#ifndef cwAudioTransforms_h #define cwAudioTransforms_h namespace cw { namespace dsp { //--------------------------------------------------------------------------------------------------------------------------------- // Window Function // namespace wnd_func { enum { kInvalidWndId = 0x000, kHannWndId = 0x001, kHammingWndId = 0x002, kTriangleWndId = 0x004, kKaiserWndId = 0x008, kHannMatlabWndId= 0x010, kUnityWndId = 0x020, kWndIdMask = 0x0ff, kNormByLengthWndFl = 0x100, // mult by 1/wndSmpCnt kNormBySumWndFl = 0x200, // mult by wndSmpCnt/sum(wndV) kSlRejIsBetaWndFl = 0x400 // kaiser beta arg. is being passed as kaiserSideLobeRejectDb arg. }; template< typename sample_t > struct obj_str { unsigned wndTypeId; // unsigned flags; // sample_t* wndV; // wndV[ wndN ] unsigned wndN; // length of wndV[] and outV[] sample_t* outV; // outV[ wndN ] double kaiserSLRejectDb; // }; typedef struct obj_str fobj_t; typedef struct obj_str dobj_t; const char* wndIdToLabel( unsigned id ); // window type to label unsigned wndLabelToId( const char* label ); // window type label to id template< typename sample_t > rc_t create( struct obj_str*& p, unsigned wndId, unsigned wndSmpCnt, double kaiserSideLobeRejectDb ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str< sample_t > >(); p->wndV = mem::allocZ(wndSmpCnt); p->outV = mem::allocZ(wndSmpCnt); p->wndN = wndSmpCnt; p->wndTypeId = wndId & kWndIdMask; p->flags = wndId & ~kWndIdMask; switch( p->wndTypeId ) { case kHannWndId: hann(p->wndV,p->wndN); break; case kHammingWndId: hamming(p->wndV,p->wndN); break; case kTriangleWndId: triangle(p->wndV,p->wndN); break; case kKaiserWndId: { sample_t beta = (p->flags & kSlRejIsBetaWndFl) ? kaiserSideLobeRejectDb : kaiser_beta_from_sidelobe_reject(kaiserSideLobeRejectDb); kaiser(p->wndV,p->wndN, beta); } break; case kHannMatlabWndId: hann_matlab(p->wndV, p->wndN); break; case kUnityWndId: vop::fill(p->wndV,p->wndN,1); break; default: rc = cwLogError(kInvalidArgRC,"The window id '%i' (0x%x) is not valid.", wndId, wndId ); } sample_t den = 0; sample_t num = 1; if( cwIsFlag(p->flags,kNormBySumWndFl) ) { den = vop::sum(p->wndV, p->wndN); num = wndSmpCnt; } if( cwIsFlag(p->flags,kNormByLengthWndFl) ) den += wndSmpCnt; if( den > 0 ) { vop::mul(p->wndV,num,p->wndN); vop::div(p->wndV,den,p->wndN); } if( rc != kOkRC ) destroy(p); return rc; } template< typename sample_t > rc_t destroy( struct obj_str*& p ) { if( p != nullptr ) { mem::release(p->outV); mem::release(p->wndV); mem::release(p); } return kOkRC; } template< typename sample_t > rc_t exec( struct obj_str* p, const sample_t* sigV, unsigned sigN, sample_t* outV=nullptr, unsigned outN=0 ) { rc_t rc = kOkRC; if( outN > p->wndN ) return cwLogError(kInvalidArgRC,"The signal size (%i) is greater than the window size (%i). ",outN,p->wndN); if( outV == nullptr ) { outV = p->outV; outN = p->wndN; } vop::mul( outV, p->wndV, sigV, outN ); return rc; } rc_t test( const cw::object_t* args ); } //--------------------------------------------------------------------------------------------------------------------------------- // Overlap Add // namespace ola { template< typename sample_t > struct obj_str { wnd_func::obj_str* wf; // unsigned