f) *
                        (1.0f - pM) * 128.0f);

#if attr_type >= 2
uint32_t accu2 = 0;
pM2 += dpM;
// hard sync
if (pM2 >= 1.0f) {
  pM2 -= 1.0f;
  pS2 = (int32_t)((pM2 * _dpM) * dpS);
}
accu = 0;
for (int i = 0; i < 8; i++) {
  pS2 += dpS_8;
  int32_t m = ___SMMUL(pS2, idpM);
  accu2 += (m ^ (m >> 1));
}
#if attr_type == 2
outlet_wave += (int32_t)((___SMMUL(accu2 << 5, _dpS << 3) - pM2 * 3500000.0f) *
                         (1.0f - pM2) * 128.0f);
#endif
#if attr_type > 2
outlet_wave -= (int32_t)((___SMMUL(accu2 << 5, _dpS << 3) - pM2 * 3500000.0f) *
                         (1.0f - pM2) * 128.0f);
#endif
#endif