From f0e82c0d4b99bd1df5e97949c8bef4ca8d19b79f Mon Sep 17 00:00:00 2001 From: vaclav Date: Mon, 25 Aug 2025 12:13:26 +0200 Subject: [PATCH] port CR --- lib_com/ivas_prot.h | 5 ++ lib_com/options.h | 1 + lib_dec/ivas_dirac_dec.c | 133 +++++++++++++++++++++++++++++++++++- lib_dec/ivas_spar_decoder.c | 17 +++++ lib_dec/lib_dec.c | 5 ++ 5 files changed, 160 insertions(+), 1 deletion(-) diff --git a/lib_com/ivas_prot.h b/lib_com/ivas_prot.h index b46381291..243f62e78 100644 --- a/lib_com/ivas_prot.h +++ b/lib_com/ivas_prot.h @@ -3747,8 +3747,13 @@ void ivas_dirac_dec_render_sf( Decoder_Struct *st_ivas, /* i/o: IVAS decoder structure */ float *output_f[], /* i/o: synthesized core-coder transport channels/DirAC output */ const int16_t nchan_transport, /* i : number of transport channels */ +#ifdef FIX_1319_STACK_SBA_DECODER + float *pppQMfFrame_ts_re[HOA3_CHANNELS][CLDFB_NO_COL_MAX], + float *pppQMfFrame_ts_im[HOA3_CHANNELS][CLDFB_NO_COL_MAX] +#else float *pppQMfFrame_ts_re[IVAS_MAX_FB_MIXER_IN_CH][CLDFB_NO_COL_MAX], float *pppQMfFrame_ts_im[IVAS_MAX_FB_MIXER_IN_CH][CLDFB_NO_COL_MAX] +#endif ); void computeDiffuseness_mdft( diff --git a/lib_com/options.h b/lib_com/options.h index d8e6dbc66..eadec6987 100644 --- a/lib_com/options.h +++ b/lib_com/options.h @@ -220,6 +220,7 @@ #define FIX_1369_HQ_LR_OVERFLOW /* FhG: fix BASOP overflow in hq_lr_enc(), brings floating-point code inline with FX */ #define NONBE_1118_EVS_LR_HQ_BITERROR /* VA: issue 1118: fix EVS decoder crash in LR-HQ in case of bit errors */ #define FIX_1139_REV_COLORATION_SHORT_T60 /* Nokia,FhG: Fix issue 1139, prevent sound coloration artefacts at very low reverberation times */ +#define FIX_1319_STACK_SBA_DECODER /* VA: issue 1319: Optimize the definition of buffer lengths in the SBA decoder */ /* #################### End BASOP porting switches ############################ */ diff --git a/lib_dec/ivas_dirac_dec.c b/lib_dec/ivas_dirac_dec.c index 8fc0712d5..a0c00758a 100644 --- a/lib_dec/ivas_dirac_dec.c +++ b/lib_dec/ivas_dirac_dec.c @@ -1639,6 +1639,82 @@ void ivas_dirac_dec_render( } +#ifdef FIX_1319_STACK_SBA_DECODER +/*------------------------------------------------------------------------- + * Local functions to perform binaural rendering with optimized stack + *------------------------------------------------------------------------*/ + +static void binRenderer_split( + BINAURAL_RENDERER_HANDLE hBinRenderer, /* i/o: binaural renderer handle */ + ISAR_DEC_SPLIT_REND_WRAPPER_HANDLE hSplitBinRend, /* i/o: ISAR split binaural rendering handle */ + COMBINED_ORIENTATION_HANDLE hCombinedOrientationData, /* i : combined head and external orientation handle */ + const int16_t numTimeSlots, /* i : number of time slots to render */ + float Cldfb_RealBuffer_Binaural[][BINAURAL_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX], /* o : Rotated Binaural signals */ + float Cldfb_ImagBuffer_Binaural[][BINAURAL_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX], /* o : Rotated Binaural signals */ + float RealBuffer[][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX], /* i : LS signals */ + float ImagBuffer[][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX], /* i : LS signals */ + const int16_t slot_idx_start, + const int16_t num_freq_bands, + const int16_t nchan_out ) +{ + int16_t pos_idx, slot_idx, ch; + float Cldfb_RealBuffer_Binaural_loc[MAX_HEAD_ROT_POSES][BINAURAL_CHANNELS][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; + float Cldfb_ImagBuffer_Binaural_loc[MAX_HEAD_ROT_POSES][BINAURAL_CHANNELS][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; + + /* Perform binaural rendering */ + ivas_binRenderer( hBinRenderer, &hSplitBinRend->splitrend.multiBinPoseData, hCombinedOrientationData, numTimeSlots, +#ifdef SPLIT_REND_WITH_HEAD_ROT_DEBUG + NULL, +#endif + Cldfb_RealBuffer_Binaural_loc, Cldfb_ImagBuffer_Binaural_loc, RealBuffer, ImagBuffer ); + + for ( slot_idx = 0; slot_idx < numTimeSlots; slot_idx++ ) + { + for ( ch = 0; ch < BINAURAL_CHANNELS; ch++ ) + { + mvr2r( Cldfb_RealBuffer_Binaural_loc[0][ch][slot_idx], Cldfb_RealBuffer_Binaural[0][ch][slot_idx], num_freq_bands ); + mvr2r( Cldfb_ImagBuffer_Binaural_loc[0][ch][slot_idx], Cldfb_ImagBuffer_Binaural[0][ch][slot_idx], num_freq_bands ); + } + } + + for ( pos_idx = 0; pos_idx < hBinRenderer->numPoses; pos_idx++ ) + { + for ( slot_idx = 0; slot_idx < numTimeSlots; slot_idx++ ) + { + for ( ch = 0; ch < nchan_out; ch++ ) + { + mvr2r( Cldfb_RealBuffer_Binaural_loc[pos_idx][ch][slot_idx], hSplitBinRend->hMultiBinCldfbData->Cldfb_RealBuffer_Binaural[( pos_idx * BINAURAL_CHANNELS ) + ch][slot_idx_start + slot_idx], num_freq_bands ); + mvr2r( Cldfb_ImagBuffer_Binaural_loc[pos_idx][ch][slot_idx], hSplitBinRend->hMultiBinCldfbData->Cldfb_ImagBuffer_Binaural[( pos_idx * BINAURAL_CHANNELS ) + ch][slot_idx_start + slot_idx], num_freq_bands ); + } + } + } + + return; +} + + +static void binRenderer( + BINAURAL_RENDERER_HANDLE hBinRenderer, /* i/o: binaural renderer handle */ + COMBINED_ORIENTATION_HANDLE hCombinedOrientationData, /* i : combined head and external orientation handle */ + const int16_t numTimeSlots, /* i : number of time slots to render */ + float Cldfb_RealBuffer_Binaural[][BINAURAL_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX], /* o : Rotated Binaural signals */ + float Cldfb_ImagBuffer_Binaural[][BINAURAL_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX], /* o : Rotated Binaural signals */ + float RealBuffer[][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX], /* i : LS signals */ + float ImagBuffer[][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX] /* i : LS signals */ +) +{ + /* Perform binaural rendering */ + ivas_binRenderer( hBinRenderer, NULL, hCombinedOrientationData, numTimeSlots, +#ifdef SPLIT_REND_WITH_HEAD_ROT_DEBUG + NULL, +#endif + Cldfb_RealBuffer_Binaural, Cldfb_ImagBuffer_Binaural, RealBuffer, ImagBuffer ); + + return; +} +#endif + + /*------------------------------------------------------------------------- * ivas_dirac_dec_render_sf() * @@ -1649,8 +1725,14 @@ void ivas_dirac_dec_render_sf( Decoder_Struct *st_ivas, /* i/o: IVAS decoder structure */ float *output_f[], /* i/o: synthesized core-coder transport channels/DirAC output */ const int16_t nchan_transport, /* i : number of transport channels */ +#ifdef FIX_1319_STACK_SBA_DECODER + float *pppQMfFrame_ts_re[HOA3_CHANNELS][CLDFB_NO_COL_MAX], + float *pppQMfFrame_ts_im[HOA3_CHANNELS][CLDFB_NO_COL_MAX] +#else float *pppQMfFrame_ts_re[IVAS_MAX_FB_MIXER_IN_CH][CLDFB_NO_COL_MAX], - float *pppQMfFrame_ts_im[IVAS_MAX_FB_MIXER_IN_CH][CLDFB_NO_COL_MAX] ) + float *pppQMfFrame_ts_im[IVAS_MAX_FB_MIXER_IN_CH][CLDFB_NO_COL_MAX] +#endif +) { int16_t i, ch, idx_in, idx_lfe; DIRAC_DEC_HANDLE hDirAC; @@ -1665,12 +1747,19 @@ void ivas_dirac_dec_render_sf( int16_t slot_idx_start, slot_idx_start_cldfb_synth, md_idx; /*CLDFB: last output channels reserved to LFT for CICPx*/ +#ifdef FIX_1319_STACK_SBA_DECODER + float Cldfb_RealBuffer[MAX_OUTPUT_CHANNELS][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; + float Cldfb_ImagBuffer[MAX_OUTPUT_CHANNELS][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; + float Cldfb_RealBuffer_Binaural[1][BINAURAL_CHANNELS][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; + float Cldfb_ImagBuffer_Binaural[1][BINAURAL_CHANNELS][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; +#else float Cldfb_RealBuffer[MAX_OUTPUT_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX]; float Cldfb_ImagBuffer[MAX_OUTPUT_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX]; float Cldfb_RealBuffer_Binaural[MAX_HEAD_ROT_POSES][BINAURAL_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX]; float Cldfb_ImagBuffer_Binaural[MAX_HEAD_ROT_POSES][BINAURAL_CHANNELS][MAX_PARAM_SPATIAL_SUBFRAMES][CLDFB_NO_CHANNELS_MAX]; float Cldfb_RealBuffer_Temp[2][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; float Cldfb_ImagBuffer_Temp[2][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX]; +#endif int16_t index, num_freq_bands; /* local copies of azi, ele, diffuseness */ @@ -1847,17 +1936,29 @@ void ivas_dirac_dec_render_sf( /* CLDFB Analysis*/ for ( ch = 0; ch < nchan_transport; ch++ ) { +#ifdef FIX_1319_STACK_SBA_DECODER + cldfbAnalysis_ts( &( st_ivas->hTcBuffer->tc[hDirACRend->sba_map_tc[ch]][hSpatParamRendCom->num_freq_bands * index_slot] ), + Cldfb_RealBuffer_Binaural[0][ch][slot_idx], /* note: it is a tmp. buffer at this