wndSmpCnt; // unsigned hopSmpCnt; // unsigned procSmpCnt; // sample_t* bufV; // bufV[wndSmpCnt] overlap add buffer sample_t* outV; // outV[hopSmpCnt] output vector sample_t* outPtr; // outPtr[procSmpCnt] output vector unsigned idx; // idx of next val in bufV[] to be moved to outV[] }; typedef struct obj_str fobj_t; typedef struct obj_str dobj_t; // hopSmpCnt must be <= wndSmpCnt. // hopSmpCnt must be an even multiple of procSmpCnt. // Call exec() at the spectral frame rate. // Call execOut() at the time domain audio frame rate. // Set wndTypeId to one of the cmWndFuncXXX enumerated widnow type id's. template< typename sample_t > rc_t create( struct obj_str*& p, unsigned wndSmpCnt, unsigned hopSmpCnt, unsigned procSmpCnt, unsigned wndTypeId ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str >(); if((rc = wnd_func::create( p->wf, wndTypeId, wndSmpCnt, 0)) != kOkRC ) return rc; p->bufV = mem::allocZ( wndSmpCnt ); p->outV = mem::allocZ( hopSmpCnt ); p->outPtr = p->outV + hopSmpCnt; // hopSmpCnt must be an even multiple of procSmpCnt assert( hopSmpCnt % procSmpCnt == 0 ); assert( wndSmpCnt >= hopSmpCnt ); p->wndSmpCnt = wndSmpCnt; p->hopSmpCnt = hopSmpCnt; p->procSmpCnt = procSmpCnt; p->idx = 0; return rc; } template< typename sample_t > rc_t destroy( struct obj_str*& p ) { rc_t rc = kOkRC; if( p != nullptr ) { wnd_func::destroy(p->wf); mem::release( p->bufV ); mem::release( p->outV ); mem::release( p ); } return rc; } template< typename sample_t > rc_t exec( struct obj_str* p, const sample_t* sp, unsigned sN ) { rc_t rc = kOkRC; assert( sN == p->wndSmpCnt ); const sample_t* ep = sp + sN; const sample_t* wp = p->wf->wndV; int i,j,k,n; // [Sum head of incoming samples with tail of ola buf] // fill outV with the bufV[idx:idx+hopSmpCnt] + sp[hopSmpCnt] for(i=0; i<(int)p->hopSmpCnt; ++i) { p->outV[i] = p->bufV[p->idx++] + (*sp++ * *wp++); if( p->idx == p->wndSmpCnt ) p->idx = 0; } // [Sum middle of incoming samples with middle of ola buf] // sum next wndSmpCnt - hopSmpCnt samples of sp[] into bufV[] n = p->wndSmpCnt - (2*p->hopSmpCnt); k = p->idx; for(j=0; jbufV[k++] += (*sp++ * *wp++); if( k == (int)p->wndSmpCnt ) k = 0; } // [Assign tail of incoming to tail of ola buf] // assign ending samples from sp[] into bufV[] while( sp < ep ) { p->bufV[k++] = (*sp++ * *wp++); if( k == (int)p->wndSmpCnt ) k = 0; } p->outPtr = p->outV; return rc; } template< typename sample_t > const sample_t* execOut( struct obj_str* p ) { const sample_t* sp = p->outPtr; if( sp >= p->outV + p->hopSmpCnt ) return NULL; p->outPtr += p->procSmpCnt; return sp; } rc_t test( const cw::object_t* args ); } //--------------------------------------------------------------------------------------------------------------------------------- // Shift Buffer // namespace shift_buf { template< typename sample_t > struct obj_str { unsigned bufSmpCnt; // wndSmpCnt + hopSmpCnt sample_t* bufV; // bufV[bufSmpCnt] all other pointers use this memory sample_t* outV; // output window outV[ outN ] unsigned outN; // outN == wndSmpCnt unsigned procSmpCnt; // input sample count unsigned wndSmpCnt; // output sample count unsigned hopSmpCnt; // count of samples to