diff --git a/Makefile b/Makefile index ea2480aa..ca753cae 100644 --- a/Makefile +++ b/Makefile @@ -240,9 +240,9 @@ UIDIR=$(MOUNT_DIR)/ui Q3UIDIR=$(MOUNT_DIR)/q3_ui JPDIR=$(MOUNT_DIR)/jpeg-8c SPEEXDIR=$(MOUNT_DIR)/libspeex -OGGDIR=$(MOUNT_DIR)/libogg-1.3.0 -OPUSDIR=$(MOUNT_DIR)/opus-1.0.2 -OPUSFILEDIR=$(MOUNT_DIR)/opusfile-0.2 +OGGDIR=$(MOUNT_DIR)/libogg-1.3.1 +OPUSDIR=$(MOUNT_DIR)/opus-1.1 +OPUSFILEDIR=$(MOUNT_DIR)/opusfile-0.5 ZDIR=$(MOUNT_DIR)/zlib Q3ASMDIR=$(MOUNT_DIR)/tools/asm LBURGDIR=$(MOUNT_DIR)/tools/lcc/lburg @@ -534,6 +534,10 @@ ifeq ($(PLATFORM),mingw32) endif endif + ifeq ($(CC),) + $(error Cannot find a suitable cross compiler for $(PLATFORM)) + endif + BASE_CFLAGS = -Wall -fno-strict-aliasing -Wimplicit -Wstrict-prototypes \ -DUSE_ICON @@ -1819,10 +1823,15 @@ endif ifeq ($(USE_CODEC_OPUS),1) ifeq ($(USE_INTERNAL_OPUS),1) Q3OBJ += \ + $(B)/client/opus/analysis.o \ + $(B)/client/opus/mlp.o \ + $(B)/client/opus/mlp_data.o \ $(B)/client/opus/opus.o \ $(B)/client/opus/opus_decoder.o \ $(B)/client/opus/opus_encoder.o \ $(B)/client/opus/opus_multistream.o \ + $(B)/client/opus/opus_multistream_encoder.o \ + $(B)/client/opus/opus_multistream_decoder.o \ $(B)/client/opus/repacketizer.o \ \ $(B)/client/opus/bands.o \ @@ -1837,6 +1846,8 @@ Q3OBJ += \ $(B)/client/opus/mdct.o \ $(B)/client/opus/modes.o \ $(B)/client/opus/pitch.o \ + $(B)/client/opus/celt_encoder.o \ + $(B)/client/opus/celt_decoder.o \ $(B)/client/opus/celt_lpc.o \ $(B)/client/opus/quant_bands.o \ $(B)/client/opus/rate.o \ @@ -1955,7 +1966,8 @@ Q3OBJ += \ $(B)/client/info.o \ $(B)/client/internal.o \ $(B)/client/opusfile.o \ - $(B)/client/stream.o + $(B)/client/stream.o \ + $(B)/client/wincerts.o endif endif @@ -2701,6 +2713,10 @@ ifneq ($(BUILD_CLIENT),0) ifneq ($(BUILD_RENDERER_OPENGL2),0) $(INSTALL) $(STRIP_FLAG) -m 0755 $(BR)/renderer_opengl2_$(SHLIBNAME) $(COPYBINDIR)/renderer_opengl2_$(SHLIBNAME) endif + else + ifneq ($(BUILD_RENDERER_OPENGL2),0) + $(INSTALL) $(STRIP_FLAG) -m 0755 $(BR)/$(CLIENTBIN)_opengl2$(FULLBINEXT) $(COPYBINDIR)/$(CLIENTBIN)_opengl2$(FULLBINEXT) + endif endif endif diff --git a/README.md b/README.md index e8cd6f0e..35b72fe3 100644 --- a/README.md +++ b/README.md @@ -783,4 +783,4 @@ Significant contributions from * Aaron Gyes -[![githalytics.com alpha](https://cruel-carlota.pagodabox.com/f88af076b5015c62b699968f6772c3a5 "githalytics.com")](http://githalytics.com/ioquake/ioq3) +[![githalytics.com alpha](https://cruel-carlota.pagodabox.com/6d196bd663b47049a25dcb8caef95949 "githalytics.com")](http://githalytics.com/ioquake/ioq3) diff --git a/code/client/cl_cin.c b/code/client/cl_cin.c index 45c4dd45..599443c9 100644 --- a/code/client/cl_cin.c +++ b/code/client/cl_cin.c @@ -1472,7 +1472,9 @@ int CIN_PlayCinematic( const char *arg, int x, int y, int w, int h, int systemBi Con_Close(); - s_rawend[0] = s_soundtime; + if (!cinTable[currentHandle].silent) { + s_rawend[0] = s_soundtime; + } return currentHandle; } diff --git a/code/client/cl_main.c b/code/client/cl_main.c index daf2ed58..7877a886 100644 --- a/code/client/cl_main.c +++ b/code/client/cl_main.c @@ -670,7 +670,7 @@ void CL_StopRecord_f( void ) { CL_DemoFilename ================== */ -void CL_DemoFilename( int number, char *fileName ) { +void CL_DemoFilename( int number, char *fileName, int fileNameSize ) { int a,b,c,d; if(number < 0 || number > 9999) @@ -684,7 +684,7 @@ void CL_DemoFilename( int number, char *fileName ) { number -= c*10; d = number; - Com_sprintf( fileName, MAX_OSPATH, "demo%i%i%i%i" + Com_sprintf( fileName, fileNameSize, "demo%i%i%i%i" , a, b, c, d ); } @@ -744,7 +744,7 @@ void CL_Record_f( void ) { // scan for a free demo name for ( number = 0 ; number <= 9999 ; number++ ) { - CL_DemoFilename( number, demoName ); + CL_DemoFilename( number, demoName, sizeof( demoName ) ); #ifdef LEGACY_PROTOCOL if(clc.compat) Com_sprintf(name, sizeof(name), "demos/%s.%s%d", demoName, DEMOEXT, com_legacyprotocol->integer); diff --git a/code/client/cl_parse.c b/code/client/cl_parse.c index a424aa59..5c9c1471 100644 --- a/code/client/cl_parse.c +++ b/code/client/cl_parse.c @@ -709,7 +709,7 @@ void CL_ParseVoip ( msg_t *msg ) { const int packetsize = MSG_ReadShort(msg); const int flags = MSG_ReadBits(msg, VOIP_FLAGCNT); char encoded[1024]; - int seqdiff = sequence - clc.voipIncomingSequence[sender]; + int seqdiff; int written = 0; int i; @@ -753,6 +753,8 @@ void CL_ParseVoip ( msg_t *msg ) { Com_DPrintf("VoIP: packet accepted!\n"); + seqdiff = sequence - clc.voipIncomingSequence[sender]; + // This is a new "generation" ... a new recording started, reset the bits. if (generation != clc.voipIncomingGeneration[sender]) { Com_DPrintf("VoIP: new generation %d!\n", generation); diff --git a/code/client/cl_scrn.c b/code/client/cl_scrn.c index 095dc8ad..831d5ef7 100644 --- a/code/client/cl_scrn.c +++ b/code/client/cl_scrn.c @@ -481,7 +481,7 @@ void SCR_DrawScreenField( stereoFrame_t stereoFrame ) { // wide aspect ratio screens need to have the sides cleared // unless they are displaying game renderings - if ( uiFullscreen || (clc.state != CA_ACTIVE && clc.state != CA_CINEMATIC) ) { + if ( uiFullscreen || clc.state < CA_LOADING ) { if ( cls.glconfig.vidWidth * 480 > cls.glconfig.vidHeight * 640 ) { re.SetColor( g_color_table[0] ); re.DrawStretchPic( 0, 0, cls.glconfig.vidWidth, cls.glconfig.vidHeight, 0, 0, 0, 0, cls.whiteShader ); diff --git a/code/client/snd_dma.c b/code/client/snd_dma.c index d623ec4c..e243f406 100644 --- a/code/client/snd_dma.c +++ b/code/client/snd_dma.c @@ -982,29 +982,34 @@ void S_Base_RawSamples( int stream, int samples, int rate, int width, int s_chan int i; int src, dst; float scale; - int intVolume; + int intVolumeLeft, intVolumeRight; portable_samplepair_t *rawsamples; if ( !s_soundStarted || s_soundMuted ) { return; } - if(entityNum >= 0) - { - // FIXME: support spatialized raw streams, e.g. for VoIP - return; - } - if ( (stream < 0) || (stream >= MAX_RAW_STREAMS) ) { return; } - + rawsamples = s_rawsamples[stream]; - if(s_muted->integer) - intVolume = 0; - else - intVolume = 256 * volume * s_volume->value; + if ( s_muted->integer ) { + intVolumeLeft = intVolumeRight = 0; + } else { + int leftvol, rightvol; + + if ( entityNum >= 0 && entityNum < MAX_GENTITIES ) { + // support spatialized raw streams, e.g. for VoIP + S_SpatializeOrigin( loopSounds[ entityNum ].origin, 256, &leftvol, &rightvol ); + } else { + leftvol = rightvol = 256; + } + + intVolumeLeft = leftvol * volume * s_volume->value; + intVolumeRight = rightvol * volume * s_volume->value; + } if ( s_rawend[stream] < s_soundtime ) { Com_DPrintf( "S_Base_RawSamples: resetting minimum: %i < %i\n", s_rawend[stream], s_soundtime ); @@ -1022,8 +1027,8 @@ void S_Base_RawSamples( int stream, int samples, int rate, int width, int s_chan { dst = s_rawend[stream]&(MAX_RAW_SAMPLES-1); s_rawend[stream]++; - rawsamples[dst].left = ((short *)data)[i*2] * intVolume; - rawsamples[dst].right = ((short *)data)[i*2+1] * intVolume; + rawsamples[dst].left = ((short *)data)[i*2] * intVolumeLeft; + rawsamples[dst].right = ((short *)data)[i*2+1] * intVolumeRight; } } else @@ -1035,8 +1040,8 @@ void S_Base_RawSamples( int stream, int samples, int rate, int width, int s_chan break; dst = s_rawend[stream]&(MAX_RAW_SAMPLES-1); s_rawend[stream]++; - rawsamples[dst].left = ((short *)data)[src*2] * intVolume; - rawsamples[dst].right = ((short *)data)[src*2+1] * intVolume; + rawsamples[dst].left = ((short *)data)[src*2] * intVolumeLeft; + rawsamples[dst].right = ((short *)data)[src*2+1] * intVolumeRight; } } } @@ -1049,13 +1054,14 @@ void S_Base_RawSamples( int stream, int samples, int rate, int width, int s_chan break; dst = s_rawend[stream]&(MAX_RAW_SAMPLES-1); s_rawend[stream]++; - rawsamples[dst].left = ((short *)data)[src] * intVolume; - rawsamples[dst].right = ((short *)data)[src] * intVolume; + rawsamples[dst].left = ((short *)data)[src] * intVolumeLeft; + rawsamples[dst].right = ((short *)data)[src] * intVolumeRight; } } else if (s_channels == 2 && width == 1) { - intVolume *= 256; + intVolumeLeft *= 256; + intVolumeRight *= 256; for (i=0 ; ; i++) { @@ -1064,13 +1070,14 @@ void S_Base_RawSamples( int stream, int samples, int rate, int width, int s_chan break; dst = s_rawend[stream]&(MAX_RAW_SAMPLES-1); s_rawend[stream]++; - rawsamples[dst].left = ((char *)data)[src*2] * intVolume; - rawsamples[dst].right = ((char *)data)[src*2+1] * intVolume; + rawsamples[dst].left = ((char *)data)[src*2] * intVolumeLeft; + rawsamples[dst].right = ((char *)data)[src*2+1] * intVolumeRight; } } else if (s_channels == 1 && width == 1) { - intVolume *= 256; + intVolumeLeft *= 256; + intVolumeRight *= 256; for (i=0 ; ; i++) { @@ -1079,8 +1086,8 @@ void S_Base_RawSamples( int stream, int samples, int rate, int width, int s_chan break; dst = s_rawend[stream]&(MAX_RAW_SAMPLES-1); s_rawend[stream]++; - rawsamples[dst].left = (((byte *)data)[src]-128) * intVolume; - rawsamples[dst].right = (((byte *)data)[src]-128) * intVolume; + rawsamples[dst].left = (((byte *)data)[src]-128) * intVolumeLeft; + rawsamples[dst].right = (((byte *)data)[src]-128) * intVolumeRight; } } @@ -1417,11 +1424,7 @@ void S_Base_StartBackgroundTrack( const char *intro, const char *loop ){ return; } - if( !loop ) { - s_backgroundLoop[0] = 0; - } else { - Q_strncpyz( s_backgroundLoop, loop, sizeof( s_backgroundLoop ) ); - } + Q_strncpyz( s_backgroundLoop, loop, sizeof( s_backgroundLoop ) ); S_OpenBackgroundStream( intro ); } diff --git a/code/client/snd_local.h b/code/client/snd_local.h index 66f3bcde..e0f5b1d0 100644 --- a/code/client/snd_local.h +++ b/code/client/snd_local.h @@ -56,6 +56,7 @@ typedef struct sfx_s { qboolean soundCompressed; // not in Memory int soundCompressionMethod; int soundLength; + int soundChannels; char soundName[MAX_QPATH]; int lastTimeUsed; struct sfx_s *next; diff --git a/code/client/snd_mem.c b/code/client/snd_mem.c index 1031f34c..655ae420 100644 --- a/code/client/snd_mem.c +++ b/code/client/snd_mem.c @@ -113,47 +113,51 @@ ResampleSfx resample / decimate to the current source rate ================ */ -static void ResampleSfx( sfx_t *sfx, int inrate, int inwidth, byte *data, qboolean compressed ) { +static int ResampleSfx( sfx_t *sfx, int channels, int inrate, int inwidth, int samples, byte *data, qboolean compressed ) { int outcount; int srcsample; float stepscale; - int i; + int i, j; int sample, samplefrac, fracstep; int part; sndBuffer *chunk; stepscale = (float)inrate / dma.speed; // this is usually 0.5, 1, or 2 - outcount = sfx->soundLength / stepscale; - sfx->soundLength = outcount; + outcount = samples / stepscale; samplefrac = 0; - fracstep = stepscale * 256; + fracstep = stepscale * 256 * channels; chunk = sfx->soundData; for (i=0 ; i> 8; samplefrac += fracstep; - if( inwidth == 2 ) { - sample = ( ((short *)data)[srcsample] ); - } else { - sample = (int)( (unsigned char)(data[srcsample]) - 128) << 8; - } - part = (i&(SND_CHUNK_SIZE-1)); - if (part == 0) { - sndBuffer *newchunk; - newchunk = SND_malloc(); - if (chunk == NULL) { - sfx->soundData = newchunk; + for (j=0 ; jnext = newchunk; + sample = (int)( (unsigned char)(data[srcsample+j]) - 128) << 8; + } + part = (i*channels+j)&(SND_CHUNK_SIZE-1); + if (part == 0) { + sndBuffer *newchunk; + newchunk = SND_malloc(); + if (chunk == NULL) { + sfx->soundData = newchunk; + } else { + chunk->next = newchunk; + } + chunk = newchunk; } - chunk = newchunk; - } - chunk->sndChunk[part] = sample; + chunk->sndChunk[part] = sample; + } } + + return outcount; } /* @@ -163,11 +167,11 @@ ResampleSfx resample / decimate to the current source rate ================ */ -static int ResampleSfxRaw( short *sfx, int inrate, int inwidth, int samples, byte *data ) { +static int ResampleSfxRaw( short *sfx, int channels, int inrate, int inwidth, int samples, byte *data ) { int outcount; int srcsample; float stepscale; - int i; + int i, j; int sample, samplefrac, fracstep; stepscale = (float)inrate / dma.speed; // this is usually 0.5, 1, or 2 @@ -175,18 +179,21 @@ static int ResampleSfxRaw( short *sfx, int inrate, int inwidth, int samples, byt outcount = samples / stepscale; samplefrac = 0; - fracstep = stepscale * 256; + fracstep = stepscale * 256 * channels; for (i=0 ; i> 8; samplefrac += fracstep; - if( inwidth == 2 ) { - sample = LittleShort ( ((short *)data)[srcsample] ); - } else { - sample = (int)( (unsigned char)(data[srcsample]) - 128) << 8; + for (j=0 ; jsoundName); } - samples = Hunk_AllocateTempMemory(info.samples * sizeof(short) * 2); + samples = Hunk_AllocateTempMemory(info.channels * info.samples * sizeof(short) * 2); sfx->lastTimeUsed = Com_Milliseconds()+1; @@ -231,29 +238,30 @@ qboolean S_LoadSound( sfx_t *sfx ) // manager to do the right thing for us and page // sound in as needed - if( sfx->soundCompressed == qtrue) { + if( info.channels == 1 && sfx->soundCompressed == qtrue) { sfx->soundCompressionMethod = 1; sfx->soundData = NULL; - sfx->soundLength = ResampleSfxRaw( samples, info.rate, info.width, info.samples, data + info.dataofs ); + sfx->soundLength = ResampleSfxRaw( samples, info.channels, info.rate, info.width, info.samples, data + info.dataofs ); S_AdpcmEncodeSound(sfx, samples); #if 0 - } else if (info.samples>(SND_CHUNK_SIZE*16) && info.width >1) { + } else if (info.channels == 1 && info.samples>(SND_CHUNK_SIZE*16) && info.width >1) { sfx->soundCompressionMethod = 3; sfx->soundData = NULL; - sfx->soundLength = ResampleSfxRaw( samples, info.rate, info.width, info.samples, (data + info.dataofs) ); + sfx->soundLength = ResampleSfxRaw( samples, info.channels, info.rate, info.width, info.samples, (data + info.dataofs) ); encodeMuLaw( sfx, samples); - } else if (info.samples>(SND_CHUNK_SIZE*6400) && info.width >1) { + } else if (info.channels == 1 && info.samples>(SND_CHUNK_SIZE*6400) && info.width >1) { sfx->soundCompressionMethod = 2; sfx->soundData = NULL; - sfx->soundLength = ResampleSfxRaw( samples, info.rate, info.width, info.samples, (data + info.dataofs) ); + sfx->soundLength = ResampleSfxRaw( samples, info.channels, info.rate, info.width, info.samples, (data + info.dataofs) ); encodeWavelet( sfx, samples); #endif } else { sfx->soundCompressionMethod = 0; - sfx->soundLength = info.samples; sfx->soundData = NULL; - ResampleSfx( sfx, info.rate, info.width, data + info.dataofs, qfalse ); + sfx->soundLength = ResampleSfx( sfx, info.channels, info.rate, info.width, info.samples, data + info.dataofs, qfalse ); } + + sfx->soundChannels = info.channels; Hunk_FreeTempMemory(samples); Hunk_FreeTempMemory(data); diff --git a/code/client/snd_mix.c b/code/client/snd_mix.c index fca5e15f..38bd0aae 100644 --- a/code/client/snd_mix.c +++ b/code/client/snd_mix.c @@ -234,7 +234,7 @@ static void S_PaintChannelFrom16_altivec( channel_t *ch, const sfx_t *sc, int co portable_samplepair_t *samp; sndBuffer *chunk; short *samples; - float ooff, fdata, fdiv, fleftvol, frightvol; + float ooff, fdata[2], fdiv, fleftvol, frightvol; samp = &paintbuffer[ bufferOffset ]; @@ -242,6 +242,14 @@ static void S_PaintChannelFrom16_altivec( channel_t *ch, const sfx_t *sc, int co sampleOffset = sampleOffset*ch->oldDopplerScale; } + if ( sc->soundChannels == 2 ) { + sampleOffset *= sc->soundChannels; + + if ( sampleOffset & 1 ) { + sampleOffset &= ~1; + } + } + chunk = sc->soundData; while (sampleOffset>=SND_CHUNK_SIZE) { chunk = chunk->next; @@ -274,6 +282,10 @@ static void S_PaintChannelFrom16_altivec( channel_t *ch, const sfx_t *sc, int co while(i < count && (((unsigned long)&samp[i] & 0x1f) || ((count-i) < 8) || ((SND_CHUNK_SIZE - sampleOffset) < 8))) { data = samples[sampleOffset++]; samp[i].left += (data * leftvol)>>8; + + if ( sc->soundChannels == 2 ) { + data = samples[sampleOffset++]; + } samp[i].right += (data * rightvol)>>8; if (sampleOffset == SND_CHUNK_SIZE) { @@ -373,10 +385,10 @@ static void S_PaintChannelFrom16_altivec( channel_t *ch, const sfx_t *sc, int co for ( i=0 ; idopplerScale; + ooff = ooff + ch->dopplerScale * sc->soundChannels; boff = ooff; - fdata = 0; - for (j=aoff; jsoundChannels) { if (j == SND_CHUNK_SIZE) { chunk = chunk->next; if (!chunk) { @@ -385,11 +397,17 @@ static void S_PaintChannelFrom16_altivec( channel_t *ch, const sfx_t *sc, int co samples = chunk->sndChunk; ooff -= SND_CHUNK_SIZE; } - fdata += samples[j&(SND_CHUNK_SIZE-1)]; + if ( sc->soundChannels == 2 ) { + fdata[0] += samples[j&(SND_CHUNK_SIZE-1)]; + fdata[1] += samples[(j+1)&(SND_CHUNK_SIZE-1)]; + } else { + fdata[0] += samples[j&(SND_CHUNK_SIZE-1)]; + fdata[1] += samples[j&(SND_CHUNK_SIZE-1)]; + } } - fdiv = 256 * (boff-aoff); - samp[i].left += (fdata * fleftvol)/fdiv; - samp[i].right += (fdata * frightvol)/fdiv; + fdiv = 256 * (boff-aoff) / sc->soundChannels; + samp[i].left += (fdata[0] * fleftvol)/fdiv; + samp[i].right += (fdata[1] * frightvol)/fdiv; } } } @@ -402,7 +420,7 @@ static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int cou portable_samplepair_t *samp; sndBuffer *chunk; short *samples; - float ooff, fdata, fdiv, fleftvol, frightvol; + float ooff, fdata[2], fdiv, fleftvol, frightvol; samp = &paintbuffer[ bufferOffset ]; @@ -410,6 +428,14 @@ static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int cou sampleOffset = sampleOffset*ch->oldDopplerScale; } + if ( sc->soundChannels == 2 ) { + sampleOffset *= sc->soundChannels; + + if ( sampleOffset & 1 ) { + sampleOffset &= ~1; + } + } + chunk = sc->soundData; while (sampleOffset>=SND_CHUNK_SIZE) { chunk = chunk->next; @@ -426,6 +452,10 @@ static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int cou for ( i=0 ; i>8; + + if ( sc->soundChannels == 2 ) { + data = samples[sampleOffset++]; + } samp[i].right += (data * rightvol)>>8; if (sampleOffset == SND_CHUNK_SIZE) { @@ -447,10 +477,10 @@ static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int cou for ( i=0 ; idopplerScale; + ooff = ooff + ch->dopplerScale * sc->soundChannels; boff = ooff; - fdata = 0; - for (j=aoff; jsoundChannels) { if (j == SND_CHUNK_SIZE) { chunk = chunk->next; if (!chunk) { @@ -459,11 +489,17 @@ static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int cou samples = chunk->sndChunk; ooff -= SND_CHUNK_SIZE; } - fdata += samples[j&(SND_CHUNK_SIZE-1)]; + if ( sc->soundChannels == 2 ) { + fdata[0] += samples[j&(SND_CHUNK_SIZE-1)]; + fdata[1] += samples[(j+1)&(SND_CHUNK_SIZE-1)]; + } else { + fdata[0] += samples[j&(SND_CHUNK_SIZE-1)]; + fdata[1] += samples[j&(SND_CHUNK_SIZE-1)]; + } } - fdiv = 256 * (boff-aoff); - samp[i].left += (fdata * fleftvol)/fdiv; - samp[i].right += (fdata * frightvol)/fdiv; + fdiv = 256 * (boff-aoff) / sc->soundChannels; + samp[i].left += (fdata[0] * fleftvol)/fdiv; + samp[i].right += (fdata[1] * frightvol)/fdiv; } } } diff --git a/code/client/snd_openal.c b/code/client/snd_openal.c index e2d19990..2ba07038 100644 --- a/code/client/snd_openal.c +++ b/code/client/snd_openal.c @@ -2521,7 +2521,9 @@ qboolean S_AL_Init( soundInterface_t *si ) if( !QAL_Init( s_alDriver->string ) ) { Com_Printf( "Failed to load library: \"%s\".\n", s_alDriver->string ); - return qfalse; + if( !Q_stricmp( s_alDriver->string, ALDRIVER_DEFAULT ) || !QAL_Init( ALDRIVER_DEFAULT ) ) { + return qfalse; + } } device = s_alDevice->string; @@ -2667,11 +2669,14 @@ qboolean S_AL_Init( soundInterface_t *si ) defaultinputdevice = qalcGetString(NULL, ALC_CAPTURE_DEFAULT_DEVICE_SPECIFIER); // dump a list of available devices to a cvar for the user to see. - while((curlen = strlen(inputdevicelist))) + if (inputdevicelist) { - Q_strcat(inputdevicenames, sizeof(inputdevicenames), inputdevicelist); - Q_strcat(inputdevicenames, sizeof(inputdevicenames), "\n"); - inputdevicelist += curlen + 1; + while((curlen = strlen(inputdevicelist))) + { + Q_strcat(inputdevicenames, sizeof(inputdevicenames), inputdevicelist); + Q_strcat(inputdevicenames, sizeof(inputdevicenames), "\n"); + inputdevicelist += curlen + 1; + } } s_alAvailableInputDevices = Cvar_Get("s_alAvailableInputDevices", inputdevicenames, CVAR_ROM | CVAR_NORESTART); @@ -2680,7 +2685,7 @@ qboolean S_AL_Init( soundInterface_t *si ) // !!! FIXME: should probably open the capture device after // !!! FIXME: initializing Speex so we can change to wideband // !!! FIXME: if we like. - Com_Printf("OpenAL default capture device is '%s'\n", defaultinputdevice); + Com_Printf("OpenAL default capture device is '%s'\n", defaultinputdevice ? defaultinputdevice : "none"); alCaptureDevice = qalcCaptureOpenDevice(inputdevice, 8000, AL_FORMAT_MONO16, 4096); if( !alCaptureDevice && inputdevice ) { diff --git a/code/libogg-1.3.0/include/ogg/config_types.h b/code/libogg-1.3.1/include/ogg/config_types.h similarity index 92% rename from code/libogg-1.3.0/include/ogg/config_types.h rename to code/libogg-1.3.1/include/ogg/config_types.h index e0dd8655..28288279 100644 --- a/code/libogg-1.3.0/include/ogg/config_types.h +++ b/code/libogg-1.3.1/include/ogg/config_types.h @@ -1,6 +1,7 @@ #ifndef __CONFIG_TYPES_H__ #define __CONFIG_TYPES_H__ +/* these are filled in by configure */ /* #define INCLUDE_INTTYPES_H 1 */ #define INCLUDE_STDINT_H 1 /* #define INCLUDE_SYS_TYPES_H 1 */ diff --git a/code/libogg-1.3.0/include/ogg/ogg.h b/code/libogg-1.3.1/include/ogg/ogg.h similarity index 100% rename from code/libogg-1.3.0/include/ogg/ogg.h rename to code/libogg-1.3.1/include/ogg/ogg.h diff --git a/code/libogg-1.3.0/include/ogg/os_types.h b/code/libogg-1.3.1/include/ogg/os_types.h similarity index 100% rename from code/libogg-1.3.0/include/ogg/os_types.h rename to code/libogg-1.3.1/include/ogg/os_types.h diff --git a/code/libogg-1.3.0/src/bitwise.c b/code/libogg-1.3.1/src/bitwise.c similarity index 100% rename from code/libogg-1.3.0/src/bitwise.c rename to code/libogg-1.3.1/src/bitwise.c diff --git a/code/libogg-1.3.0/src/framing.c b/code/libogg-1.3.1/src/framing.c similarity index 98% rename from code/libogg-1.3.0/src/framing.c rename to code/libogg-1.3.1/src/framing.c index 4452cbd5..3a2f0a60 100644 --- a/code/libogg-1.3.0/src/framing.c +++ b/code/libogg-1.3.1/src/framing.c @@ -12,7 +12,7 @@ function: code raw packets into framed OggSquish stream and decode Ogg streams back into raw packets - last mod: $Id: framing.c 18052 2011-08-04 17:57:02Z giles $ + last mod: $Id: framing.c 18758 2013-01-08 16:29:56Z tterribe $ note: The CRC code is directly derived from public domain code by Ross Williams (ross@guest.adelaide.edu.au). See docs/framing.html @@ -21,6 +21,7 @@ ********************************************************************/ #include +#include #include #include @@ -236,39 +237,51 @@ int ogg_stream_destroy(ogg_stream_state *os){ /* Helpers for ogg_stream_encode; this keeps the structure and what's happening fairly clear */ -static int _os_body_expand(ogg_stream_state *os,int needed){ - if(os->body_storage<=os->body_fill+needed){ +static int _os_body_expand(ogg_stream_state *os,long needed){ + if(os->body_storage-needed<=os->body_fill){ + long body_storage; void *ret; - ret=_ogg_realloc(os->body_data,(os->body_storage+needed+1024)* - sizeof(*os->body_data)); + if(os->body_storage>LONG_MAX-needed){ + ogg_stream_clear(os); + return -1; + } + body_storage=os->body_storage+needed; + if(body_storagebody_data,body_storage*sizeof(*os->body_data)); if(!ret){ ogg_stream_clear(os); return -1; } - os->body_storage+=(needed+1024); + os->body_storage=body_storage; os->body_data=ret; } return 0; } -static int _os_lacing_expand(ogg_stream_state *os,int needed){ - if(os->lacing_storage<=os->lacing_fill+needed){ +static int _os_lacing_expand(ogg_stream_state *os,long needed){ + if(os->lacing_storage-needed<=os->lacing_fill){ + long lacing_storage; void *ret; - ret=_ogg_realloc(os->lacing_vals,(os->lacing_storage+needed+32)* - sizeof(*os->lacing_vals)); + if(os->lacing_storage>LONG_MAX-needed){ + ogg_stream_clear(os); + return -1; + } + lacing_storage=os->lacing_storage+needed; + if(lacing_storagelacing_vals,lacing_storage*sizeof(*os->lacing_vals)); if(!ret){ ogg_stream_clear(os); return -1; } os->lacing_vals=ret; - ret=_ogg_realloc(os->granule_vals,(os->lacing_storage+needed+32)* + ret=_ogg_realloc(os->granule_vals,lacing_storage* sizeof(*os->granule_vals)); if(!ret){ ogg_stream_clear(os); return -1; } os->granule_vals=ret; - os->lacing_storage+=(needed+32); + os->lacing_storage=lacing_storage; } return 0; } @@ -304,12 +317,17 @@ void ogg_page_checksum_set(ogg_page *og){ int ogg_stream_iovecin(ogg_stream_state *os, ogg_iovec_t *iov, int count, long e_o_s, ogg_int64_t granulepos){ - int bytes = 0, lacing_vals, i; + long bytes = 0, lacing_vals; + int i; if(ogg_stream_check(os)) return -1; if(!iov) return 0; - for (i = 0; i < count; ++i) bytes += (int)iov[i].iov_len; + for (i = 0; i < count; ++i){ + if(iov[i].iov_len>LONG_MAX) return -1; + if(bytes>LONG_MAX-(long)iov[i].iov_len) return -1; + bytes += (long)iov[i].iov_len; + } lacing_vals=bytes/255+1; if(os->body_returned){ diff --git a/code/opus-1.0.2/celt/celt.c b/code/opus-1.0.2/celt/celt.c deleted file mode 100644 index 9bbe8524..00000000 --- a/code/opus-1.0.2/celt/celt.c +++ /dev/null @@ -1,2906 +0,0 @@ -/* Copyright (c) 2007-2008 CSIRO - Copyright (c) 2007-2010 Xiph.Org Foundation - Copyright (c) 2008 Gregory Maxwell - Written by Jean-Marc Valin and Gregory Maxwell */ -/* - Redistribution and use in source and binary forms, with or without - modification, are permitted provided that the following conditions - are met: - - - Redistributions of source code must retain the above copyright - notice, this list of conditions and the following disclaimer. - - - Redistributions in binary form must reproduce the above copyright - notice, this list of conditions and the following disclaimer in the - documentation and/or other materials provided with the distribution. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER - OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, - EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, - PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR - PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF - LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING - NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -*/ - -#ifdef HAVE_CONFIG_H -#include "config.h" -#endif - -#define CELT_C - -#include "os_support.h" -#include "mdct.h" -#include -#include "celt.h" -#include "pitch.h" -#include "bands.h" -#include "modes.h" -#include "entcode.h" -#include "quant_bands.h" -#include "rate.h" -#include "stack_alloc.h" -#include "mathops.h" -#include "float_cast.h" -#include -#include "celt_lpc.h" -#include "vq.h" - -#ifndef OPUS_VERSION -#define OPUS_VERSION "unknown" -#endif - -#ifdef CUSTOM_MODES -#define OPUS_CUSTOM_NOSTATIC -#else -#define OPUS_CUSTOM_NOSTATIC static inline -#endif - -static const unsigned char trim_icdf[11] = {126, 124, 119, 109, 87, 41, 19, 9, 4, 2, 0}; -/* Probs: NONE: 21.875%, LIGHT: 6.25%, NORMAL: 65.625%, AGGRESSIVE: 6.25% */ -static const unsigned char spread_icdf[4] = {25, 23, 2, 0}; - -static const unsigned char tapset_icdf[3]={2,1,0}; - -#ifdef CUSTOM_MODES -static const unsigned char toOpusTable[20] = { - 0xE0, 0xE8, 0xF0, 0xF8, - 0xC0, 0xC8, 0xD0, 0xD8, - 0xA0, 0xA8, 0xB0, 0xB8, - 0x00, 0x00, 0x00, 0x00, - 0x80, 0x88, 0x90, 0x98, -}; - -static const unsigned char fromOpusTable[16] = { - 0x80, 0x88, 0x90, 0x98, - 0x40, 0x48, 0x50, 0x58, - 0x20, 0x28, 0x30, 0x38, - 0x00, 0x08, 0x10, 0x18 -}; - -static inline int toOpus(unsigned char c) -{ - int ret=0; - if (c<0xA0) - ret = toOpusTable[c>>3]; - if (ret == 0) - return -1; - else - return ret|(c&0x7); -} - -static inline int fromOpus(unsigned char c) -{ - if (c<0x80) - return -1; - else - return fromOpusTable[(c>>3)-16] | (c&0x7); -} -#endif /* CUSTOM_MODES */ - -#define COMBFILTER_MAXPERIOD 1024 -#define COMBFILTER_MINPERIOD 15 - -static int resampling_factor(opus_int32 rate) -{ - int ret; - switch (rate) - { - case 48000: - ret = 1; - break; - case 24000: - ret = 2; - break; - case 16000: - ret = 3; - break; - case 12000: - ret = 4; - break; - case 8000: - ret = 6; - break; - default: -#ifndef CUSTOM_MODES - celt_assert(0); -#endif - ret = 0; - break; - } - return ret; -} - -/** Encoder state - @brief Encoder state - */ -struct OpusCustomEncoder { - const OpusCustomMode *mode; /**< Mode used by the encoder */ - int overlap; - int channels; - int stream_channels; - - int force_intra; - int clip; - int disable_pf; - int complexity; - int