point */ + Cldfb_ImagBuffer_Binaural[0][ch][slot_idx], /* note: it is a tmp. buffer at this point */ + hSpatParamRendCom->num_freq_bands, + st_ivas->cldfbAnaDec[ch] ); +#else cldfbAnalysis_ts( &( st_ivas->hTcBuffer->tc[hDirACRend->sba_map_tc[ch]][hSpatParamRendCom->num_freq_bands * index_slot] ), Cldfb_RealBuffer_Temp[ch][slot_idx], Cldfb_ImagBuffer_Temp[ch][slot_idx], hSpatParamRendCom->num_freq_bands, st_ivas->cldfbAnaDec[ch] ); +#endif } } if ( st_ivas->ism_mode != ISM_MASA_MODE_DISC && st_ivas->ism_mode != ISM_MASA_MODE_MASA_ONE_OBJ ) { +#ifdef FIX_1319_STACK_SBA_DECODER + ivas_omasa_preProcessStereoTransportsForMovedObjects( st_ivas, Cldfb_RealBuffer_Binaural[0], Cldfb_ImagBuffer_Binaural[0], hSpatParamRendCom->num_freq_bands, subframe_idx ); +#else ivas_omasa_preProcessStereoTransportsForMovedObjects( st_ivas, Cldfb_RealBuffer_Temp, Cldfb_ImagBuffer_Temp, hSpatParamRendCom->num_freq_bands, subframe_idx ); +#endif } } @@ -1884,8 +1985,13 @@ void ivas_dirac_dec_render_sf( { for ( ch = 0; ch < nchan_transport; ch++ ) { +#ifdef FIX_1319_STACK_SBA_DECODER + mvr2r( Cldfb_RealBuffer_Binaural[0][ch][slot_idx], Cldfb_RealBuffer[ch][0], hSpatParamRendCom->num_freq_bands ); + mvr2r( Cldfb_ImagBuffer_Binaural[0][ch][slot_idx], Cldfb_ImagBuffer[ch][0], hSpatParamRendCom->num_freq_bands ); +#else mvr2r( Cldfb_RealBuffer_Temp[ch][slot_idx], Cldfb_RealBuffer[ch][0], hSpatParamRendCom->num_freq_bands ); mvr2r( Cldfb_ImagBuffer_Temp[ch][slot_idx], Cldfb_ImagBuffer[ch][0], hSpatParamRendCom->num_freq_bands ); +#endif } } else @@ -2370,6 +2476,20 @@ void ivas_dirac_dec_render_sf( } } +#ifdef FIX_1319_STACK_SBA_DECODER + /* Perform binaural rendering */ + if ( st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED || st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM ) + { + binRenderer_split( st_ivas->hBinRenderer, st_ivas->hSplitBinRend, st_ivas->hCombinedOrientationData, hSpatParamRendCom->subframe_nbslots[subframe_idx], + Cldfb_RealBuffer_Binaural, Cldfb_ImagBuffer_Binaural, Cldfb_RealBuffer, Cldfb_ImagBuffer, slot_idx_start, hSpatParamRendCom->num_freq_bands, st_ivas->hDecoderConfig->nchan_out ); + } + else + { + binRenderer( st_ivas->hBinRenderer, st_ivas->hCombinedOrientationData, hSpatParamRendCom->subframe_nbslots[subframe_idx], + Cldfb_RealBuffer_Binaural, Cldfb_ImagBuffer_Binaural, Cldfb_RealBuffer, Cldfb_ImagBuffer ); + } + +#else /* Perform binaural rendering */ ivas_binRenderer( st_ivas->hBinRenderer, ( st_ivas->hSplitBinRend == NULL ) ? NULL : &st_ivas->hSplitBinRend->splitrend.multiBinPoseData, st_ivas->hCombinedOrientationData, hSpatParamRendCom->subframe_nbslots[subframe_idx], @@ -2392,13 +2512,19 @@ void ivas_dirac_dec_render_sf( } } } +#endif /* Inverse CLDFB*/ for ( ch = 0; ch < st_ivas->hDecoderConfig->nchan_out; ch++ ) { /* open CLDFB buffer