shift the buffer by on each call to cmShiftExec() sample_t* inPtr; // ptr to location in outV[] to recv next sample bool fl; // reflects the last value returned by cmShiftBufExec(). }; typedef obj_str fobj_t; typedef obj_str dobj_t; template< typename sample_t > rc_t create( struct obj_str*& p, unsigned procSmpCnt, unsigned wndSmpCnt, unsigned hopSmpCnt ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str >(); if( hopSmpCnt > wndSmpCnt ) return cwLogError( kInvalidArgRC, "The window sample count (%i) must be greater than or equal to the hop sample count (%i).", wndSmpCnt, hopSmpCnt ); // The worst case storage requirement is where there are wndSmpCnt-1 samples in outV[] and procSmpCnt new samples arrive. p->bufSmpCnt = wndSmpCnt + procSmpCnt; p->bufV = mem::allocZ( p->bufSmpCnt ); p->outV = p->bufV; p->outN = wndSmpCnt; p->wndSmpCnt = wndSmpCnt; p->procSmpCnt = procSmpCnt; p->hopSmpCnt = hopSmpCnt; p->inPtr = p->outV; p->fl = false; return rc; } template< typename sample_t > rc_t destroy( struct obj_str*& p ) { if( p != nullptr ) { mem::release(p->outV); mem::release(p); } return kOkRC; } // Returns true if there are 'wndSmpCnt' available samples at outV[] otherwise returns false. template< typename sample_t > bool exec( struct obj_str* p, const sample_t* sp, unsigned sn ) { assert( sn <= p->procSmpCnt ); // The active samples are in outV[wndSmpCnt] // Stored samples are between outV + wndSmpCnt and inPtr. // if the previous call to this function returned true then the buffer must be // shifted by hopSmpCnt samples - AND sp[] is ignored. if( p->fl ) { // shift the output buffer to the left to remove expired samples p->outV += p->hopSmpCnt; // if there are not wndSmpCnt samples left in the buffer if( p->inPtr - p->outV < p->wndSmpCnt ) { // then copy the remaining active samples (between outV and inPtr) // to the base of the physicalbuffer unsigned n = p->inPtr - p->outV; memmove( p->bufV, p->outV, n * sizeof(sample_t)); p->inPtr = p->bufV + n; // update the input and output positions p->outV = p->bufV; } } else { // if the previous call to this function returned false then sp[sn] should not be ignored assert( p->inPtr + sn <= p->outV + p->bufSmpCnt ); // copy the incoming samples into the buffer vop::copy(p->inPtr,sp,sn); p->inPtr += sn; } // if there are at least wndSmpCnt available samples in outV[] p->fl = p->inPtr - p->outV >= p->wndSmpCnt; return p->fl; } rc_t test( const cw::object_t* args ); } //--------------------------------------------------------------------------------------------------------------------------------- // Phase to Frequency // namespace phs_to_frq { template< typename T > struct obj_str { T* hzV; // hzV[binCnt] output vector - frequency in Hertz T* phsV; // phsV[binCnt] T* wV; // bin freq in rads/hop double srate; unsigned hopSmpCnt; unsigned binCnt; }; typedef obj_str< float > fobj_t; typedef obj_str< double> dobj_t; template< typename T > rc_t create( struct obj_str*& p, const T& srate, unsigned binCnt, unsigned hopSmpCnt ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str >(); p->hzV = mem::allocZ( binCnt ); p->phsV = mem::allocZ( binCnt ); p->wV = mem::allocZ( binCnt ); p->srate = srate; p->binCnt = binCnt; p->hopSmpCnt = hopSmpCnt; for(unsigned i=0; iwV[i] = M_PI * i * hopSmpCnt / (binCnt-1); return rc; } template< typename T > rc_t destroy( struct obj_str*& p ) { if( p != nullptr ) { mem::release( p->hzV ); mem::release( p->phsV ); mem::release( p->wV ); mem::release( p ); } return kOkRC; } template< typename T > rc_t exec( struct obj_str* p, const T* phsV ) { rc_t rc = kOkRC; unsigned i; double twoPi = 2.0 * M_PI; double den = twoPi * p->hopSmpCnt; for(i=0; ibinCnt; ++i) { T dPhs = phsV[i] - p->phsV[i]; // unwrap phase - see phase_study.m for explanation T k = round( (p->wV[i] - dPhs) / twoPi); // convert phase change to Hz p->hzV[i] = (k * twoPi + dPhs) * p->srate / den; // store phase for next iteration p->phsV[i] = phsV[i]; } return rc; } } //--------------------------------------------------------------------------------------------------------------------------------- // Phase Vocoder (Analysis) // namespace pv_anl { enum { kNoCalcHzPvaFl = 0x00, kCalcHzPvaFl = 0x01, }; template< typename T > struct obj_str { struct shift_buf::obj_str* sb; struct fft::obj_str* ft; struct wnd_func::obj_str* wf; struct phs_to_frq::obj_str* pf; unsigned flags; unsigned procSmpCnt; T srate; unsigned wndSmpCnt; unsigned hopSmpCnt; unsigned binCnt; const T* magV; // amplitude NOT power - alias to ft->magV const T* phsV; // alias to ft->phsV const T* hzV; }; typedef obj_str< float > fobj_t; typedef obj_str< double> dobj_t; template< typename T > rc_t create( struct obj_str*& p, unsigned procSmpCnt, const T& srate, unsigned wndSmpCnt, unsigned hopSmpCnt, unsigned flags ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str >(); shift_buf::create( p->sb, procSmpCnt, wndSmpCnt, hopSmpCnt ); wnd_func::create( p->wf, wnd_func::kHannWndId | wnd_func::kNormByLengthWndFl, wndSmpCnt, 0 ); fft::create( p->ft, wndSmpCnt, fft::kToPolarFl); phs_to_frq::create(p->pf, srate, p->ft->binN, hopSmpCnt ); p->flags = flags; p->procSmpCnt = procSmpCnt; p->wndSmpCnt = wndSmpCnt; p->hopSmpCnt = hopSmpCnt; p->binCnt = p->ft->binN; p->magV = p->ft->magV; p->phsV = p->ft->phsV; p->hzV = p->pf->hzV; return rc; } template< typename T > rc_t destroy( struct obj_str*& p ) { if( p != nullptr ) { shift_buf::destroy( p->sb ); wnd_func::destroy( p->wf ); fft::destroy( p->ft ); phs_to_frq::destroy( p->pf ); mem::release( p ); } return kOkRC; } template< typename T > bool exec( struct obj_str* p, const T* x, unsigned xN ) { bool fl = false; while( shift_buf::exec(p->sb,x,xN) ) { wnd_func::exec(p->wf, p->sb->outV, p->sb->wndSmpCnt ); fft::exec(p->ft, p->wf->outV, p->wf->wndN); if( cwIsFlag(p->flags,kCalcHzPvaFl) ) phs_to_frq::exec(p->pf,p->phsV); fl = true; } return fl; } } //--------------------------------------------------------------------------------------------------------------------------------- // Phase Vocoder (Synthesis) // namespace pv_syn { template< typename T > struct obj_str { ifft::obj_str* ft; wnd_func::obj_str* wf; ola::obj_str* ola; T* minRphV; T* maxRphV; T* itrV; T* phs0V; T* mag0V; T* phsV; T* magV; double outSrate; unsigned procSmpCnt; unsigned wndSmpCnt; unsigned hopSmpCnt; unsigned binCnt; }; typedef obj_str< float > fobj_t; typedef obj_str< double> dobj_t; template< typename T > rc_t create( struct obj_str*& p, unsigned procSmpCnt, const T& outSrate, unsigned wndSmpCnt, unsigned hopSmpCnt, unsigned wndTypeId=wnd_func::kHannWndId ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str >(); int k; double twoPi = 2.0 * M_PI; bool useHannFl = true; int m = useHannFl ? 