upsample; - int start, end; - - opus_int32 bitrate; - int vbr; - int signalling; - int constrained_vbr; /* If zero, VBR can do whatever it likes with the rate */ - int loss_rate; - int lsb_depth; - - /* Everything beyond this point gets cleared on a reset */ -#define ENCODER_RESET_START rng - - opus_uint32 rng; - int spread_decision; - opus_val32 delayedIntra; - int tonal_average; - int lastCodedBands; - int hf_average; - int tapset_decision; - - int prefilter_period; - opus_val16 prefilter_gain; - int prefilter_tapset; -#ifdef RESYNTH - int prefilter_period_old; - opus_val16 prefilter_gain_old; - int prefilter_tapset_old; -#endif - int consec_transient; - - opus_val32 preemph_memE[2]; - opus_val32 preemph_memD[2]; - - /* VBR-related parameters */ - opus_int32 vbr_reservoir; - opus_int32 vbr_drift; - opus_int32 vbr_offset; - opus_int32 vbr_count; - -#ifdef RESYNTH - celt_sig syn_mem[2][2*MAX_PERIOD]; -#endif - - celt_sig in_mem[1]; /* Size = channels*mode->overlap */ - /* celt_sig prefilter_mem[], Size = channels*COMBFILTER_MAXPERIOD */ - /* opus_val16 oldBandE[], Size = channels*mode->nbEBands */ - /* opus_val16 oldLogE[], Size = channels*mode->nbEBands */ - /* opus_val16 oldLogE2[], Size = channels*mode->nbEBands */ -#ifdef RESYNTH - /* opus_val16 overlap_mem[], Size = channels*overlap */ -#endif -}; - -int celt_encoder_get_size(int channels) -{ - CELTMode *mode = opus_custom_mode_create(48000, 960, NULL); - return opus_custom_encoder_get_size(mode, channels); -} - -OPUS_CUSTOM_NOSTATIC int opus_custom_encoder_get_size(const CELTMode *mode, int channels) -{ - int size = sizeof(struct CELTEncoder) - + (channels*mode->overlap-1)*sizeof(celt_sig) /* celt_sig in_mem[channels*mode->overlap]; */ - + channels*COMBFILTER_MAXPERIOD*sizeof(celt_sig) /* celt_sig prefilter_mem[channels*COMBFILTER_MAXPERIOD]; */ - + 3*channels*mode->nbEBands*sizeof(opus_val16); /* opus_val16 oldBandE[channels*mode->nbEBands]; */ - /* opus_val16 oldLogE[channels*mode->nbEBands]; */ - /* opus_val16 oldLogE2[channels*mode->nbEBands]; */ -#ifdef RESYNTH - size += channels*mode->overlap*sizeof(celt_sig); /* celt_sig overlap_mem[channels*mode->nbEBands]; */ -#endif - return size; -} - -#ifdef CUSTOM_MODES -CELTEncoder *opus_custom_encoder_create(const CELTMode *mode, int channels, int *error) -{ - int ret; - CELTEncoder *st = (CELTEncoder *)opus_alloc(opus_custom_encoder_get_size(mode, channels)); - /* init will handle the NULL case */ - ret = opus_custom_encoder_init(st, mode, channels); - if (ret != OPUS_OK) - { - opus_custom_encoder_destroy(st); - st = NULL; - } - if (error) - *error = ret; - return st; -} -#endif /* CUSTOM_MODES */ - -int celt_encoder_init(CELTEncoder *st, opus_int32 sampling_rate, int channels) -{ - int ret; - ret = opus_custom_encoder_init(st, opus_custom_mode_create(48000, 960, NULL), channels); - if (ret != OPUS_OK) - return ret; - st->upsample = resampling_factor(sampling_rate); - return OPUS_OK; -} - -OPUS_CUSTOM_NOSTATIC int opus_custom_encoder_init(CELTEncoder *st, const CELTMode *mode, int channels) -{ - if (channels < 0 || channels > 2) - return OPUS_BAD_ARG; - - if (st==NULL || mode==NULL) - return OPUS_ALLOC_FAIL; - - OPUS_CLEAR((char*)st, opus_custom_encoder_get_size(mode, channels)); - - st->mode = mode; - st->overlap = mode->overlap; - st->stream_channels = st->channels = channels; - - st->upsample = 1; - st->start = 0; - st->end = st->mode->effEBands; - st->signalling = 1; - - st->constrained_vbr = 1; - st->clip = 1; - - st->bitrate = OPUS_BITRATE_MAX; - st->vbr = 0; - st->force_intra = 0; - st->complexity = 5; - st->lsb_depth=24; - - opus_custom_encoder_ctl(st, OPUS_RESET_STATE); - - return OPUS_OK; -} - -#ifdef CUSTOM_MODES -void opus_custom_encoder_destroy(CELTEncoder *st) -{ - opus_free(st); -} -#endif /* CUSTOM_MODES */ - -static inline opus_val16 SIG2WORD16(celt_sig x) -{ -#ifdef FIXED_POINT - x = PSHR32(x, SIG_SHIFT); - x = MAX32(x, -32768); - x = MIN32(x, 32767); - return EXTRACT16(x); -#else - return (opus_val16)x; -#endif -} - -static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int C, - int overlap) -{ - int i; - VARDECL(opus_val16, tmp); - opus_val32 mem0=0,mem1=0; - int is_transient = 0; - int block; - int N; - VARDECL(opus_val16, bins); - SAVE_STACK; - ALLOC(tmp, len, opus_val16); - - block = overlap/2; - N=len/block; - ALLOC(bins, N, opus_val16); - if (C==1) - { - for (i=0;i=3) - is_transient=1; - conseq = 0; - for (j=i+1;j=7) - is_transient=1; - } - RESTORE_STACK; -#ifdef FUZZING - is_transient = rand()&0x1; -#endif - return is_transient; -} - -/** Apply window and compute the MDCT for all sub-frames and - all channels in a frame */ -static void compute_mdcts(const CELTMode *mode, int shortBlocks, celt_sig * OPUS_RESTRICT in, celt_sig * OPUS_RESTRICT out, int C, int LM) -{ - if (C==1 && !shortBlocks) - { - const int overlap = OVERLAP(mode); - clt_mdct_forward(&mode->mdct, in, out, mode->window, overlap, mode->maxLM-LM, 1); - } else { - const int overlap = OVERLAP(mode); - int N = mode->shortMdctSize<shortMdctSize; - B = shortBlocks; - } - c=0; do { - for (b=0;bmdct, in+c*(B*N+overlap)+b*N, &out[b+c*N*B], mode->window, overlap, shortBlocks ? mode->maxLM : mode->maxLM-LM, B); - } - } while (++cshortMdctSize<shortMdctSize; - B = shortBlocks; - } - /* Prevents problems from the imdct doing the overlap-add */ - OPUS_CLEAR(x, overlap); - - for (b=0;bmdct, &X[b+c*N2*B], x+N2*b, mode->window, overlap, shortBlocks ? mode->maxLM : mode->maxLM-LM, B); - } - - for (j=0;j>LM;j++) - L2 = MAC16_16(L2, tmp[(j<eBands[len]-m->eBands[len-1])<eBands[i+1]-m->eBands[i])<eBands[i]<eBands[i]<>LM); - best_L1 = L1; - /*printf ("%f ", L1);*/ - for (k=0;k>(LM-k), 1<<(LM-k)); - else - haar1(tmp, N>>k, 1<>LM); - - if (L1 < best_L1) - { - best_L1 = L1; - best_level = k+1; - } - } - /*printf ("%d ", isTransient ? LM-best_level : best_level);*/ - if (isTransient) - metric[i] = best_level; - else - metric[i] = -best_level; - *tf_sum += metric[i]; - } - /*printf("\n");*/ - /* NOTE: Future optimized implementations could detect extreme transients and set - tf_select = 1 but so far we have not found a reliable way of making this useful */ - tf_select = 0; - - cost0 = 0; - cost1 = isTransient ? 0 : lambda; - /* Viterbi forward pass */ - for (i=1;i=0;i--) - { - if (tf_res[i+1] == 1) - tf_res[i] = path1[i+1]; - else - tf_res[i] = path0[i+1]; - } - RESTORE_STACK; -#ifdef FUZZING - tf_select = rand()&0x1; - tf_res[0] = rand()&0x1; - for (i=1;istorage*8; - tell = ec_tell(enc); - logp = isTransient ? 2 : 4; - /* Reserve space to code the tf_select decision. */ - tf_select_rsv = LM>0 && tell+logp+1 <= budget; - budget -= tf_select_rsv; - curr = tf_changed = 0; - for (i=start;istorage*8; - tell = ec_tell(dec); - logp = isTransient ? 2 : 4; - tf_select_rsv = LM>0 && tell+logp+1<=budget; - budget -= tf_select_rsv; - tf_changed = curr = 0; - for (i=start;inbEBands;i++) - { - int N; - N=(m->eBands[i+1]-m->eBands[i])<cache.caps[m->nbEBands*(2*LM+C-1)+i]+64)*C*N>>2; - } -} - -static int alloc_trim_analysis(const CELTMode *m, const celt_norm *X, - const opus_val16 *bandLogE, int end, int LM, int C, int N0) -{ - int i; - opus_val32 diff=0; - int c; - int trim_index = 5; - if (C==2) - { - opus_val16 sum = 0; /* Q10 */ - /* Compute inter-channel correlation for low frequencies */ - for (i=0;i<8;i++) - { - int j; - opus_val32 partial = 0; - for (j=m->eBands[i]<eBands[i+1]< QCONST16(.995f,10)) - trim_index-=4; - else if (sum > QCONST16(.92f,10)) - trim_index-=3; - else if (sum > QCONST16(.85f,10)) - trim_index-=2; - else if (sum > QCONST16(.8f,10)) - trim_index-=1; - } - - /* Estimate spectral tilt */ - c=0; do { - for (i=0;inbEBands]*(opus_int32)(2+2*i-m->nbEBands); - } - } while (++c QCONST16(2.f, DB_SHIFT)) - trim_index--; - if (diff > QCONST16(8.f, DB_SHIFT)) - trim_index--; - if (diff < -QCONST16(4.f, DB_SHIFT)) - trim_index++; - if (diff < -QCONST16(10.f, DB_SHIFT)) - trim_index++; - - if (trim_index<0) - trim_index = 0; - if (trim_index>10) - trim_index = 10; -#ifdef FUZZING - trim_index = rand()%11; -#endif - return trim_index; -} - -static int stereo_analysis(const CELTMode *m, const celt_norm *X, - int LM, int N0) -{ - int i; - int thetas; - opus_val32 sumLR = EPSILON, sumMS = EPSILON; - - /* Use the L1 norm to model the entropy of the L/R signal vs the M/S signal */ - for (i=0;i<13;i++) - { - int j; - for (j=m->eBands[i]<eBands[i+1]<eBands[13]<<(LM+1))+thetas, sumMS) - > MULT16_32_Q15(m->eBands[13]<<(LM+1), sumLR); -} - -int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes, ec_enc *enc) -{ - int i, c, N; - opus_int32 bits; - ec_enc _enc; - VARDECL(celt_sig, in); - VARDECL(celt_sig, freq); - VARDECL(celt_norm, X); - VARDECL(celt_ener, bandE); - VARDECL(opus_val16, bandLogE); - VARDECL(int, fine_quant); - VARDECL(opus_val16, error); - VARDECL(int, pulses); - VARDECL(int, cap); - VARDECL(int, offsets); - VARDECL(int, fine_priority); - VARDECL(int, tf_res); - VARDECL(unsigned char, collapse_masks); - celt_sig *prefilter_mem; - opus_val16 *oldBandE, *oldLogE, *oldLogE2; - int shortBlocks=0; - int isTransient=0; - const int CC = st->channels; - const int C = st->stream_channels; - int LM, M; - int tf_select; - int nbFilledBytes, nbAvailableBytes; - int effEnd; - int codedBands; - int tf_sum; - int alloc_trim; - int pitch_index=COMBFILTER_MINPERIOD; - opus_val16 gain1 = 0; - int intensity=0; - int dual_stereo=0; - int effectiveBytes; - opus_val16 pf_threshold; - int dynalloc_logp; - opus_int32 vbr_rate; - opus_int32 total_bits; - opus_int32 total_boost; - opus_int32 balance; - opus_int32 tell; - int prefilter_tapset=0; - int pf_on; - int anti_collapse_rsv; - int anti_collapse_on=0; - int silence=0; - ALLOC_STACK; - - if (nbCompressedBytes<2 || pcm==NULL) - return OPUS_BAD_ARG; - - frame_size *= st->upsample; - for (LM=0;LM<=st->mode->maxLM;LM++) - if (st->mode->shortMdctSize<st->mode->maxLM) - return OPUS_BAD_ARG; - M=1<mode->shortMdctSize; - - prefilter_mem = st->in_mem+CC*(st->overlap); - oldBandE = (opus_val16*)(st->in_mem+CC*(st->overlap+COMBFILTER_MAXPERIOD)); - oldLogE = oldBandE + CC*st->mode->nbEBands; - oldLogE2 = oldLogE + CC*st->mode->nbEBands; - - if (enc==NULL) - { - tell=1; - nbFilledBytes=0; - } else { - tell=ec_tell(enc); - nbFilledBytes=(tell+4)>>3; - } - -#ifdef CUSTOM_MODES - if (st->signalling && enc==NULL) - { - int tmp = (st->mode->effEBands-st->end)>>1; - st->end = IMAX(1, st->mode->effEBands-tmp); - compressed[0] = tmp<<5; - compressed[0] |= LM<<3; - compressed[0] |= (C==2)<<2; - /* Convert "standard mode" to Opus header */ - if (st->mode->Fs==48000 && st->mode->shortMdctSize==120) - { - int c0 = toOpus(compressed[0]); - if (c0<0) - return OPUS_BAD_ARG; - compressed[0] = c0; - } - compressed++; - nbCompressedBytes--; - } -#else - celt_assert(st->signalling==0); -#endif - - /* Can't produce more than 1275 output bytes */ - nbCompressedBytes = IMIN(nbCompressedBytes,1275); - nbAvailableBytes = nbCompressedBytes - nbFilledBytes; - - if (st->vbr && st->bitrate!=OPUS_BITRATE_MAX) - { - opus_int32 den=st->mode->Fs>>BITRES; - vbr_rate=(st->bitrate*frame_size+(den>>1))/den; -#ifdef CUSTOM_MODES - if (st->signalling) - vbr_rate -= 8<>(3+BITRES); - } else { - opus_int32 tmp; - vbr_rate = 0; - tmp = st->bitrate*frame_size; - if (tell>1) - tmp += tell; - if (st->bitrate!=OPUS_BITRATE_MAX) - nbCompressedBytes = IMAX(2, IMIN(nbCompressedBytes, - (tmp+4*st->mode->Fs)/(8*st->mode->Fs)-!!st->signalling)); - effectiveBytes = nbCompressedBytes; - } - - if (enc==NULL) - { - ec_enc_init(&_enc, compressed, nbCompressedBytes); - enc = &_enc; - } - - if (vbr_rate>0) - { - /* Computes the max bit-rate allowed in VBR mode to avoid violating the - target rate and buffering. - We must do this up front so that bust-prevention logic triggers - correctly if we don't have enough bits. */ - if (st->constrained_vbr) - { - opus_int32 vbr_bound; - opus_int32 max_allowed; - /* We could use any multiple of vbr_rate as bound (depending on the - delay). - This is clamped to ensure we use at least two bytes if the encoder - was entirely empty, but to allow 0 in hybrid mode. */ - vbr_bound = vbr_rate; - max_allowed = IMIN(IMAX(tell==1?2:0, - (vbr_rate+vbr_bound-st->vbr_reservoir)>>(BITRES+3)), - nbAvailableBytes); - if(max_allowed < nbAvailableBytes) - { - nbCompressedBytes = nbFilledBytes+max_allowed; - nbAvailableBytes = max_allowed; - ec_enc_shrink(enc, nbCompressedBytes); - } - } - } - total_bits = nbCompressedBytes*8; - - effEnd = st->end; - if (effEnd > st->mode->effEBands) - effEnd = st->mode->effEBands; - - ALLOC(in, CC*(N+st->overlap), celt_sig); - - /* Find pitch period and gain */ - { - VARDECL(celt_sig, _pre); - celt_sig *pre[2]; - SAVE_STACK; - ALLOC(_pre, CC*(N+COMBFILTER_MAXPERIOD), celt_sig); - - pre[0] = _pre; - pre[1] = _pre + (N+COMBFILTER_MAXPERIOD); - - silence = 1; - c=0; do { - int count = 0; - const opus_val16 * OPUS_RESTRICT pcmp = pcm+c; - celt_sig * OPUS_RESTRICT inp = in+c*(N+st->overlap)+st->overlap; - - for (i=0;iclip) - x = MAX32(-65536.f, MIN32(65536.f,x)); -#endif - if (++count==st->upsample) - { - count=0; - pcmp+=CC; - } else { - x = 0; - } - /* Apply pre-emphasis */ - tmp = MULT16_16(st->mode->preemph[2], x); - *inp = tmp + st->preemph_memE[c]; - st->preemph_memE[c] = MULT16_32_Q15(st->mode->preemph[1], *inp) - - MULT16_32_Q15(st->mode->preemph[0], tmp); - silence = silence && *inp == 0; - inp++; - } - OPUS_COPY(pre[c], prefilter_mem+c*COMBFILTER_MAXPERIOD, COMBFILTER_MAXPERIOD); - OPUS_COPY(pre[c]+COMBFILTER_MAXPERIOD, in+c*(N+st->overlap)+st->overlap, N); - } while (++c0) - { - effectiveBytes=nbCompressedBytes=IMIN(nbCompressedBytes, nbFilledBytes+2); - total_bits=nbCompressedBytes*8; - nbAvailableBytes=2; - ec_enc_shrink(enc, nbCompressedBytes); - } - /* Pretend we've filled all the remaining bits with zeros - (that's what the initialiser did anyway) */ - tell = nbCompressedBytes*8; - enc->nbits_total+=tell-ec_tell(enc); - } - if (nbAvailableBytes>12*C && st->start==0 && !silence && !st->disable_pf && st->complexity >= 5) - { - VARDECL(opus_val16, pitch_buf); - ALLOC(pitch_buf, (COMBFILTER_MAXPERIOD+N)>>1, opus_val16); - - pitch_downsample(pre, pitch_buf, COMBFILTER_MAXPERIOD+N, CC); - pitch_search(pitch_buf+(COMBFILTER_MAXPERIOD>>1), pitch_buf, N, - COMBFILTER_MAXPERIOD-COMBFILTER_MINPERIOD, &pitch_index); - pitch_index = COMBFILTER_MAXPERIOD-pitch_index; - - gain1 = remove_doubling(pitch_buf, COMBFILTER_MAXPERIOD, COMBFILTER_MINPERIOD, - N, &pitch_index, st->prefilter_period, st->prefilter_gain); - if (pitch_index > COMBFILTER_MAXPERIOD-2) - pitch_index = COMBFILTER_MAXPERIOD-2; - gain1 = MULT16_16_Q15(QCONST16(.7f,15),gain1); - if (st->loss_rate>2) - gain1 = HALF32(gain1); - if (st->loss_rate>4) - gain1 = HALF32(gain1); - if (st->loss_rate>8) - gain1 = 0; - prefilter_tapset = st->tapset_decision; - } else { - gain1 = 0; - } - - /* Gain threshold for enabling the prefilter/postfilter */ - pf_threshold = QCONST16(.2f,15); - - /* Adjusting the threshold based on rate and continuity */ - if (abs(pitch_index-st->prefilter_period)*10>pitch_index) - pf_threshold += QCONST16(.2f,15); - if (nbAvailableBytes<25) - pf_threshold += QCONST16(.1f,15); - if (nbAvailableBytes<35) - pf_threshold += QCONST16(.1f,15); - if (st->prefilter_gain > QCONST16(.4f,15)) - pf_threshold -= QCONST16(.1f,15); - if (st->prefilter_gain > QCONST16(.55f,15)) - pf_threshold -= QCONST16(.1f,15); - - /* Hard threshold at 0.2 */ - pf_threshold = MAX16(pf_threshold, QCONST16(.2f,15)); - if (gain1start==0 && tell+16<=total_bits) - ec_enc_bit_logp(enc, 0, 1); - gain1 = 0; - pf_on = 0; - } else { - /*This block is not gated by a total bits check only because - of the nbAvailableBytes check above.