up to CLDFB_NO_CHANNELS_MAX bands for 48kHz */ +#ifdef FIX_1319_STACK_SBA_DECODER + float *RealBuffer[CLDFB_SLOTS_PER_SUBFRAME]; + float *ImagBuffer[CLDFB_SLOTS_PER_SUBFRAME]; +#else float *RealBuffer[MAX_PARAM_SPATIAL_SUBFRAMES]; float *ImagBuffer[MAX_PARAM_SPATIAL_SUBFRAMES]; +#endif for ( i = 0; i < hSpatParamRendCom->subframe_nbslots[subframe_idx]; i++ ) { @@ -2422,8 +2548,13 @@ void ivas_dirac_dec_render_sf( } else { +#ifdef FIX_1319_STACK_SBA_DECODER + float *RealBuffer[CLDFB_SLOTS_PER_SUBFRAME]; + float *ImagBuffer[CLDFB_SLOTS_PER_SUBFRAME]; +#else float *RealBuffer[MAX_PARAM_SPATIAL_SUBFRAMES]; float *ImagBuffer[MAX_PARAM_SPATIAL_SUBFRAMES]; +#endif int16_t outchannels; idx_in = 0; diff --git a/lib_dec/ivas_spar_decoder.c b/lib_dec/ivas_spar_decoder.c index ae43ad1e5..7716ce401 100644 --- a/lib_dec/ivas_spar_decoder.c +++ b/lib_dec/ivas_spar_decoder.c @@ -1388,15 +1388,28 @@ void ivas_spar_dec_upmixer_sf( ) { int16_t cldfb_band, num_cldfb_bands, numch_in, numch_out; +#ifdef FIX_1319_STACK_SBA_DECODER + float *cldfb_in_ts_re[HOA3_CHANNELS][CLDFB_NO_COL_MAX]; + float *cldfb_in_ts_im[HOA3_CHANNELS][CLDFB_NO_COL_MAX]; +#else float *cldfb_in_ts_re[MAX_OUTPUT_CHANNELS + MAX_NUM_OBJECTS][CLDFB_NO_COL_MAX]; float *cldfb_in_ts_im[MAX_OUTPUT_CHANNELS + MAX_NUM_OBJECTS][CLDFB_NO_COL_MAX]; +#endif int16_t i, b, ts, out_ch, in_ch; int16_t num_spar_bands, spar_band, nchan_transport; int16_t num_in_ingest, split_band; int16_t slot_size, slot_idx_start; +#ifdef FIX_1319_STACK_SBA_DECODER + float *p_tc[HOA3_CHANNELS]; +#else float *p_tc[MAX_OUTPUT_CHANNELS + MAX_NUM_OBJECTS]; +#endif int16_t md_idx; +#ifdef FIX_1319_STACK_SBA_DECODER + float Pcm_tmp[HOA3_CHANNELS][2 /* Re, Im*/ * L_FRAME48k / MAX_PARAM_SPATIAL_SUBFRAMES]; +#else float Pcm_tmp[MAX_OUTPUT_CHANNELS + MAX_NUM_OBJECTS][L_FRAME48k]; +#endif int16_t numch_out_dirac; float mixer_mat[IVAS_SPAR_MAX_CH][IVAS_SPAR_MAX_CH][IVAS_MAX_NUM_BANDS]; int16_t b_skip_mat[IVAS_SPAR_MAX_CH][IVAS_SPAR_MAX_CH]; @@ -1511,7 +1524,11 @@ void ivas_spar_dec_upmixer_sf( if ( hDecoderConfig->output_config != IVAS_AUDIO_CONFIG_FOA ) { /* at this point, output channels are used as intermediate procesing buffers */ +#ifdef FIX_1319_STACK_SBA_DECODER + for ( in_ch = 0; in_ch < HOA3_CHANNELS; in_ch++ ) +#else for ( in_ch = 0; in_ch < MAX_OUTPUT_CHANNELS + MAX_NUM_OBJECTS; in_ch++ ) +#endif { for ( ts = 0; ts < MAX_PARAM_SPATIAL_SUBFRAMES; ts++ ) { diff --git a/lib_dec/lib_dec.c b/lib_dec/lib_dec.c index 868fc4e06..da8dd318a 100644 --- a/lib_dec/lib_dec.c +++ b/lib_dec/lib_dec.c @@ -4022,6 +4022,11 @@ static void ivas_destroy_handle_isar( { free( ( *hSplitBinRend )->hCldfbDataOut ); ( *hSplitBinRend )->hCldfbDataOut = NULL; +#ifdef FIX_1319_STACK_SBA_DECODER + + free( ( *hSplitBinRend ) ); + ( *hSplitBinRend ) = NULL; +#endif } } -- GitLab