2 : 1; p->outSrate = outSrate; p->procSmpCnt = procSmpCnt; p->wndSmpCnt = wndSmpCnt; p->hopSmpCnt = hopSmpCnt; p->binCnt = wndSmpCnt / 2 + 1; p->minRphV = mem::allocZ( p->binCnt ); p->maxRphV = mem::allocZ( p->binCnt ); p->itrV = mem::allocZ( p->binCnt ); p->phs0V = mem::allocZ( p->binCnt ); p->phsV = mem::allocZ( p->binCnt ); p->mag0V = mem::allocZ( p->binCnt ); p->magV = mem::allocZ( p->binCnt ); wnd_func::create( p->wf, wndTypeId, wndSmpCnt, 0); ifft::create( p->ft, p->binCnt ); ola::create( p->ola, wndSmpCnt, hopSmpCnt, procSmpCnt, wndTypeId ); for(k=0; k<(int)p->binCnt; ++k) { // complete revolutions per hop in radians p->itrV[k] = twoPi * floor((double)k * hopSmpCnt / wndSmpCnt ); p->minRphV[k] = ((T)(k-m)) * hopSmpCnt * twoPi / wndSmpCnt; p->maxRphV[k] = ((T)(k+m)) * hopSmpCnt * twoPi / wndSmpCnt; //printf("%f %f %f\n",p->itrV[k],p->minRphV[k],p->maxRphV[k]); } return rc; } template< typename T > rc_t destroy( struct obj_str*& p ) { if( p != nullptr ) { wnd_func::destroy(p->wf); ifft::destroy(p->ft); ola::destroy(p->ola); mem::release(p->minRphV); mem::release(p->maxRphV); mem::release(p->itrV); mem::release(p->phs0V); mem::release(p->phsV); mem::release(p->mag0V); mem::release(p->magV); mem::release( p ); } return kOkRC; } template< typename T > rc_t exec( struct obj_str* p, const T* magV, const T* phsV ) { double twoPi = 2.0 * M_PI; unsigned k; for(k=0; kbinCnt; ++k) { // phase dist between cur and prv frame T dp = phsV[k] - p->phs0V[k]; // dist must be positive (accum phase always increases) if( dp < -0.00001 ) dp += twoPi; // add in complete revolutions based on the bin frequency // (these would have been lost from 'dp' due to phase wrap) dp += p->itrV[k]; // constrain the phase change to lie within the range of the kth bin if( dp < p->minRphV[k] ) dp += twoPi; if( dp > p->maxRphV[k] ) dp -= twoPi; p->phsV[k] = p->phs0V[k] + dp; p->magV[k] = p->mag0V[k]; p->phs0V[k] = phsV[k]; p->mag0V[k] = magV[k]; } ifft::exec_polar( p->ft, magV, phsV ); ola::exec( p->ola, p->ft->outV, p->ft->outN ); //printf("%i %i\n",p->binCnt,p->ft.binCnt ); //cmVOR_Print( p->obj.ctx->outFuncPtr, 1, p->binCnt, magV ); //cmVOR_Print( p->obj.ctx->outFuncPtr, 1, p->binCnt, p->phsV ); //cmVOS_Print( p->obj.ctx->outFuncPtr, 1, 10, p->ft.outV ); return kOkRC; } } //--------------------------------------------------------------------------------------------------------------------------------- // Spectral Distortion // namespace spec_dist { template< typename T0, typename T1 > struct obj_str { T1 ceiling; T1 expo; T1 mix; T1 thresh; T1 uprSlope; T1 lwrSlope; T0 ogain; T0* outMagV; T0* outPhsV; }; typedef struct obj_str fobj_t; typedef struct obj_str dobj_t; template< typename T0, typename T1 > rc_t create( struct obj_str*& p, unsigned binN, T1 ceiling=30, T1 expo=2, T1 thresh=60, T1 uprSlope=0, T1 lwrSlope=2, T1 