*/ - int qg; - int octave; - - if (ABS16(gain1-st->prefilter_gain)prefilter_gain; - -#ifdef FIXED_POINT - qg = ((gain1+1536)>>10)/3-1; -#else - qg = (int)floor(.5f+gain1*32/3)-1; -#endif - qg = IMAX(0, IMIN(7, qg)); - ec_enc_bit_logp(enc, 1, 1); - pitch_index += 1; - octave = EC_ILOG(pitch_index)-5; - ec_enc_uint(enc, octave, 6); - ec_enc_bits(enc, pitch_index-(16<mode->shortMdctSize-st->mode->overlap; - st->prefilter_period=IMAX(st->prefilter_period, COMBFILTER_MINPERIOD); - OPUS_COPY(in+c*(N+st->overlap), st->in_mem+c*(st->overlap), st->overlap); - if (offset) - comb_filter(in+c*(N+st->overlap)+st->overlap, pre[c]+COMBFILTER_MAXPERIOD, - st->prefilter_period, st->prefilter_period, offset, -st->prefilter_gain, -st->prefilter_gain, - st->prefilter_tapset, st->prefilter_tapset, NULL, 0); - - comb_filter(in+c*(N+st->overlap)+st->overlap+offset, pre[c]+COMBFILTER_MAXPERIOD+offset, - st->prefilter_period, pitch_index, N-offset, -st->prefilter_gain, -gain1, - st->prefilter_tapset, prefilter_tapset, st->mode->window, st->mode->overlap); - OPUS_COPY(st->in_mem+c*(st->overlap), in+c*(N+st->overlap)+N, st->overlap); - - if (N>COMBFILTER_MAXPERIOD) - { - OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, pre[c]+N, COMBFILTER_MAXPERIOD); - } else { - OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, prefilter_mem+c*COMBFILTER_MAXPERIOD+N, COMBFILTER_MAXPERIOD-N); - OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD+COMBFILTER_MAXPERIOD-N, pre[c]+COMBFILTER_MAXPERIOD, N); - } - } while (++c0 && ec_tell(enc)+3<=total_bits) - { - if (st->complexity > 1) - { - isTransient = transient_analysis(in, N+st->overlap, CC, - st->overlap); - if (isTransient) - shortBlocks = M; - } - ec_enc_bit_logp(enc, isTransient, 3); - } - - ALLOC(freq, CC*N, celt_sig); /**< Interleaved signal MDCTs */ - ALLOC(bandE,st->mode->nbEBands*CC, celt_ener); - ALLOC(bandLogE,st->mode->nbEBands*CC, opus_val16); - /* Compute MDCTs */ - compute_mdcts(st->mode, shortBlocks, in, freq, CC, LM); - - if (CC==2&&C==1) - { - for (i=0;iupsample != 1) - { - c=0; do - { - int bound = N/st->upsample; - for (i=0;iupsample; - for (;imode, freq, bandE, effEnd, C, M); - - amp2Log2(st->mode, effEnd, st->end, bandE, bandLogE, C); - - /* Band normalisation */ - normalise_bands(st->mode, freq, X, bandE, effEnd, C, M); - - ALLOC(tf_res, st->mode->nbEBands, int); - tf_select = tf_analysis(st->mode, effEnd, C, isTransient, tf_res, effectiveBytes, X, N, LM, st->start, &tf_sum); - for (i=effEnd;iend;i++) - tf_res[i] = tf_res[effEnd-1]; - - ALLOC(error, C*st->mode->nbEBands, opus_val16); - quant_coarse_energy(st->mode, st->start, st->end, effEnd, bandLogE, - oldBandE, total_bits, error, enc, - C, LM, nbAvailableBytes, st->force_intra, - &st->delayedIntra, st->complexity >= 4, st->loss_rate); - - tf_encode(st->start, st->end, isTransient, tf_res, LM, tf_select, enc); - - if (ec_tell(enc)+4<=total_bits) - { - if (shortBlocks || st->complexity < 3 - || nbAvailableBytes < 10*C || st->start!=0) - { - if (st->complexity == 0) - st->spread_decision = SPREAD_NONE; - else - st->spread_decision = SPREAD_NORMAL; - } else { - st->spread_decision = spreading_decision(st->mode, X, - &st->tonal_average, st->spread_decision, &st->hf_average, - &st->tapset_decision, pf_on&&!shortBlocks, effEnd, C, M); - } - ec_enc_icdf(enc, st->spread_decision, spread_icdf, 5); - } - - ALLOC(cap, st->mode->nbEBands, int); - ALLOC(offsets, st->mode->nbEBands, int); - - init_caps(st->mode,cap,LM,C); - for (i=0;imode->nbEBands;i++) - offsets[i] = 0; - /* Dynamic allocation code */ - /* Make sure that dynamic allocation can't make us bust the budget */ - if (effectiveBytes > 50 && LM>=1) - { - int t1, t2; - if (LM <= 1) - { - t1 = 3; - t2 = 5; - } else { - t1 = 2; - t2 = 4; - } - for (i=st->start+1;iend-1;i++) - { - opus_val32 d2; - d2 = 2*bandLogE[i]-bandLogE[i-1]-bandLogE[i+1]; - if (C==2) - d2 = HALF32(d2 + 2*bandLogE[i+st->mode->nbEBands]- - bandLogE[i-1+st->mode->nbEBands]-bandLogE[i+1+st->mode->nbEBands]); -#ifdef FUZZING - if((rand()&0xF)==0) - { - offsets[i] += 1; - if((rand()&0x3)==0) - offsets[i] += 1+(rand()&0x3); - } -#else - if (d2 > SHL16(t1,DB_SHIFT)) - offsets[i] += 1; - if (d2 > SHL16(t2,DB_SHIFT)) - offsets[i] += 1; -#endif - } - } - dynalloc_logp = 6; - total_bits<<=BITRES; - total_boost = 0; - tell = ec_tell_frac(enc); - for (i=st->start;iend;i++) - { - int width, quanta; - int dynalloc_loop_logp; - int boost; - int j; - width = C*(st->mode->eBands[i+1]-st->mode->eBands[i])<mode, X, bandLogE, - st->end, LM, C, N); - ec_enc_icdf(enc, alloc_trim, trim_icdf, 7); - tell = ec_tell_frac(enc); - } - - /* Variable bitrate */ - if (vbr_rate>0) - { - opus_val16 alpha; - opus_int32 delta; - /* The target rate in 8th bits per frame */ - opus_int32 target; - opus_int32 min_allowed; - int lm_diff = st->mode->maxLM - LM; - - /* Don't attempt to use more than 510 kb/s, even for frames smaller than 20 ms. - The CELT allocator will just not be able to use more than that anyway. */ - nbCompressedBytes = IMIN(nbCompressedBytes,1275>>(3-LM)); - target = vbr_rate + (st->vbr_offset>>lm_diff) - ((40*C+20)<end-st->start)) - target = 7*target/4; - else if (tf_sum < -(st->end-st->start)) - target = 3*target/2; - else if (M > 1) - target-=(target+14)/28; - - /* The current offset is removed from the target and the space used - so far is added*/ - target=target+tell; - - /* In VBR mode the frame size must not be reduced so much that it would - result in the encoder running out of bits. - The margin of 2 bytes ensures that none of the bust-prevention logic - in the decoder will have triggered so far. */ - min_allowed = ((tell+total_boost+(1<<(BITRES+3))-1)>>(BITRES+3)) + 2 - nbFilledBytes; - - nbAvailableBytes = (target+(1<<(BITRES+2)))>>(BITRES+3); - nbAvailableBytes = IMAX(min_allowed,nbAvailableBytes); - nbAvailableBytes = IMIN(nbCompressedBytes,nbAvailableBytes+nbFilledBytes) - nbFilledBytes; - - /* By how much did we "miss" the target on that frame */ - delta = target - vbr_rate; - - target=nbAvailableBytes<<(BITRES+3); - - /*If the frame is silent we don't adjust our drift, otherwise - the encoder will shoot to very high rates after hitting a - span of silence, but we do allow the bitres to refill. - This means that we'll undershoot our target in CVBR/VBR modes - on files with lots of silence. */ - if(silence) - { - nbAvailableBytes = 2; - target = 2*8<vbr_count < 970) - { - st->vbr_count++; - alpha = celt_rcp(SHL32(EXTEND32(st->vbr_count+20),16)); - } else - alpha = QCONST16(.001f,15); - /* How many bits have we used in excess of what we're allowed */ - if (st->constrained_vbr) - st->vbr_reservoir += target - vbr_rate; - /*printf ("%d\n", st->vbr_reservoir);*/ - - /* Compute the offset we need to apply in order to reach the target */ - st->vbr_drift += (opus_int32)MULT16_32_Q15(alpha,(delta*(1<vbr_offset-st->vbr_drift); - st->vbr_offset = -st->vbr_drift; - /*printf ("%d\n", st->vbr_drift);*/ - - if (st->constrained_vbr && st->vbr_reservoir < 0) - { - /* We're under the min value -- increase rate */ - int adjust = (-st->vbr_reservoir)/(8<vbr_reservoir = 0; - /*printf ("+%d\n", adjust);*/ - } - nbCompressedBytes = IMIN(nbCompressedBytes,nbAvailableBytes+nbFilledBytes); - /* This moves the raw bits to take into account the new compressed size */ - ec_enc_shrink(enc, nbCompressedBytes); - } - if (C==2) - { - int effectiveRate; - - /* Always use MS for 2.5 ms frames until we can do a better analysis */ - if (LM!=0) - dual_stereo = stereo_analysis(st->mode, X, LM, N); - - /* Account for coarse energy */ - effectiveRate = (8*effectiveBytes - 80)>>LM; - - /* effectiveRate in kb/s */ - effectiveRate = 2*effectiveRate/5; - if (effectiveRate<35) - intensity = 8; - else if (effectiveRate<50) - intensity = 12; - else if (effectiveRate<68) - intensity = 16; - else if (effectiveRate<84) - intensity = 18; - else if (effectiveRate<102) - intensity = 19; - else if (effectiveRate<130) - intensity = 20; - else - intensity = 100; - intensity = IMIN(st->end,IMAX(st->start, intensity)); - } - - /* Bit allocation */ - ALLOC(fine_quant, st->mode->nbEBands, int); - ALLOC(pulses, st->mode->nbEBands, int); - ALLOC(fine_priority, st->mode->nbEBands, int); - - /* bits = packet size - where we are - safety*/ - bits = (((opus_int32)nbCompressedBytes*8)<=2&&bits>=((LM+2)<mode, st->start, st->end, offsets, cap, - alloc_trim, &intensity, &dual_stereo, bits, &balance, pulses, - fine_quant, fine_priority, C, LM, enc, 1, st->lastCodedBands); - st->lastCodedBands = codedBands; - - quant_fine_energy(st->mode, st->start, st->end, oldBandE, error, fine_quant, enc, C); - -#ifdef MEASURE_NORM_MSE - float X0[3000]; - float bandE0[60]; - c=0; do - for (i=0;imode->nbEBands;i++) - bandE0[i] = bandE[i]; -#endif - - /* Residual quantisation */ - ALLOC(collapse_masks, C*st->mode->nbEBands, unsigned char); - quant_all_bands(1, st->mode, st->start, st->end, X, C==2 ? X+N : NULL, collapse_masks, - bandE, pulses, shortBlocks, st->spread_decision, dual_stereo, intensity, tf_res, - nbCompressedBytes*(8<rng); - - if (anti_collapse_rsv > 0) - { - anti_collapse_on = st->consec_transient<2; -#ifdef FUZZING - anti_collapse_on = rand()&0x1; -#endif - ec_enc_bits(enc, anti_collapse_on, 1); - } - quant_energy_finalise(st->mode, st->start, st->end, oldBandE, error, fine_quant, fine_priority, nbCompressedBytes*8-ec_tell(enc), enc, C); - - if (silence) - { - for (i=0;imode->nbEBands;i++) - oldBandE[i] = -QCONST16(28.f,DB_SHIFT); - } - -#ifdef RESYNTH - /* Re-synthesis of the coded audio if required */ - { - celt_sig *out_mem[2]; - celt_sig *overlap_mem[2]; - - log2Amp(st->mode, st->start, st->end, bandE, oldBandE, C); - if (silence) - { - for (i=0;imode->nbEBands;i++) - bandE[i] = 0; - } - -#ifdef MEASURE_NORM_MSE - measure_norm_mse(st->mode, X, X0, bandE, bandE0, M, N, C); -#endif - if (anti_collapse_on) - { - anti_collapse(st->mode, X, collapse_masks, LM, C, N, - st->start, st->end, oldBandE, oldLogE, oldLogE2, pulses, st->rng); - } - - /* Synthesis */ - denormalise_bands(st->mode, X, freq, bandE, effEnd, C, M); - - OPUS_MOVE(st->syn_mem[0], st->syn_mem[0]+N, MAX_PERIOD); - if (CC==2) - OPUS_MOVE(st->syn_mem[1], st->syn_mem[1]+N, MAX_PERIOD); - - c=0; do - for (i=0;imode->eBands[st->start];i++) - freq[c*N+i] = 0; - while (++cmode->eBands[st->end];isyn_mem[0]+MAX_PERIOD; - if (CC==2) - out_mem[1] = st->syn_mem[1]+MAX_PERIOD; - - overlap_mem[0] = (celt_sig*)(oldLogE2 + CC*st->mode->nbEBands); - if (CC==2) - overlap_mem[1] = overlap_mem[0] + st->overlap; - - compute_inv_mdcts(st->mode, shortBlocks, freq, out_mem, overlap_mem, CC, LM); - - c=0; do { - st->prefilter_period=IMAX(st->prefilter_period, COMBFILTER_MINPERIOD); - st->prefilter_period_old=IMAX(st->prefilter_period_old, COMBFILTER_MINPERIOD); - comb_filter(out_mem[c], out_mem[c], st->prefilter_period_old, st->prefilter_period, st->mode->shortMdctSize, - st->prefilter_gain_old, st->prefilter_gain, st->prefilter_tapset_old, st->prefilter_tapset, - st->mode->window, st->overlap); - if (LM!=0) - comb_filter(out_mem[c]+st->mode->shortMdctSize, out_mem[c]+st->mode->shortMdctSize, st->prefilter_period, pitch_index, N-st->mode->shortMdctSize, - st->prefilter_gain, gain1, st->prefilter_tapset, prefilter_tapset, - st->mode->window, st->mode->overlap); - } while (++cupsample, st->mode->preemph, st->preemph_memD); - st->prefilter_period_old = st->prefilter_period; - st->prefilter_gain_old = st->prefilter_gain; - st->prefilter_tapset_old = st->prefilter_tapset; - } -#endif - - st->prefilter_period = pitch_index; - st->prefilter_gain = gain1; - st->prefilter_tapset = prefilter_tapset; -#ifdef RESYNTH - if (LM!