EXPORT_FCN void MWDSP_blms_an_wn_CC( const creal32_T *inSigU,
const creal32_T *deSigU,
const real32_T muU,
creal32_T *inBuff,
creal32_T *wgtBuff,
const int_T FilterLength,
const int_T BlockLength,
const real32_T LkgFactor,
const int_T FrmLen,
creal32_T *outY,
creal32_T *errY)
{
int_T i,j,k,m=0;
const int_T FiltLen_minus_1 = FilterLength-1;
const int_T NumberOfFrame = (int_T)(FrmLen/BlockLength + 0.5); /* To avoid precision problem */
const int_T bytesPerFiltLen = FilterLength*sizeof(creal32_T);
const int_T bytesPerBlkLen = BlockLength*sizeof(creal32_T);
memset(outY,0,sizeof(creal32_T)*FrmLen);
for (i=0; i<NumberOfFrame; i++)
{
/* Step-1: Copy new BlockLength samples at the END of the linear buffer (has length = FilterLength+BlockLength */
memmove(inBuff, (inBuff + BlockLength), bytesPerFiltLen);
memcpy((inBuff + FilterLength), inSigU + i*BlockLength, bytesPerBlkLen);
/* Step-2: convolve inBuff_vector (length= FilterLength+BlockLength) and wgtIC_vector(length=FilterLength) (not yet updated) and */
/* resultantLen = FilterLength+BlockLength + FilterLength - 1, but we need only FilterLength+1:(FilterLength+BlockLength) */
for (j=0; j <BlockLength; j++)
{
int_T j_plus_1 = j+1;
for (k=0; k < FilterLength; k++)
{
outY[m].re += CMULT_RE(wgtBuff[FiltLen_minus_1-k],inBuff[k+j_plus_1]);
outY[m].im += CMULT_IM(wgtBuff[FiltLen_minus_1-k],inBuff[k+j_plus_1]);
}
/* Step-3: get error for the current sample in the block */
errY[m].re = deSigU[m].re - outY[m].re;
errY[m].im = deSigU[m].im - outY[m].im;
m++;
}
/* Step-4: correlate inBuff_vector (length= FilterLength+BlockLength) and errY_vector(length=BlockLength) and */
/* resultantLen = FilterLength+BlockLength + FilterLength - 1, but we need only BlockLength+1:(FilterLength+BlockLength) */
m=i*BlockLength;
for (j=0; j <FilterLength; j++)
{
creal32_T tmpSum = {0,0};
int_T j_plus_1 = j+1;
int_T FiltLen_minus_1_minus_j = FiltLen_minus_1-j;
for (k=0; k < BlockLength; k++)
{
tmpSum.re += muU * CMULT_YCONJ_RE(errY[m+k], inBuff[k+j_plus_1]);
tmpSum.im += muU * CMULT_YCONJ_IM(errY[m+k], inBuff[k+j_plus_1]);
}
wgtBuff[FiltLen_minus_1_minus_j].re = tmpSum.re + LkgFactor*(wgtBuff[FiltLen_minus_1_minus_j].re);
wgtBuff[FiltLen_minus_1_minus_j].im = tmpSum.im + LkgFactor*(wgtBuff[FiltLen_minus_1_minus_j].im);
}
m=(i+1)*BlockLength;
}
}
EXPORT_FCN void MWDSP_2ChSBank_DF_ZZ(
const creal_T *inputToLongFilt,
const creal_T *inputToShortFilt,
creal_T *out,
creal_T *longFiltTapBuf,
creal_T *shortFiltTapBuf,
const creal_T *const longFilt,
const creal_T *const shortFilt,
int32_T *longFiltTapIdx,
int32_T *shortFiltTapIdx,
const int_T numChans,
const int_T inFrameSize,
const int_T polyphaseLongFiltLen,
const int_T polyphaseShortFiltLen
)
{
const int_T iFactor = 2;
/* initialize local variables to rid compiler warnings */
int_T k = numChans;
int_T curTapLong = 0;
int_T curTapShort = 0;
int_T jlong = 0;
int_T jshort = 0;
creal_T sum;
sum.re = 0.0;
sum.im = 0.0;
/* loop through k=numChans channels */
do {
int_T i = inFrameSize;
/* make per channel copy of long and short tap indices which are common to all channels */
curTapLong = *longFiltTapIdx;
curTapShort = *shortFiltTapIdx;
/* loop through inFrameSize samples in each input frame */
while (i--) {
int_T m = iFactor;
/* make per sample copies of filter addresses and input tap delay buffer addresses
which are common to all input samples */
const creal_T *cffLong = longFilt;
const creal_T *cffShort = shortFilt;
creal_T *memLong = longFiltTapBuf + curTapLong;
creal_T *memShort = shortFiltTapBuf + curTapShort;
/* read input sample into the tap delay buffer locations */
*memLong = *inputToLongFilt++;
*memShort = *inputToShortFilt++;
/* loop through iFactor interpolation phases
output results to output buffer location pointed at by out */
while (m--) {
/* perform direct form FIR filtering on the first curTapLong taps
of the Long Tap Delay Buffer (longFiltTapBuf) */
for (jlong = 0; jlong <= curTapLong; jlong++) {
sum.re += CMULT_RE(*memLong, *cffLong);
sum.im += CMULT_IM(*memLong, *cffLong);
--memLong;
++cffLong;
}
/* perform direct form FIR filtering on the first curTapShort taps
of the Short Tap Delay Buffer (shortFiltTapBuf) and add the
intermediate result to previous sum */
for (jshort = 0; jshort <= curTapShort; jshort++) {
sum.re += CMULT_RE(*memShort, *cffShort);
sum.im += CMULT_IM(*memShort, *cffShort);
--memShort;
++cffShort;
}
/* point memLong and memShort to the last tap delay buffer locations */
memLong += polyphaseLongFiltLen;
memShort += polyphaseShortFiltLen;
/* perform direct form FIR filtering on the remaining taps for the
long and short filters and add their results */
while(jlong++ < polyphaseLongFiltLen) {
sum.re += CMULT_RE(*memLong, *cffLong);
sum.im += CMULT_IM(*memLong, *cffLong);
--memLong;
++cffLong;
}
while(jshort++ < polyphaseShortFiltLen) {
sum.re += CMULT_RE(*memShort, *cffShort);
sum.im += CMULT_IM(*memShort, *cffShort);
--memShort;
++cffShort;
}
/* output sum to out and reset sum */
*out++ = sum;
sum.re = 0.0;
sum.im = 0.0;
}
/* increment circular buffer pointers : curTapLong and curTapShort */
if ( (++curTapLong) >= polyphaseLongFiltLen ) curTapLong = 0;
if ( (++curTapShort) >= polyphaseShortFiltLen ) curTapShort = 0;
} /* inFrameSize */
/* increment for next channel */
longFiltTapBuf += polyphaseLongFiltLen;
shortFiltTapBuf += polyphaseShortFiltLen;
} while ((--k) > 0);
/* Update stored indices for next function call */
*longFiltTapIdx = curTapLong;
*shortFiltTapIdx = curTapShort;
}