mix=0 ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str >(); p->ceiling = ceiling; p->expo = expo; p->thresh = thresh; p->uprSlope = uprSlope; p->lwrSlope = lwrSlope; p->mix = mix; p->ogain = 1; p->outMagV = mem::allocZ( binN ); p->outPhsV = mem::allocZ( binN ); return rc; } template< typename T0, typename T1 > rc_t destroy( struct obj_str*& p ) { rc_t rc = kOkRC; if( p != nullptr ) { mem::release(p->outMagV); mem::release(p->outPhsV); mem::release(p); } return rc; } template< typename T0, typename T1 > void _cmSpecDist2Bump( struct obj_str* p, double* x, unsigned binCnt, double thresh, double expo) { unsigned i = 0; double minDb = -100.0; thresh = -fabs(thresh); for(i=0; i thresh ) y = 1; else { y = (minDb - x[i])/(minDb - thresh); y += y - pow(y,expo); } x[i] = minDb + (-minDb) * y; } } template< typename T0, typename T1 > void _cmSpecDist2BasicMode( struct obj_str* p, double* X1m, unsigned binCnt, double thresh, double upr, double lwr ) { unsigned i=0; if( lwr < 0.3 ) lwr = 0.3; for(i=0; i 0 ) X1m[i] -= (lwr*d); else X1m[i] -= (upr*d); } } template< typename T0, typename T1 > rc_t exec( struct obj_str* p, const T0* magV, const T0* phsV, unsigned binN ) { rc_t rc = kOkRC; double X0m[binN]; double X1m[binN]; // take the mean of the the input magntitude spectrum double u0 = vop::mean(magV,binN); // convert magnitude to db (range=-1000.0 to 0.0) vop::ampl_to_db(X0m, magV, binN ); vop::copy(X1m,X0m,binN); // bump transform X0m _cmSpecDist2Bump(p,X0m, binN, p->ceiling, p->expo); // mix bump output with raw input: X1m = (X0m*mix) + (X1m*(1.0-mix)) vop::mul(X0m, p->mix, binN ); vop::mul(X1m, 1.0 - p->mix, binN ); vop::add(X1m, X0m, binN ); // basic transform _cmSpecDist2BasicMode(p,X1m,binN,p->thresh,p->uprSlope,p->lwrSlope); // convert db back to magnitude vop::db_to_ampl(X1m, X1m, binN ); // convert the mean input magnitude to db double idb = 20*log10(u0); // get the mean output magnitude spectra double u1 = vop::mean(X1m,binN); if( idb > -150.0 ) { // set the output gain such that the mean output magnitude // will match the mean input magnitude p->ogain = u0/u1; } else { T0 a0 = 0.9; p->ogain *= a0; } // apply the output gain vop::mul( p->outMagV, X1m, std::min((T1)4.0,p->ogain), binN); vop::copy( p->outPhsV, phsV, binN); return rc; } } //--------------------------------------------------------------------------------------------------------------------------------- // Data Recorder // namespace data_recorder { template< typename T > struct block_str { T* buf; // buf[frameN,sigN] struct block_str* link; // link to next block in chain }; template< typename T > struct obj_str { unsigned sigN; // count of channels per frame unsigned frameN; // count of frames per block struct block_str* head; // first block struct block_str* tail; // last block and the one being currrently filled unsigned frameIdx; // index into tail of frame to fill char* fn; char** colLabelA; unsigned colLabelN; bool enableFl; }; typedef struct obj_str fobj_t; typedef struct obj_str dobj_t; template< typename T > struct block_str* _block_alloc( struct obj_str* p ) { struct block_str* block = mem::allocZ< struct block_str >(); block->buf = mem::alloc( p->frameN * p->sigN ); if( p->head == nullptr ) p->head = block; if( p->tail != nullptr ) p->tail->link = block; p->tail = block; return block; } template< typename T > rc_t create( struct obj_str*& p, unsigned sigN, unsigned frameCacheN, const char* fn, const char** colLabelA, unsigned colLabelN, bool enableFl ) { rc_t rc = kOkRC; p = mem::allocZ< struct obj_str >(); p->frameN = frameCacheN; p->sigN = sigN; p->fn = mem::duplStr(fn); p->colLabelN = colLabelN; p->colLabelA = mem::allocZ< char* >( colLabelN ); p->enableFl = enableFl; for(unsigned i=0; icolLabelA[i] = mem::duplStr(colLabelA[i]); _block_alloc(p); return rc; } template< typename T > rc_t create( struct obj_str*& p, const object_t* cfg ) { rc_t rc = kOkRC; bool enableFl = true; unsigned sigN = 0; unsigned frameN = 0; const char* filename = nullptr; const object_t* colLabelL = nullptr; // parse the recorder spec if((rc = cfg->getv("enableFl", enableFl, "sigN", sigN, "frameN", frameN, "filename", filename, "colLabelL",colLabelL)) != kOkRC ) { rc = cwLogError(rc,"Record cfg. parse failed."); } else { unsigned labelN = colLabelL->child_count(); const char* labelL[ labelN ]; for(unsigned i=0; ichild_ele(i)->value( labelL[i] ); rc = create(p, sigN, frameN, filename, labelL, labelN, enableFl ); } return rc; } template< typename T> rc_t destroy( struct obj_str*& p ) { if( p != nullptr ) { if( p->enableFl && p->fn != nullptr && textLength(p->fn)!=0 ) write_as_csv(p,p->fn); for(unsigned i=0; icolLabelN; ++i) mem::release( p->colLabelA[i] ); struct block_str* b0 = p->head; while( b0 != nullptr ) { struct block_str* b1 = b0->link; mem::release(b0->buf); mem::release(b0); b0 = b1; } mem::release(p->fn); mem::release(p); } return kOkRC; } template< typename T> rc_t exec( struct obj_str* p, const T* xV, unsigned xN, unsigned chIdx=0, bool advance_fl = true ) { struct block_str* b = p->tail; if( chIdx + xN > p->sigN ) return cwLogError(kInvalidArgRC,"Channel index (%i) plus channel count (%i) is out of range of the allocated channe count (%i).", chIdx, xN, p->sigN ); if( p->enableFl ) { for(unsigned i=chIdx; i-chIdxframeIdx * p->sigN + i < p->frameN * p->sigN ); b->buf[ p->frameIdx * p->sigN + i ] = xV[i-chIdx]; } if( advance_fl ) advance(p); } return kOkRC; } template< typename T> rc_t advance( struct obj_str* p, unsigned frameN=1 ) { for(unsigned i=0; iframeIdx += 1; if( p->frameIdx >= p->frameN ) { _block_alloc(p); p->frameIdx = 0; } } return kOkRC; } template< typename T> rc_t write_as_csv( const struct obj_str* p, const char* fn ) { rc_t rc = kOkRC; file::handle_t h; struct block_str* b = p->head; if((rc = file::open(h,fn,file::kWriteFl)) != kOkRC ) { rc = cwLogError(kOpenFailRC,"Create failed on the data recorder output file '%s'.", cwStringNullGuard(fn)); goto errLabel; } for(; b!=nullptr; b=b->link) { unsigned frameN = b->link==NULL ? p->frameIdx : p->frameN; for(unsigned fi=0, rowIdx=0; ficolLabelN; ++ci) file::printf(h,"%s, ", p->colLabelA[ci] ); } else { for(unsigned ci=0; cisigN; ++ci) file::printf(h,"%f,", b->buf[ fi*p->sigN + ci ]); } file::print(h,"\n"); } } errLabel: file::close(h); return rc; } } rc_t test( const cw::object_t* args ); } // dsp } // cw #endif