=0) - { - st->prefilter_period_old = st->prefilter_period; - st->prefilter_gain_old = st->prefilter_gain; - st->prefilter_tapset_old = st->prefilter_tapset; - } -#endif - - if (CC==2&&C==1) { - for (i=0;imode->nbEBands;i++) - oldBandE[st->mode->nbEBands+i]=oldBandE[i]; - } - - if (!isTransient) - { - for (i=0;imode->nbEBands;i++) - oldLogE2[i] = oldLogE[i]; - for (i=0;imode->nbEBands;i++) - oldLogE[i] = oldBandE[i]; - } else { - for (i=0;imode->nbEBands;i++) - oldLogE[i] = MIN16(oldLogE[i], oldBandE[i]); - } - /* In case start or end were to change */ - c=0; do - { - for (i=0;istart;i++) - { - oldBandE[c*st->mode->nbEBands+i]=0; - oldLogE[c*st->mode->nbEBands+i]=oldLogE2[c*st->mode->nbEBands+i]=-QCONST16(28.f,DB_SHIFT); - } - for (i=st->end;imode->nbEBands;i++) - { - oldBandE[c*st->mode->nbEBands+i]=0; - oldLogE[c*st->mode->nbEBands+i]=oldLogE2[c*st->mode->nbEBands+i]=-QCONST16(28.f,DB_SHIFT); - } - } while (++cconsec_transient++; - else - st->consec_transient=0; - st->rng = enc->rng; - - /* If there's any room left (can only happen for very high rates), - it's already filled with zeros */ - ec_enc_done(enc); - -#ifdef CUSTOM_MODES - if (st->signalling) - nbCompressedBytes++; -#endif - - RESTORE_STACK; - if (ec_get_error(enc)) - return OPUS_INTERNAL_ERROR; - else - return nbCompressedBytes; -} - - -#ifdef CUSTOM_MODES - -#ifdef FIXED_POINT -int opus_custom_encode(CELTEncoder * OPUS_RESTRICT st, const opus_int16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes) -{ - return celt_encode_with_ec(st, pcm, frame_size, compressed, nbCompressedBytes, NULL); -} - -#ifndef DISABLE_FLOAT_API -int opus_custom_encode_float(CELTEncoder * OPUS_RESTRICT st, const float * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes) -{ - int j, ret, C, N; - VARDECL(opus_int16, in); - ALLOC_STACK; - - if (pcm==NULL) - return OPUS_BAD_ARG; - - C = st->channels; - N = frame_size; - ALLOC(in, C*N, opus_int16); - - for (j=0;jchannels; - N=frame_size; - ALLOC(in, C*N, celt_sig); - for (j=0;j10) - goto bad_arg; - st->complexity = value; - } - break; - case CELT_SET_START_BAND_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - if (value<0 || value>=st->mode->nbEBands) - goto bad_arg; - st->start = value; - } - break; - case CELT_SET_END_BAND_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - if (value<1 || value>st->mode->nbEBands) - goto bad_arg; - st->end = value; - } - break; - case CELT_SET_PREDICTION_REQUEST: - { - int value = va_arg(ap, opus_int32); - if (value<0 || value>2) - goto bad_arg; - st->disable_pf = value<=1; - st->force_intra = value==0; - } - break; - case OPUS_SET_PACKET_LOSS_PERC_REQUEST: - { - int value = va_arg(ap, opus_int32); - if (value<0 || value>100) - goto bad_arg; - st->loss_rate = value; - } - break; - case OPUS_SET_VBR_CONSTRAINT_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - st->constrained_vbr = value; - } - break; - case OPUS_SET_VBR_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - st->vbr = value; - } - break; - case OPUS_SET_BITRATE_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - if (value<=500 && value!=OPUS_BITRATE_MAX) - goto bad_arg; - value = IMIN(value, 260000*st->channels); - st->bitrate = value; - } - break; - case CELT_SET_CHANNELS_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - if (value<1 || value>2) - goto bad_arg; - st->stream_channels = value; - } - break; - case OPUS_SET_LSB_DEPTH_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - if (value<8 || value>24) - goto bad_arg; - st->lsb_depth=value; - } - break; - case OPUS_GET_LSB_DEPTH_REQUEST: - { - opus_int32 *value = va_arg(ap, opus_int32*); - *value=st->lsb_depth; - } - break; - case OPUS_RESET_STATE: - { - int i; - opus_val16 *oldBandE, *oldLogE, *oldLogE2; - oldBandE = (opus_val16*)(st->in_mem+st->channels*(st->overlap+COMBFILTER_MAXPERIOD)); - oldLogE = oldBandE + st->channels*st->mode->nbEBands; - oldLogE2 = oldLogE + st->channels*st->mode->nbEBands; - OPUS_CLEAR((char*)&st->ENCODER_RESET_START, - opus_custom_encoder_get_size(st->mode, st->channels)- - ((char*)&st->ENCODER_RESET_START - (char*)st)); - for (i=0;ichannels*st->mode->nbEBands;i++) - oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT); - st->vbr_offset = 0; - st->delayedIntra = 1; - st->spread_decision = SPREAD_NORMAL; - st->tonal_average = 256; - st->hf_average = 0; - st->tapset_decision = 0; - } - break; -#ifdef CUSTOM_MODES - case CELT_SET_INPUT_CLIPPING_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - st->clip = value; - } - break; -#endif - case CELT_SET_SIGNALLING_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - st->signalling = value; - } - break; - case CELT_GET_MODE_REQUEST: - { - const CELTMode ** value = va_arg(ap, const CELTMode**); - if (value==0) - goto bad_arg; - *value=st->mode; - } - break; - case OPUS_GET_FINAL_RANGE_REQUEST: - { - opus_uint32 * value = va_arg(ap, opus_uint32 *); - if (value==0) - goto bad_arg; - *value=st->rng; - } - break; - default: - goto bad_request; - } - va_end(ap); - return OPUS_OK; -bad_arg: - va_end(ap); - return OPUS_BAD_ARG; -bad_request: - va_end(ap); - return OPUS_UNIMPLEMENTED; -} - -/**********************************************************************/ -/* */ -/* DECODER */ -/* */ -/**********************************************************************/ -#define DECODE_BUFFER_SIZE 2048 - -/** Decoder state - @brief Decoder state - */ -struct OpusCustomDecoder { - const OpusCustomMode *mode; - int overlap; - int channels; - int stream_channels; - - int downsample; - int start, end; - int signalling; - - /* Everything beyond this point gets cleared on a reset */ -#define DECODER_RESET_START rng - - opus_uint32 rng; - int error; - int last_pitch_index; - int loss_count; - int postfilter_period; - int postfilter_period_old; - opus_val16 postfilter_gain; - opus_val16 postfilter_gain_old; - int postfilter_tapset; - int postfilter_tapset_old; - - celt_sig preemph_memD[2]; - - celt_sig _decode_mem[1]; /* Size = channels*(DECODE_BUFFER_SIZE+mode->overlap) */ - /* opus_val16 lpc[], Size = channels*LPC_ORDER */ - /* opus_val16 oldEBands[], Size = 2*mode->nbEBands */ - /* opus_val16 oldLogE[], Size = 2*mode->nbEBands */ - /* opus_val16 oldLogE2[], Size = 2*mode->nbEBands */ - /* opus_val16 backgroundLogE[], Size = 2*mode->nbEBands */ -}; - -int celt_decoder_get_size(int channels) -{ - const CELTMode *mode = opus_custom_mode_create(48000, 960, NULL); - return opus_custom_decoder_get_size(mode, channels); -} - -OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_get_size(const CELTMode *mode, int channels) -{ - int size = sizeof(struct CELTDecoder) - + (channels*(DECODE_BUFFER_SIZE+mode->overlap)-1)*sizeof(celt_sig) - + channels*LPC_ORDER*sizeof(opus_val16) - + 4*2*mode->nbEBands*sizeof(opus_val16); - return size; -} - -#ifdef CUSTOM_MODES -CELTDecoder *opus_custom_decoder_create(const CELTMode *mode, int channels, int *error) -{ - int ret; - CELTDecoder *st = (CELTDecoder *)opus_alloc(opus_custom_decoder_get_size(mode, channels)); - ret = opus_custom_decoder_init(st, mode, channels); - if (ret != OPUS_OK) - { - opus_custom_decoder_destroy(st); - st = NULL; - } - if (error) - *error = ret; - return st; -} -#endif /* CUSTOM_MODES */ - -int celt_decoder_init(CELTDecoder *st, opus_int32 sampling_rate, int channels) -{ - int ret; - ret = opus_custom_decoder_init(st, opus_custom_mode_create(48000, 960, NULL), channels); - if (ret != OPUS_OK) - return ret; - st->downsample = resampling_factor(sampling_rate); - if (st->downsample==0) - return OPUS_BAD_ARG; - else - return OPUS_OK; -} - -OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_init(CELTDecoder *st, const CELTMode *mode, int channels) -{ - if (channels < 0 || channels > 2) - return OPUS_BAD_ARG; - - if (st==NULL) - return OPUS_ALLOC_FAIL; - - OPUS_CLEAR((char*)st, opus_custom_decoder_get_size(mode, channels)); - - st->mode = mode; - st->overlap = mode->overlap; - st->stream_channels = st->channels = channels; - - st->downsample = 1; - st->start = 0; - st->end = st->mode->effEBands; - st->signalling = 1; - - st->loss_count = 0; - - opus_custom_decoder_ctl(st, OPUS_RESET_STATE); - - return OPUS_OK; -} - -#ifdef CUSTOM_MODES -void opus_custom_decoder_destroy(CELTDecoder *st) -{ - opus_free(st); -} -#endif /* CUSTOM_MODES */ - -static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, opus_val16 * OPUS_RESTRICT pcm, int N, int LM) -{ - int c; - int pitch_index; - opus_val16 fade = Q15ONE; - int i, len; - const int C = st->channels; - int offset; - celt_sig *out_mem[2]; - celt_sig *decode_mem[2]; - celt_sig *overlap_mem[2]; - opus_val16 *lpc; - opus_val32 *out_syn[2]; - opus_val16 *oldBandE, *oldLogE, *oldLogE2, *backgroundLogE; - const OpusCustomMode *mode; - int nbEBands; - int overlap; - const opus_int16 *eBands; - SAVE_STACK; - - mode = st->mode; - nbEBands = mode->nbEBands; - overlap = mode->overlap; - eBands = mode->eBands; - - c=0; do { - decode_mem[c] = st->_decode_mem + c*(DECODE_BUFFER_SIZE+st->overlap); - out_mem[c] = decode_mem[c]+DECODE_BUFFER_SIZE-MAX_PERIOD; - overlap_mem[c] = decode_mem[c]+DECODE_BUFFER_SIZE; - } while (++c_decode_mem+(DECODE_BUFFER_SIZE+st->overlap)*C); - oldBandE = lpc+C*LPC_ORDER; - oldLogE = oldBandE + 2*nbEBands; - oldLogE2 = oldLogE + 2*nbEBands; - backgroundLogE = oldLogE2 + 2*nbEBands; - - c=0; do { - out_syn[c] = out_mem[c]+MAX_PERIOD-N; - } while (++closs_count >= 5 || st->start!=0) - { - /* Noise-based PLC/CNG */ - VARDECL(celt_sig, freq); - VARDECL(celt_norm, X); - VARDECL(celt_ener, bandE); - opus_uint32 seed; - int effEnd; - - effEnd = st->end; - if (effEnd > mode->effEBands) - effEnd = mode->effEBands; - - ALLOC(freq, C*N, celt_sig); /**< Interleaved signal MDCTs */ - ALLOC(X, C*N, celt_norm); /**< Interleaved normalised MDCTs */ - ALLOC(bandE, nbEBands*C, celt_ener); - - if (st->loss_count >= 5) - log2Amp(mode, st->start, st->end, bandE, backgroundLogE, C); - else { - /* Energy decay */ - opus_val16 decay = st->loss_count==0 ? QCONST16(1.5f, DB_SHIFT) : QCONST16(.5f, DB_SHIFT); - c=0; do - { - for (i=st->start;iend;i++) - oldBandE[c*nbEBands+i] -= decay; - } while (++cstart, st->end, bandE, oldBandE, C); - } - seed = st->rng; - for (c=0;cmode->eBands[st->start]<start;ieffEBands;i++) - { - int j; - int boffs; - int blen; - boffs = N*c+(eBands[i]<>20); - } - renormalise_vector(X+boffs, blen, Q15ONE); - } - for (i=(st->mode->eBands[st->end]<rng = seed; - - denormalise_bands(mode, X, freq, bandE, mode->effEBands, C, 1<mode->eBands[st->start]<downsample!=1) - bound = IMIN(bound, N/st->downsample); - for (i=bound;iloss_count == 0) - { - opus_val16 pitch_buf[DECODE_BUFFER_SIZE>>1]; - /* Corresponds to a min pitch of 67 Hz. It's possible to save CPU in this - search by using only part of the decode buffer */ - int poffset = 720; - pitch_downsample(decode_mem, pitch_buf, DECODE_BUFFER_SIZE, C); - /* Max pitch is 100 samples (480 Hz) */ - pitch_search(pitch_buf+((poffset)>>1), pitch_buf, DECODE_BUFFER_SIZE-poffset, - poffset-100, &pitch_index); - pitch_index = poffset-pitch_index; - st->last_pitch_index = pitch_index; - } else { - pitch_index = st->last_pitch_index; - fade = QCONST16(.8f,15); - } - - ALLOC(etmp, overlap, opus_val32); - c=0; do { - opus_val16 exc[MAX_PERIOD]; - opus_val32 ac[LPC_ORDER+1]; - opus_val16 decay; - opus_val16 attenuation; - opus_val32 S1=0; - opus_val16 mem[LPC_ORDER]={0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0}; - opus_val32 *e = out_syn[c]; - - - offset = MAX_PERIOD-pitch_index; - for (i=0;iloss_count == 0) - { - _celt_autocorr(exc, ac, mode->window, overlap, - LPC_ORDER, MAX_PERIOD); - - /* Noise floor -40 dB */ -#ifdef FIXED_POINT - ac[0] += SHR32(ac[0],13); -#else - ac[0] *= 1.0001f; -#endif - /* Lag windowing */ - for (i=1;i<=LPC_ORDER;i++) - { - /*ac[i] *= exp(-.5*(2*M_PI*.002*i)*(2*M_PI*.002*i));*/ -#ifdef FIXED_POINT - ac[i] -= MULT16_32_Q15(2*i*i, ac[i]); -#else - ac[i] -= ac[i]*(.008f*i)*(.008f*i); -#endif - } - - _celt_lpc(lpc+c*LPC_ORDER, ac, LPC_ORDER); - } - /* Samples just before the beginning of exc */ - for (i=0;i E2) - E1 = E2; - decay = celt_sqrt(frac_div32(SHR32(E1,1),E2)); - attenuation = decay; - } - - /* Move memory one frame to the left */ - OPUS_MOVE(decode_mem[c], decode_mem[c]+N, DECODE_BUFFER_SIZE-N+overlap); - - /* Extrapolate excitation with the right period, taking decay into account */ - for (i=0;i= MAX_PERIOD) - { - offset -= pitch_index; - attenuation = MULT16_16_Q15(attenuation, decay); - } - e[i] = SHL32(EXTEND32(MULT16_16_Q15(attenuation, exc[offset+i])), SIG_SHIFT); - /* Compute the energy of the previously decoded signal whose - excitation we're copying */ - tmp = ROUND16(out_mem[c][-N+offset+i],SIG_SHIFT); - S1 += SHR32(MULT16_16(tmp,tmp),8); - } - - /* Copy the last decoded samples (prior to the overlap region) to - synthesis filter memory so we can have a continuous signal. */ - for (i=0;i SHR32(S2,2))) -#else - /* Float test is written this way to catch NaNs at the same time */ - if (!(S1 > 0.2f*S2)) -#endif - { - for (i=0;iwindow[i], Q15ONE-ratio); - e[i] = MULT16_32_Q15(tmp_g, e[i]); - } - for (i=overlap;ipostfilter_period, st->postfilter_period, st->overlap, - -st->postfilter_gain, -st->postfilter_gain, st->postfilter_tapset, st->postfilter_tapset, - NULL, 0); - - /* Simulate TDAC on the concealed audio so that it blends with the - MDCT of next frames. */ - for (i=0;iwindow[i], etmp[overlap-1-i]) + - MULT16_32_Q15(mode->window[overlap-i-1], etmp[i ]); - out_mem[c][MAX_PERIOD+i] = MULT16_32_Q15(mode->window[overlap-i-1], tmp); - out_mem[c][MAX_PERIOD+overlap-i-1] = MULT16_32_Q15(mode->window[i], tmp); - } - } while (++cdownsample, mode->preemph, st->preemph_memD); - - st->loss_count++; - - RESTORE_STACK; -} - -int celt_decode_with_ec(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, opus_val16 * OPUS_RESTRICT pcm, int frame_size, ec_dec *dec) -{ - int c, i, N; - int spread_decision; - opus_int32 bits; - ec_dec _dec; - VARDECL(celt_sig, freq); - VARDECL(celt_norm, X); - VARDECL(celt_ener, bandE); - VARDECL(int, fine_quant); - VARDECL(int, pulses); - VARDECL(int, cap); - VARDECL(int, offsets); - VARDECL(int, fine_priority); - VARDECL(int, tf_res); - VARDECL(unsigned char, collapse_masks); - celt_sig *out_mem[2]; - celt_sig *decode_mem[2]; - celt_sig *overlap_mem[2]; - celt_sig *out_syn[2]; - opus_val16 *lpc; - opus_val16 *oldBandE, *oldLogE, *oldLogE2, *backgroundLogE; - - int shortBlocks; - int isTransient; - int intra_ener; - const int CC = st->channels; - int LM, M; - int effEnd; - int codedBands; - int alloc_trim; - int postfilter_pitch; - opus_val16 postfilter_gain; - int intensity=0; - int dual_stereo=0; - opus_int32 total_bits; - opus_int32 balance; - opus_int32 tell; - int dynalloc_logp; - int postfilter_tapset; - int anti_collapse_rsv; - int anti_collapse_on=0; - int silence; - int C = st->stream_channels; - ALLOC_STACK; - - frame_size *= st->downsample; - - c=0; do { - decode_mem[c] = st->_decode_mem + c*(DECODE_BUFFER_SIZE+st->overlap); - out_mem[c] = decode_mem[c]+DECODE_BUFFER_SIZE-MAX_PERIOD; - overlap_mem[c] = decode_mem[c]+DECODE_BUFFER_SIZE; - } while (++c_decode_mem+(DECODE_BUFFER_SIZE+st->overlap)*CC); - oldBandE = lpc+CC*LPC_ORDER; - oldLogE = oldBandE + 2*st->mode->nbEBands; - oldLogE2 = oldLogE + 2*st->mode->nbEBands; - backgroundLogE = oldLogE2 + 2*st->mode->nbEBands; - -#ifdef CUSTOM_MODES - if (st->signalling && data!=NULL) - { - int data0=data[0]; - /* Convert "standard mode" to Opus header */ - if (st->mode->Fs==48000 && st->mode->shortMdctSize==120) - { - data0 = fromOpus(data0); - if (data0<0) - return OPUS_INVALID_PACKET; - } - st->end = IMAX(1, st->mode->effEBands-2*(data0>>5)); - LM = (data0>>3)&0x3; - C = 1 + ((data0>>2)&0x1); - data++; - len--; - if (LM>st->mode->maxLM) - return OPUS_INVALID_PACKET; - if (frame_size < st->mode->shortMdctSize<mode->shortMdctSize<mode->maxLM;LM++) - if (st->mode->shortMdctSize<st->mode->maxLM) - return OPUS_BAD_ARG; - } - M=1<1275 || pcm==NULL) - return OPUS_BAD_ARG; - - N = M*st->mode->shortMdctSize; - - effEnd = st->end; - if (effEnd > st->mode->effEBands) - effEnd = st->mode->effEBands; - - if (data == NULL || len<=1) - { - celt_decode_lost(st, pcm, N, LM); - RESTORE_STACK; - return frame_size/st->downsample; - } - - ALLOC(freq, IMAX(CC,C)*N, celt_sig); /**< Interleaved signal MDCTs */ - ALLOC(X, C*N, celt_norm); /**< Interleaved normalised MDCTs */ - ALLOC(bandE, st->mode->nbEBands*C, celt_ener); - c=0; do - for (i=0;imode->eBands[st->start];i++) - X[c*N+i] = 0; - while (++cmode->eBands[effEnd];imode->nbEBands;i++) - oldBandE[i]=MAX16(oldBandE[i],oldBandE[st->mode->nbEBands+i]); - } - - total_bits = len*8; - tell = ec_tell(dec); - - if (tell >= total_bits) - silence = 1; - else if (tell==1) - silence = ec_dec_bit_logp(dec, 15); - else - silence = 0; - if (silence) - { - /* Pretend we've read all the remaining bits */ - tell = len*8; - dec->nbits_total+=tell-ec_tell(dec); - } - - postfilter_gain = 0; - postfilter_pitch = 0; - postfilter_tapset = 0; - if (st->start==0 && tell+16 <= total_bits) - { - if(ec_dec_bit_logp(dec, 1)) - { - int qg, octave; - octave = ec_dec_uint(dec, 6); - postfilter_pitch = (16< 0 && tell+3 <= total_bits) - { - isTransient = ec_dec_bit_logp(dec, 3); - tell = ec_tell(dec); - } - else - isTransient = 0; - - if (isTransient) - shortBlocks = M; - else - shortBlocks = 0; - - /* Decode the global flags (first symbols in the stream) */ - intra_ener = tell+3<=total_bits ? ec_dec_bit_logp(dec, 3) : 0; - /* Get band energies */ - unquant_coarse_energy(st->mode, st->start, st->end, oldBandE, - intra_ener, dec, C, LM); - - ALLOC(tf_res, st->mode->nbEBands, int); - tf_decode(st->start, st->end, isTransient, tf_res, LM, dec); - - tell = ec_tell(dec); - spread_decision = SPREAD_NORMAL; - if (tell+4 <= total_bits) - spread_decision = ec_dec_icdf(dec, spread_icdf, 5); - - ALLOC(pulses, st->mode->nbEBands, int); - ALLOC(cap, st->mode->nbEBands, int); - ALLOC(offsets, st->mode->nbEBands, int); - ALLOC(fine_priority, st->mode->nbEBands, int); - - init_caps(st->mode,cap,LM,C); - - dynalloc_logp = 6; - total_bits<<=BITRES; - tell = ec_tell_frac(dec); - for (i=st->start;iend;i++) - { - int width, quanta; - int dynalloc_loop_logp; - int boost; - width = C*(st->mode->eBands[i+1]-st->mode->eBands[i])<0) - dynalloc_logp = IMAX(2, dynalloc_logp-1); - } - - ALLOC(fine_quant, st->mode->nbEBands, int); - alloc_trim = tell+(6<=2&&bits>=((LM+2)<mode, st->start, st->end, offsets, cap, - alloc_trim, &intensity, &dual_stereo, bits, &balance, pulses, - fine_quant, fine_priority, C, LM, dec, 0, 0); - - unquant_fine_energy(st->mode, st->start, st->end, oldBandE, fine_quant, dec, C); - - /* Decode fixed codebook */ - ALLOC(collapse_masks, C*st->mode->nbEBands, unsigned char); - quant_all_bands(0, st->mode, st->start, st->end, X, C==2 ? X+N : NULL, collapse_masks, - NULL, pulses, shortBlocks, spread_decision, dual_stereo, intensity, tf_res, - len*(8<rng); - - if (anti_collapse_rsv > 0) - { - anti_collapse_on = ec_dec_bits(dec, 1); - } - - unquant_energy_finalise(st->mode, st->start, st->end, oldBandE, - fine_quant, fine_priority, len*8-ec_tell(dec), dec, C); - - if (anti_collapse_on) - anti_collapse(st->mode, X, collapse_masks, LM, C, N, - st->start, st->end, oldBandE, oldLogE, oldLogE2, pulses, st->rng); - - log2Amp(st->mode, st->start, st->end, bandE, oldBandE, C); - - if (silence) - { - for (i=0;imode->nbEBands;i++) - { - bandE[i] = 0; - oldBandE[i] = -QCONST16(28.f,DB_SHIFT); - } - } - /* Synthesis */ - denormalise_bands(st->mode, X, freq, bandE, effEnd, C, M); - - OPUS_MOVE(decode_mem[0], decode_mem[0]+N, DECODE_BUFFER_SIZE-N); - if (CC==2) - OPUS_MOVE(decode_mem[1], decode_mem[1]+N, DECODE_BUFFER_SIZE-N); - - c=0; do - for (i=0;imode->eBands[st->start];i++) - freq[c*N+i] = 0; - while (++cmode->eBands[effEnd]; - if (st->downsample!=1) - bound = IMIN(bound, N/st->downsample); - for (i=bound;imode, shortBlocks, freq, out_syn, overlap_mem, CC, LM); - - c=0; do { - st->postfilter_period=IMAX(st->postfilter_period, COMBFILTER_MINPERIOD); - st->postfilter_period_old=IMAX(st->postfilter_period_old, COMBFILTER_MINPERIOD); - comb_filter(out_syn[c], out_syn[c], st->postfilter_period_old, st->postfilter_period, st->mode->shortMdctSize, - st->postfilter_gain_old, st->postfilter_gain, st->postfilter_tapset_old, st->postfilter_tapset, - st->mode->window, st->overlap); - if (LM!=0) - comb_filter(out_syn[c]+st->mode->shortMdctSize, out_syn[c]+st->mode->shortMdctSize, st->postfilter_period, postfilter_pitch, N-st->mode->shortMdctSize, - st->postfilter_gain, postfilter_gain, st->postfilter_tapset, postfilter_tapset, - st->mode->window, st->mode->overlap); - - } while (++cpostfilter_period_old = st->postfilter_period; - st->postfilter_gain_old = st->postfilter_gain; - st->postfilter_tapset_old = st->postfilter_tapset; - st->postfilter_period = postfilter_pitch; - st->postfilter_gain = postfilter_gain; - st->postfilter_tapset = postfilter_tapset; - if (LM!=0) - { - st->postfilter_period_old = st->postfilter_period; - st->postfilter_gain_old = st->postfilter_gain; - st->postfilter_tapset_old = st->postfilter_tapset; - } - - if (C==1) { - for (i=0;imode->nbEBands;i++) - oldBandE[st->mode->nbEBands+i]=oldBandE[i]; - } - - /* In case start or end were to change */ - if (!isTransient) - { - for (i=0;i<2*st->mode->nbEBands;i++) - oldLogE2[i] = oldLogE[i]; - for (i=0;i<2*st->mode->nbEBands;i++) - oldLogE[i] = oldBandE[i]; - for (i=0;i<2*st->mode->nbEBands;i++) - backgroundLogE[i] = MIN16(backgroundLogE[i] + M*QCONST16(0.001f,DB_SHIFT), oldBandE[i]); - } else { - for (i=0;i<2*st->mode->nbEBands;i++) - oldLogE[i] = MIN16(oldLogE[i], oldBandE[i]); - } - c=0; do - { - for (i=0;istart;i++) - { - oldBandE[c*st->mode->nbEBands+i]=0; - oldLogE[c*st->mode->nbEBands+i]=oldLogE2[c*st->mode->nbEBands+i]=-QCONST16(28.f,DB_SHIFT); - } - for (i=st->end;imode->nbEBands;i++) - { - oldBandE[c*st->mode->nbEBands+i]=0; - oldLogE[c*st->mode->nbEBands+i]=oldLogE2[c*st->mode->nbEBands+i]=-QCONST16(28.f,DB_SHIFT); - } - } while (++c<2); - st->rng = dec->rng; - - deemphasis(out_syn, pcm, N, CC, st->downsample, st->mode->preemph, st->preemph_memD); - st->loss_count = 0; - RESTORE_STACK; - if (ec_tell(dec) > 8*len) - return OPUS_INTERNAL_ERROR; - if(ec_get_error(dec)) - st->error = 1; - return frame_size/st->downsample; -} - - -#ifdef CUSTOM_MODES - -#ifdef FIXED_POINT -int opus_custom_decode(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, opus_int16 * OPUS_RESTRICT pcm, int frame_size) -{ - return celt_decode_with_ec(st, data, len, pcm, frame_size, NULL); -} - -#ifndef DISABLE_FLOAT_API -int opus_custom_decode_float(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, float * OPUS_RESTRICT pcm, int frame_size) -{ - int j, ret, C, N; - VARDECL(opus_int16, out); - ALLOC_STACK; - - if (pcm==NULL) - return OPUS_BAD_ARG; - - C = st->channels; - N = frame_size; - - ALLOC(out, C*N, opus_int16); - ret=celt_decode_with_ec(st, data, len, out, frame_size, NULL); - if (ret>0) - for (j=0;jchannels; - N = frame_size; - ALLOC(out, C*N, celt_sig); - - ret=celt_decode_with_ec(st, data, len, out, frame_size, NULL); - - if (ret>0) - for (j=0;j=st->mode->nbEBands) - goto bad_arg; - st->start = value; - } - break; - case CELT_SET_END_BAND_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - if (value<1 || value>st->mode->nbEBands) - goto bad_arg; - st->end = value; - } - break; - case CELT_SET_CHANNELS_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - if (value<1 || value>2) - goto bad_arg; - st->stream_channels = value; - } - break; - case CELT_GET_AND_CLEAR_ERROR_REQUEST: - { - opus_int32 *value = va_arg(ap, opus_int32*); - if (value==NULL) - goto bad_arg; - *value=st->error; - st->error = 0; - } - break; - case OPUS_GET_LOOKAHEAD_REQUEST: - { - opus_int32 *value = va_arg(ap, opus_int32*); - if (value==NULL) - goto bad_arg; - *value = st->overlap/st->downsample; - } - break; - case OPUS_RESET_STATE: - { - int i; - opus_val16 *lpc, *oldBandE, *oldLogE, *oldLogE2; - lpc = (opus_val16*)(st->_decode_mem+(DECODE_BUFFER_SIZE+st->overlap)*st->channels); - oldBandE = lpc+st->channels*LPC_ORDER; - oldLogE = oldBandE + 2*st->mode->nbEBands; - oldLogE2 = oldLogE + 2*st->mode->nbEBands; - OPUS_CLEAR((char*)&st->DECODER_RESET_START, - opus_custom_decoder_get_size(st->mode, st->channels)- - ((char*)&st->DECODER_RESET_START - (char*)st)); - for (i=0;i<2*st->mode->nbEBands;i++) - oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT); - } - break; - case OPUS_GET_PITCH_REQUEST: - { - opus_int32 *value = va_arg(ap, opus_int32*); - if (value==NULL) - goto bad_arg; - *value = st->postfilter_period; - } - break; - case CELT_GET_MODE_REQUEST: - { - const CELTMode ** value = va_arg(ap, const CELTMode**); - if (value==0) - goto bad_arg; - *value=st->mode; - } - break; - case CELT_SET_SIGNALLING_REQUEST: - { - opus_int32 value = va_arg(ap, opus_int32); - st->signalling = value; - } - break; - case OPUS_GET_FINAL_RANGE_REQUEST: - { - opus_uint32 * value = va_arg(ap, opus_uint32 *); - if (value==0) - goto bad_arg; - *value=st->rng; - } - break; - default: - goto bad_request; - } - va_end(ap); - return OPUS_OK; -bad_arg: - va_end(ap); - return OPUS_BAD_ARG; -bad_request: - va_end(ap); - return OPUS_UNIMPLEMENTED; -} - - - -const char *opus_strerror(int error) -{ - static const char * const error_strings[8] = { - "success", - "invalid argument", - "buffer too small", - "internal error", - "corrupted stream", - "request not implemented", - "invalid state", - "memory allocation failed" - }; - if (error > 0 || error < -7) - return "unknown error"; - else - return error_strings[-error]; -} - -const char *opus_get_version_string(void) -{ - return "libopus " OPUS_VERSION -#ifdef FIXED_POINT - "-fixed" -#endif -#ifdef FUZZING - "-fuzzing" -#endif - ; -} diff --git a/code/opus-1.0.2/celt/celt_lpc.c b/code/opus-1.0.2/celt/celt_lpc.c deleted file mode 100644 index d2addbf2..00000000 --- a/code/opus-1.0.2/celt/celt_lpc.c +++ /dev/null @@ -1,188 +0,0 @@ -/* Copyright (c) 2009-2010 Xiph.Org Foundation - Written by Jean-Marc Valin */ -/* - Redistribution and use in source and binary forms, with or without - modification, are permitted provided that the following conditions - are met: - - - Redistributions of source code must retain the above copyright - notice, this list of conditions and the following disclaimer. - - - Redistributions in binary form must reproduce the above copyright - notice, this list of conditions and the following disclaimer in the - documentation and/or other materials provided with the distribution. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER - OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, - EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, - PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR - PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF - LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING - NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -*/ - -#ifdef HAVE_CONFIG_H -#include "config.h" -#endif - -#include "celt_lpc.h" -#include "stack_alloc.h" -#include "mathops.h" - -void _celt_lpc( - opus_val16 *_lpc, /* out: [0...p-1] LPC coefficients */ -const opus_val32 *ac, /* in: [0...p] autocorrelation values */ -int p -) -{ - int i, j; - opus_val32 r; - opus_val32 error = ac[0]; -#ifdef FIXED_POINT - opus_val32 lpc[LPC_ORDER]; -#else - float *lpc = _lpc; -#endif - - for (i = 0; i < p; i++) - lpc[i] = 0; - if (ac[0] != 0) - { - for (i = 0; i < p; i++) { - /* Sum up this iteration's reflection coefficient */ - opus_val32 rr = 0; - for (j = 0; j < i; j++) - rr += MULT32_32_Q31(lpc[j],ac[i - j]); - rr += SHR32(ac[i + 1],3); - r = -frac_div32(SHL32(rr,3), error); - /* Update LPC coefficients and total error */ - lpc[i] = SHR32(r,3); - for (j = 0; j < (i+1)>>1; j++) - { - opus_val32 tmp1, tmp2; - tmp1 = lpc[j]; - tmp2 = lpc[i-1-j]; - lpc[j] = tmp1 + MULT32_32_Q31(r,tmp2); - lpc[i-1-j] = tmp2 + MULT32_32_Q31(r,tmp1); - } - - error = error - MULT32_32_Q31(MULT32_32_Q31(r,r),error); - /* Bail out once we get 30 dB gain */ -#ifdef FIXED_POINT - if (error=1;j--) - { - mem[j]=mem[j-1]; - } - mem[0] = x[i]; - y[i] = ROUND16(sum, SIG_SHIFT); - } -} - -void celt_iir(const opus_val32 *x, - const opus_val16 *den, - opus_val32 *y, - int N, - int ord, - opus_val16 *mem) -{ - int i,j; - for (i=0;i=1;j--) - { - mem[j]=mem[j-1]; - } - mem[0] = ROUND16(sum,SIG_SHIFT); - y[i] = sum; - } -} - -void _celt_autocorr( - const opus_val16 *x, /* in: [0...n-1] samples x */ - opus_val32 *ac, /* out: [0...lag-1] ac values */ - const opus_val16 *window, - int overlap, - int lag, - int n - ) -{ - opus_val32 d; - int i; - VARDECL(opus_val16, xx); - SAVE_STACK; - ALLOC(xx, n, opus_val16); - celt_assert(n>0); - celt_assert(overlap>=0); - for (i=0;i=0) - { - for (i = lag, d = 0; i < n; i++) - d += xx[i] * xx[i-lag]; - ac[lag] = d; - /*printf ("%f ", ac[lag]);*/ - lag--; - } - /*printf ("\n");*/ - ac[0] += 10; - - RESTORE_STACK; -} diff --git a/code/opus-1.0.2/celt/cwrs.c b/code/opus-1.0.2/celt/cwrs.c deleted file mode 100644 index 8edc919d..00000000 --- a/code/opus-1.0.2/celt/cwrs.c +++ /dev/null @@ -1,645 +0,0 @@ -/* Copyright (c) 2007-2008 CSIRO - Copyright (c) 2007-2009 Xiph.Org Foundation - Copyright (c) 2007-2009 Timothy B. Terriberry - Written by Timothy B. Terriberry and Jean-Marc Valin */ -/* - Redistribution and use in source and binary forms, with or without - modification, are permitted provided that the following conditions - are met: - - - Redistributions of source code must retain the above copyright - notice, this list of conditions and the following disclaimer. - - - Redistributions in binary form must reproduce the above copyright - notice, this list of conditions and the following disclaimer in the - documentation and/or other materials provided with the distribution. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER - OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, - EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, - PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR - PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF - LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING - NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -*/ - -#ifdef HAVE_CONFIG_H -#include "config.h" -#endif - -#include "os_support.h" -#include "cwrs.h" -#include "mathops.h" -#include "arch.h" - -#ifdef CUSTOM_MODES - -/*Guaranteed to return a conservatively large estimate of the binary logarithm - with frac bits of fractional precision. - Tested for all possible 32-bit inputs with frac=4, where the maximum - overestimation is 0.06254243 bits.*/ -int log2_frac(opus_uint32 val, int frac) -{ - int l; - l=EC_ILOG(val); - if(val&(val-1)){ - /*This is (val>>l-16), but guaranteed to round up, even if adding a bias - before the shift would cause overflow (e.g., for 0xFFFFxxxx). - Doesn't work for val=0, but that case fails the test above.*/ - if(l>16)val=((val-1)>>(l-16))+1; - else val<<=16-l; - l=(l-1)<>16); - l+=b<>b; - val=(val*val+0x7FFF)>>15; - } - while(frac-->0); - /*If val is not exactly 0x8000, then we have to round up the remainder.*/ - return l+(val>0x8000); - } - /*Exact powers of two require no rounding.*/ - else return (l-1)<0); - celt_assert(_d<=54); - shift=EC_ILOG(_d^(_d-1)); - inv=INV_TABLE[(_d-1)>>shift]; - shift--; - one=1<>shift)-(_c>>shift)+ - ((_a*(_b&mask)+one-(_c&mask))>>shift)-1)*inv&MASK32; -} - -#endif /* SMALL_FOOTPRINT */ - -/*Although derived separately, the pulse vector coding scheme is equivalent to - a Pyramid Vector Quantizer \cite{Fis86}. - Some additional notes about an early version appear at - http://people.xiph.org/~tterribe/notes/cwrs.html, but the codebook ordering - and the definitions of some terms have evolved since that was written. - - The conversion from a pulse vector to an integer index (encoding) and back - (decoding) is governed by two related functions, V(N,K) and U(N,K). - - V(N,K) = the number of combinations, with replacement, of N items, taken K - at a time, when a sign bit is added to each item taken at least once (i.e., - the number of N-dimensional unit pulse vectors with K pulses). - One way to compute this is via - V(N,K) = K>0 ? sum(k=1...K,2**k*choose(N,k)*choose(K-1,k-1)) : 1, - where choose() is the binomial function. - A table of values for N<10 and K<10 looks like: - V[10][10] = { - {1, 0, 0, 0, 0, 0, 0, 0, 0, 0}, - {1, 2, 2, 2, 2, 2, 2, 2, 2, 2}, - {1, 4, 8, 12, 16, 20, 24, 28, 32, 36}, - {1, 6, 18, 38, 66, 102, 146, 198, 258, 326}, - {1, 8, 32, 88, 192, 360, 608, 952, 1408, 1992}, - {1, 10, 50, 170, 450, 1002, 1970, 3530, 5890, 9290}, - {1, 12, 72, 292, 912, 2364, 5336, 10836, 20256, 35436}, - {1, 14, 98, 462, 1666, 4942, 12642, 28814, 59906, 115598}, - {1, 16, 128, 688, 2816, 9424, 27008, 68464, 157184, 332688}, - {1, 18, 162, 978, 4482, 16722, 53154, 148626, 374274, 864146} - }; - - U(N,K) = the number of such combinations wherein N-1 objects are taken at - most K-1 at a time. - This is given by - U(N,K) = sum(k=0...K-1,V(N-1,k)) - = K>0 ? (V(N-1,K-1) + V(N,K-1))/2 : 0. - The latter expression also makes clear that U(N,K) is half the number of such - combinations wherein the first object is taken at least once. - Although it may not be clear from either of these definitions, U(N,K) is the - natural function to work with when enumerating the pulse vector codebooks, - not V(N,K). - U(N,K) is not well-defined for N=0, but with the extension - U(0,K) = K>0 ? 0 : 1, - the function becomes symmetric: U(N,K) = U(K,N), with a similar table: - U[10][10] = { - {1, 0, 0, 0, 0, 0, 0, 0, 0, 0}, - {0, 1, 1, 1, 1, 1, 1, 1, 1, 1}, - {0, 1, 3, 5, 7, 9, 11, 13, 15, 17}, - {0, 1, 5, 13, 25, 41, 61, 85, 113, 145}, - {0, 1, 7, 25, 63, 129, 231, 377, 575, 833}, - {0, 1, 9, 41, 129, 321, 681, 1289, 2241, 3649}, - {0, 1, 11, 61, 231, 681, 1683, 3653, 7183, 13073}, - {0, 1, 13, 85, 377, 1289, 3653, 8989, 19825, 40081}, - {0, 1, 15, 113, 575, 2241, 7183, 19825, 48639, 108545}, - {0, 1, 17, 145, 833, 3649, 13073, 40081, 108545, 265729} - }; - - With this extension, V(N,K) may be written in terms of U(N,K): - V(N,K) = U(N,K) + U(N,K+1) - for all N>=0, K>=0. - Thus U(N,K+1) represents the number of combinations where the first element - is positive or zero, and U(N,K) represents the number of combinations where - it is negative. - With a large enough table of U(N,K) values, we could write O(N) encoding - and O(min(N*log(K),N+K)) decoding routines, but such a table would be - prohibitively large for small embedded devices (K may be as large as 32767 - for small N, and N may be as large as 200). - - Both functions obey the same recurrence relation: - V(N,K) = V(N-1,K) + V(N,K-1) + V(N-1,K-1), - U(N,K) = U(N-1,K) + U(N,K-1) + U(N-1,K-1), - for all N>0, K>0, with different initial conditions at N=0 or K=0. - This allows us to construct a row of one of the tables above given the - previous row or the next row. - Thus we can derive O(NK) encoding and decoding routines with O(K) memory - using only addition and subtraction. - - When encoding, we build up from the U(2,K) row and work our way forwards. - When decoding, we need to start at the U(N,K) row and work our way backwards, - which requires a means of computing U(N,K). - U(N,K) may be computed from two previous values with the same N: - U(N,K) = ((2*N-1)*U(N,K-1) - U(N,K-2))/(K-1) + U(N,K-2) - for all N>1, and since U(N,K) is symmetric, a similar relation holds for two - previous values with the same K: - U(N,K>1) = ((2*K-1)*U(N-1,K) - U(N-2,K))/(N-1) + U(N-2,K) - for all K>1. - This allows us to construct an arbitrary row of the U(N,K) table by starting - with the first two values, which are constants. - This saves roughly 2/3 the work in our O(NK) decoding routine, but costs O(K) - multiplications. - Similar relations can be derived for V(N,K), but are not used here. - - For N>0 and K>0, U(N,K) and V(N,K) take on the form of an (N-1)-degree - polynomial for fixed N. - The first few are - U(1,K) = 1, - U(2,K) = 2*K-1, - U(3,K) = (2*K-2)*K+1, - U(4,K) = (((4*K-6)*K+8)*K-3)/3, - U(5,K) = ((((2*K-4)*K+10)*K-8)*K+3)/3, - and - V(1,K) = 2, - V(2,K) = 4*K, - V(3,K) = 4*K*K+2, - V(4,K) = 8*(K*K+2)*K/3, - V(5,K) = ((4*K*K+20)*K*K+6)/3, - for all K>0. - This allows us to derive O(N) encoding and O(N*log(K)) decoding routines for - small N (and indeed decoding is also O(N) for N<3). - - @ARTICLE{Fis86, - author="Thomas R. Fischer", - title="A Pyramid Vector Quantizer", - journal="IEEE Transactions on Information Theory", - volume="IT-32", - number=4, - pages="568--583", - month=Jul, - year=1986 - }*/ - -#ifndef SMALL_FOOTPRINT -/*Compute U(2,_k). - Note that this may be called with _k=32768 (maxK[2]+1).*/ -static inline unsigned ucwrs2(unsigned _k){ - celt_assert(_k>0); - return _k+(_k-1); -} - -/*Compute V(2,_k).*/ -static inline opus_uint32 ncwrs2(int _k){ - celt_assert(_k>0); - return 4*(opus_uint32)_k; -} - -/*Compute U(3,_k). - Note that this may be called with _k=32768 (maxK[3]+1).*/ -static inline opus_uint32 ucwrs3(unsigned _k){ - celt_assert(_k>0); - return (2*(opus_uint32)_k-2)*_k+1; -} - -/*Compute V(3,_k).*/ -static inline opus_uint32 ncwrs3(int _k){ - celt_assert(_k>0); - return 2*(2*(unsigned)_k*(opus_uint32)_k+1); -} - -/*Compute U(4,_k).*/ -static inline opus_uint32 ucwrs4(int _k){ - celt_assert(_k>0); - return imusdiv32odd(2*_k,(2*_k-3)*(opus_uint32)_k+4,3,1); -} - -/*Compute V(4,_k).*/ -static inline opus_uint32 ncwrs4(int _k){ - celt_assert(_k>0); - return ((_k*(opus_uint32)_k+2)*_k)/3<<3; -} - -#endif /* SMALL_FOOTPRINT */ - -/*Computes the next row/column of any recurrence that obeys the relation - u[i][j]=u[i-1][j]+u[i][j-1]+u[i-1][j-1]. - _ui0 is the base case for the new row/column.*/ -static inline void unext(opus_uint32 *_ui,unsigned _len,opus_uint32 _ui0){ - opus_uint32 ui1; - unsigned j; - /*This do-while will overrun the array if we don't have storage for at least - 2 values.*/ - j=1; do { - ui1=UADD32(UADD32(_ui[j],_ui[j-1]),_ui0); - _ui[j-1]=_ui0; - _ui0=ui1; - } while (++j<_len); - _ui[j-1]=_ui0; -} - -/*Computes the previous row/column of any recurrence that obeys the relation - u[i-1][j]=u[i][j]-u[i][j-1]-u[i-1][j-1]. - _ui0 is the base case for the new row/column.*/ -static inline void uprev(opus_uint32 *_ui,unsigned _n,opus_uint32 _ui0){ - opus_uint32 ui1; - unsigned j; - /*This do-while will overrun the array if we don't have storage for at least - 2 values.*/ - j=1; do { - ui1=USUB32(USUB32(_ui[j],_ui[j-1]),_ui0); - _ui[j-1]=_ui0; - _ui0=ui1; - } while (++j<_n); - _ui[j-1]=_ui0; -} - -/*Compute V(_n,_k), as well as U(_n,0..._k+1). - _u: On exit, _u[i] contains U(_n,i) for i in [0..._k+1].*/ -static opus_uint32 ncwrs_urow(unsigned _n,unsigned _k,opus_uint32 *_u){ - opus_uint32 um2; - unsigned len; - unsigned k; - len=_k+2; - /*We require storage at least 3 values (e.g., _k>0).*/ - celt_assert(len>=3); - _u[0]=0; - _u[1]=um2=1; -#ifndef SMALL_FOOTPRINT - /*_k>52 doesn't work in the false branch due to the limits of INV_TABLE, - but _k isn't tested here because k<=52 for n=7*/ - if(_n<=6) -#endif - { - /*If _n==0, _u[0] should be 1 and the rest should be 0.*/ - /*If _n==1, _u[i] should be 1 for i>1.*/ - celt_assert(_n>=2); - /*If _k==0, the following do-while loop will overflow the buffer.*/ - celt_assert(_k>0); - k=2; - do _u[k]=(k<<1)-1; - while(++k=len)break; - _u[k]=um1=imusdiv32odd(n2m1,um2,um1,(k-1)>>1)+um1; - } - } -#endif /* SMALL_FOOTPRINT */ - return _u[_k]+_u[_k+1]; -} - -#ifndef SMALL_FOOTPRINT - -/*Returns the _i'th combination of _k elements (at most 32767) chosen from a - set of size 1 with associated sign bits. - _y: Returns the vector of pulses.*/ -static inline void cwrsi1(int _k,opus_uint32 _i,int *_y){ - int s; - s=-(int)_i; - _y[0]=(_k+s)^s; -} - -/*Returns the _i'th combination of _k elements (at most 32767) chosen from a - set of size 2 with associated sign bits. - _y: Returns the vector of pulses.*/ -static inline void cwrsi2(int _k,opus_uint32 _i,int *_y){ - opus_uint32 p; - int s; - int yj; - p=ucwrs2(_k+1U); - s=-(_i>=p); - _i-=p&s; - yj=_k; - _k=(_i+1)>>1; - p=_k?ucwrs2(_k):0; - _i-=p; - yj-=_k; - _y[0]=(yj+s)^s; - cwrsi1(_k,_i,_y+1); -} - -/*Returns the _i'th combination of _k elements (at most 32767) chosen from a - set of size 3 with associated sign bits. - _y: Returns the vector of pulses.*/ -static void cwrsi3(int _k,opus_uint32 _i,int *_y){ - opus_uint32 p; - int s; - int yj; - p=ucwrs3(_k+1U); - s=-(_i>=p); - _i-=p&s; - yj=_k; - /*Finds the maximum _k such that ucwrs3(_k)<=_i (tested for all - _i<2147418113=U(3,32768)).*/ - _k=_i>0?(isqrt32(2*_i-1)+1)>>1:0; - p=_k?ucwrs3(_k):0; - _i-=p; - yj-=_k; - _y[0]=(yj+s)^s; - cwrsi2(_k,_i,_y+1); -} - -/*Returns the _i'th combination of _k elements (at most 1172) chosen from a set - of size 4 with associated sign bits. - _y: Returns the vector of pulses.*/ -static void cwrsi4(int _k,opus_uint32 _i,int *_y){ - opus_uint32 p; - int s; - int yj; - int kl; - int kr; - p=ucwrs4(_k+1); - s=-(_i>=p); - _i-=p&s; - yj=_k; - /*We could solve a cubic for k here, but the form of the direct solution does - not lend itself well to exact integer arithmetic. - Instead we do a binary search on U(4,K).*/ - kl=0; - kr=_k; - for(;;){ - _k=(kl+kr)>>1; - p=_k?ucwrs4(_k):0; - if(p<_i){ - if(_k>=kr)break; - kl=_k+1; - } - else if(p>_i)kr=_k-1; - else break; - } - _i-=p; - yj-=_k; - _y[0]=(yj+s)^s; - cwrsi3(_k,_i,_y+1); -} - -#endif /* SMALL_FOOTPRINT */ - -/*Returns the _i'th combination of _k elements chosen from a set of size _n - with associated sign bits. - _y: Returns the vector of pulses. - _u: Must contain entries [0..._k+1] of row _n of U() on input. - Its contents will be destructively modified.*/ -static void cwrsi(int _n,int _k,opus_uint32 _i,int *_y,opus_uint32 *_u){ - int j; - celt_assert(_n>0); - j=0; - do{ - opus_uint32 p; - int s; - int yj; - p=_u[_k+1]; - s=-(_i>=p); - _i-=p&s; - yj=_k; - p=_u[_k]; - while(p>_i)p=_u[--_k]; - _i-=p; - yj-=_k; - _y[j]=(yj+s)^s; - uprev(_u,_k+2,0); - } - while(++j<_n); -} - -/*Returns the index of the given combination of K elements chosen from a set - of size 1 with associated sign bits. - _y: The vector of pulses, whose sum of absolute values is K. - _k: Returns K.*/ -static inline opus_uint32 icwrs1(const int *_y,int *_k){ - *_k=abs(_y[0]); - return _y[0]<0; -} - -#ifndef SMALL_FOOTPRINT - -/*Returns the index of the given combination of K elements chosen from a set - of size 2 with associated sign bits. - _y: The vector of pulses, whose sum of absolute values is K. - _k: Returns K.*/ -static inline opus_uint32 icwrs2(const int *_y,int *_k){ - opus_uint32 i; - int k; - i=icwrs1(_y+1,&k); - i+=k?ucwrs2(k):0; - k+=abs(_y[0]); - if(_y[0]<0)i+=ucwrs2(k+1U); - *_k=k; - return i; -} - -/*Returns the index of the given combination of K elements chosen from a set - of size 3 with associated sign bits. - _y: The vector of pulses, whose sum of absolute values is K. - _k: Returns K.*/ -static inline opus_uint32 icwrs3(const int *_y,int *_k){ - opus_uint32 i; - int k; - i=icwrs2(_y+1,&k); - i+=k?ucwrs3(k):0; - k+=abs(_y[0]); - if(_y[0]<0)i+=ucwrs3(k+1U); - *_k=k; - return i; -} - -/*Returns the index of the given combination of K elements chosen from a set - of size 4 with associated sign bits. - _y: The vector of pulses, whose sum of absolute values is K. - _k: Returns K.*/ -static inline opus_uint32 icwrs4(const int *_y,int *_k){ - opus_uint32 i; - int k; - i=icwrs3(_y+1,&k); - i+=k?ucwrs4(k):0; - k+=abs(_y[0]); - if(_y[0]<0)i+=ucwrs4(k+1); - *_k=k; - return i; -} - -#endif /* SMALL_FOOTPRINT */ - -/*Returns the index of the given combination of K elements chosen from a set - of size _n with associated sign bits. - _y: The vector of pulses, whose sum of absolute values must be _k. - _nc: Returns V(_n,_k).*/ -static inline opus_uint32 icwrs(int _n,int _k,opus_uint32 *_nc,const int *_y, - opus_uint32 *_u){ - opus_uint32 i; - int j; - int k; - /*We can't unroll the first two iterations of the loop unless _n>=2.*/ - celt_assert(_n>=2); - _u[0]=0; - for(k=1;k<=_k+1;k++)_u[k]=(k<<1)-1; - i=icwrs1(_y+_n-1,&k); - j=_n-2; - i+=_u[k]; - k+=abs(_y[j]); - if(_y[j]<0)i+=_u[k+1]; - while(j-->0){ - unext(_u,_k+2,0); - i+=_u[k]; - k+=abs(_y[j]); - if(_y[j]<0)i+=_u[k+1]; - } - *_nc=_u[k]+_u[k+1]; - return i; -} - -#ifdef CUSTOM_MODES -void get_required_bits(opus_int16 *_bits,int _n,int _maxk,int _frac){ - int k; - /*_maxk==0 => there's nothing to do.*/ - celt_assert(_maxk>0); - _bits[0]=0; - if (_n==1) - { - for (k=1;k<=_maxk;k++) - _bits[k] = 1<<_frac; - } - else { - VARDECL(opus_uint32,u); - SAVE_STACK; - ALLOC(u,_maxk+2U,opus_uint32); - ncwrs_urow(_n,_maxk,u); - for(k=1;k<=_maxk;k++) - _bits[k]=log2_frac(u[k]+u[k+1],_frac); - RESTORE_STACK; - } -} -#endif /* CUSTOM_MODES */ - -void encode_pulses(const int *_y,int _n,int _k,ec_enc *_enc){ - opus_uint32 i; - celt_assert(_k>0); -#ifndef SMALL_FOOTPRINT - switch(_n){ - case 2:{ - i=icwrs2(_y,&_k); - ec_enc_uint(_enc,i,ncwrs2(_k)); - }break; - case 3:{ - i=icwrs3(_y,&_k); - ec_enc_uint(_enc,i,ncwrs3(_k)); - }break; - case 4:{ - i=icwrs4(_y,&_k); - ec_enc_uint(_enc,i,ncwrs4(_k)); - }break; - default: - { -#endif - VARDECL(opus_uint32,u); - opus_uint32 nc; - SAVE_STACK; - ALLOC(u,_k+2U,opus_uint32); - i=icwrs(_n,_k,&nc,_y,u); - ec_enc_uint(_enc,i,nc); - RESTORE_STACK; -#ifndef SMALL_FOOTPRINT - } - break; - } -#endif -} - -void decode_pulses(int *_y,int _n,int _k,ec_dec *_dec) -{ - celt_assert(_k>0); -#ifndef SMALL_FOOTPRINT - switch(_n){ - case 2:cwrsi2(_k,ec_dec_uint(_dec,ncwrs2(_k)),_y);break; - case 3:cwrsi3(_k,ec_dec_uint(_dec,ncwrs3(_k)),_y);break; - case 4:cwrsi4(_k,ec_dec_uint(_dec,ncwrs4(_k)),_y);break; - default: - { -#endif - VARDECL(opus_uint32,u); - SAVE_STACK; - ALLOC(u,_k+2U,opus_uint32); - cwrsi(_n,_k,ec_dec_uint(_dec,ncwrs_urow(_n,_k,u)),_y,u); - RESTORE_STACK; -#ifndef SMALL_FOOTPRINT - } - break; - } -#endif -} diff --git a/code/opus-1.0.2/src/opus_multistream.c b/code/opus-1.0.2/src/opus_multistream.c deleted file mode 100644 index a7f25a52..00000000 --- a/code/opus-1.0.2/src/opus_multistream.c +++ /dev/null @@ -1,1027 +0,0 @@ -/* Copyright (c) 2011 Xiph.Org Foundation - Written by Jean-Marc Valin */ -/* - Redistribution and use in source and binary forms, with or without - modification, are permitted provided that the following conditions - are met: - - - Redistributions of source code must retain the above copyright - notice, this list of conditions and the following disclaimer. - - - Redistributions in binary form must reproduce the above copyright - notice, this list of conditions and the following disclaimer in the - documentation and/or other materials provided with the distribution. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER - OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, - EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, - PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR - PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF - LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING - NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -*/ - -#ifdef HAVE_CONFIG_H -#include "config.h" -#endif - -#include "opus_multistream.h" -#include "opus.h" -#include "opus_private.h" -#include "stack_alloc.h" -#include -#include "float_cast.h" -#include "os_support.h" - -typedef struct ChannelLayout { - int nb_channels; - int nb_streams; - int nb_coupled_streams; - unsigned char mapping[256]; -} ChannelLayout; - -struct OpusMSEncoder { - ChannelLayout layout; - int bitrate; - /* Encoder states go here */ -}; - -struct OpusMSDecoder { - ChannelLayout layout; - /* Decoder states go here */ -}; - -#ifdef FIXED_POINT -#define opus_encode_native opus_encode -#else -#define opus_encode_native opus_encode_float -#endif - -static int validate_layout(const ChannelLayout *layout) -{ - int i, max_channel; - - max_channel = layout->nb_streams+layout->nb_coupled_streams; - if (max_channel>255) - return 0; - for (i=0;inb_channels;i++) - { - if (layout->mapping[i] >= max_channel && layout->mapping[i] != 255) - return 0; - } - return 1; -} - - -static int get_left_channel(const ChannelLayout *layout, int stream_id, int prev) -{ - int i; - i = (prev<0) ? 0 : prev+1; - for (;inb_channels;i++) - { - if (layout->mapping[i]==stream_id*2) - return i; - } - return -1; -} - -static int get_right_channel(const ChannelLayout *layout, int stream_id, int prev) -{ - int i; - i = (prev<0) ? 0 : prev+1; - for (;inb_channels;i++) - { - if (layout->mapping[i]==stream_id*2+1) - return i; - } - return -1; -} - -static int get_mono_channel(const ChannelLayout *layout, int stream_id, int prev) -{ - int i; - i = (prev<0) ? 0 : prev+1; - for (;inb_channels;i++) - { - if (layout->mapping[i]==stream_id+layout->nb_coupled_streams) - return i; - } - return -1; -} - -static int validate_encoder_layout(const ChannelLayout *layout) -{ - int s; - for (s=0;snb_streams;s++) - { - if (s < layout->nb_coupled_streams) - { - if (get_left_channel(layout, s, -1)==-1) - return 0; - if (get_right_channel(layout, s, -1)==-1) - return 0; - } else { - if (get_mono_channel(layout, s, -1)==-1) - return 0; - } - } - return 1; -} - -opus_int32 opus_multistream_encoder_get_size(int nb_streams, int nb_coupled_streams) -{ - int coupled_size; - int mono_size; - - if(nb_streams<1||nb_coupled_streams>nb_streams||nb_coupled_streams<0)return 0; - coupled_size = opus_encoder_get_size(2); - mono_size = opus_encoder_get_size(1); - return align(sizeof(OpusMSEncoder)) - + nb_coupled_streams * align(coupled_size) - + (nb_streams-nb_coupled_streams) * align(mono_size); -} - - - -int opus_multistream_encoder_init( - OpusMSEncoder *st, - opus_int32 Fs, - int channels, - int streams, - int coupled_streams, - const unsigned char *mapping, - int application -) -{ - int coupled_size; - int mono_size; - int i, ret; - char *ptr; - - if ((channels>255) || (channels<1) || (coupled_streams>streams) || - (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) - return OPUS_BAD_ARG; - - st->layout.nb_channels = channels; - st->layout.nb_streams = streams; - st->layout.nb_coupled_streams = coupled_streams; - - for (i=0;ilayout.nb_channels;i++) - st->layout.mapping[i] = mapping[i]; - if (!validate_layout(&st->layout) || !validate_encoder_layout(&st->layout)) - return OPUS_BAD_ARG; - ptr = (char*)st + align(sizeof(OpusMSEncoder)); - coupled_size = opus_encoder_get_size(2); - mono_size = opus_encoder_get_size(1); - - for (i=0;ilayout.nb_coupled_streams;i++) - { - ret = opus_encoder_init((OpusEncoder*)ptr, Fs, 2, application); - if(ret!=OPUS_OK)return ret; - ptr += align(coupled_size); - } - for (;ilayout.nb_streams;i++) - { - ret = opus_encoder_init((OpusEncoder*)ptr, Fs, 1, application); - if(ret!=OPUS_OK)return ret; - ptr += align(mono_size); - } - return OPUS_OK; -} - -OpusMSEncoder *opus_multistream_encoder_create( - opus_int32 Fs, - int channels, - int streams, - int coupled_streams, - const unsigned char *mapping, - int application, - int *error -) -{ - int ret; - OpusMSEncoder *st; - if ((channels>255) || (channels<1) || (coupled_streams>streams) || - (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) - { - if (error) - *error = OPUS_BAD_ARG; - return NULL; - } - st = (OpusMSEncoder *)opus_alloc(opus_multistream_encoder_get_size(streams, coupled_streams)); - if (st==NULL) - { - if (error) - *error = OPUS_ALLOC_FAIL; - return NULL; - } - ret = opus_multistream_encoder_init(st, Fs, channels, streams, coupled_streams, mapping, application); - if (ret != OPUS_OK) - { - opus_free(st); - st = NULL; - } - if (error) - *error = ret; - return st; -} - -typedef void (*opus_copy_channel_in_func)( - opus_val16 *dst, - int dst_stride, - const void *src, - int src_stride, - int src_channel, - int frame_size -); - -/* Max size in case the encoder decides to return three frames */ -#define MS_FRAME_TMP (3*1275+7) -static int opus_multistream_encode_native -( - OpusMSEncoder *st, - opus_copy_channel_in_func copy_channel_in, - const void *pcm, - int frame_size, - unsigned char *data, - opus_int32 max_data_bytes -) -{ - opus_int32 Fs; - int coupled_size; - int mono_size; - int s; - char *ptr; - int tot_size; - VARDECL(opus_val16, buf); - unsigned char tmp_data[MS_FRAME_TMP]; - OpusRepacketizer rp; - ALLOC_STACK; - - ptr = (char*)st + align(sizeof(OpusMSEncoder)); - opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_SAMPLE_RATE(&Fs)); - /* Validate frame_size before using it to allocate stack space. - This mirrors the checks in opus_encode[_float](). */ - if (400*frame_size != Fs && 200*frame_size != Fs && - 100*frame_size != Fs && 50*frame_size != Fs && - 25*frame_size != Fs && 50*frame_size != 3*Fs) - { - RESTORE_STACK; - return OPUS_BAD_ARG; - } - ALLOC(buf, 2*frame_size, opus_val16); - coupled_size = opus_encoder_get_size(2); - mono_size = opus_encoder_get_size(1); - - if (max_data_bytes < 4*st->layout.nb_streams-1) - { - RESTORE_STACK; - return OPUS_BUFFER_TOO_SMALL; - } - /* Counting ToC */ - tot_size = 0; - for (s=0;slayout.nb_streams;s++) - { - OpusEncoder *enc; - int len; - int curr_max; - - opus_repacketizer_init(&rp); - enc = (OpusEncoder*)ptr; - if (s < st->layout.nb_coupled_streams) - { - int left, right; - left = get_left_channel(&st->layout, s, -1); - right = get_right_channel(&st->layout, s, -1); - (*copy_channel_in)(buf, 2, - pcm, st->layout.nb_channels, left, frame_size); - (*copy_channel_in)(buf+1, 2, - pcm, st->layout.nb_channels, right, frame_size); - ptr += align(coupled_size); - } else { - int chan = get_mono_channel(&st->layout, s, -1); - (*copy_channel_in)(buf, 1, - pcm, st->layout.nb_channels, chan, frame_size); - ptr += align(mono_size); - } - /* number of bytes left (+Toc) */ - curr_max = max_data_bytes - tot_size; - /* Reserve three bytes for the last stream and four for the others */ - curr_max -= IMAX(0,4*(st->layout.nb_streams-s-1)-1); - curr_max = IMIN(curr_max,MS_FRAME_TMP); - len = opus_encode_native(enc, buf, frame_size, tmp_data, curr_max); - if (len<0) - { - RESTORE_STACK; - return len; - } - /* We need to use the repacketizer to add the self-delimiting lengths - while taking into account the fact that the encoder can now return - more than one frame at a time (e.g. 60 ms CELT-only) */ - opus_repacketizer_cat(&rp, tmp_data, len); - len = opus_repacketizer_out_range_impl(&rp, 0, opus_repacketizer_get_nb_frames(&rp), data, max_data_bytes-tot_size, s != st->layout.nb_streams-1); - data += len; - tot_size += len; - } - RESTORE_STACK; - return tot_size; - -} - -#if !defined(DISABLE_FLOAT_API) -static void opus_copy_channel_in_float( - opus_val16 *dst, - int dst_stride, - const void *src, - int src_stride, - int src_channel, - int frame_size -) -{ - const float *float_src; - int i; - float_src = (const float *)src; - for (i=0;ilayout.nb_streams + st->layout.nb_coupled_streams; - value /= chan; - for (s=0;slayout.nb_streams;s++) - { - OpusEncoder *enc; - enc = (OpusEncoder*)ptr; - if (s < st->layout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - opus_encoder_ctl(enc, request, value * (s < st->layout.nb_coupled_streams ? 2 : 1)); - } - } - break; - case OPUS_GET_BITRATE_REQUEST: - { - int s; - opus_int32 *value = va_arg(ap, opus_int32*); - *value = 0; - for (s=0;slayout.nb_streams;s++) - { - opus_int32 rate; - OpusEncoder *enc; - enc = (OpusEncoder*)ptr; - if (s < st->layout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - opus_encoder_ctl(enc, request, &rate); - *value += rate; - } - } - break; - case OPUS_GET_LSB_DEPTH_REQUEST: - case OPUS_GET_VBR_REQUEST: - case OPUS_GET_APPLICATION_REQUEST: - case OPUS_GET_BANDWIDTH_REQUEST: - case OPUS_GET_COMPLEXITY_REQUEST: - case OPUS_GET_PACKET_LOSS_PERC_REQUEST: - case OPUS_GET_DTX_REQUEST: - case OPUS_GET_VOICE_RATIO_REQUEST: - case OPUS_GET_VBR_CONSTRAINT_REQUEST: - case OPUS_GET_SIGNAL_REQUEST: - case OPUS_GET_LOOKAHEAD_REQUEST: - case OPUS_GET_SAMPLE_RATE_REQUEST: - case OPUS_GET_INBAND_FEC_REQUEST: - { - OpusEncoder *enc; - /* For int32* GET params, just query the first stream */ - opus_int32 *value = va_arg(ap, opus_int32*); - enc = (OpusEncoder*)ptr; - ret = opus_encoder_ctl(enc, request, value); - } - break; - case OPUS_GET_FINAL_RANGE_REQUEST: - { - int s; - opus_uint32 *value = va_arg(ap, opus_uint32*); - opus_uint32 tmp; - *value=0; - for (s=0;slayout.nb_streams;s++) - { - OpusEncoder *enc; - enc = (OpusEncoder*)ptr; - if (s < st->layout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - ret = opus_encoder_ctl(enc, request, &tmp); - if (ret != OPUS_OK) break; - *value ^= tmp; - } - } - break; - case OPUS_SET_LSB_DEPTH_REQUEST: - case OPUS_SET_COMPLEXITY_REQUEST: - case OPUS_SET_VBR_REQUEST: - case OPUS_SET_VBR_CONSTRAINT_REQUEST: - case OPUS_SET_BANDWIDTH_REQUEST: - case OPUS_SET_SIGNAL_REQUEST: - case OPUS_SET_APPLICATION_REQUEST: - case OPUS_SET_INBAND_FEC_REQUEST: - case OPUS_SET_PACKET_LOSS_PERC_REQUEST: - case OPUS_SET_DTX_REQUEST: - case OPUS_SET_FORCE_MODE_REQUEST: - { - int s; - /* This works for int32 params */ - opus_int32 value = va_arg(ap, opus_int32); - for (s=0;slayout.nb_streams;s++) - { - OpusEncoder *enc; - - enc = (OpusEncoder*)ptr; - if (s < st->layout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - ret = opus_encoder_ctl(enc, request, value); - if (ret != OPUS_OK) - break; - } - } - break; - case OPUS_MULTISTREAM_GET_ENCODER_STATE_REQUEST: - { - int s; - opus_int32 stream_id; - OpusEncoder **value; - stream_id = va_arg(ap, opus_int32); - if (stream_id<0 || stream_id >= st->layout.nb_streams) - ret = OPUS_BAD_ARG; - value = va_arg(ap, OpusEncoder**); - for (s=0;slayout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - } - *value = (OpusEncoder*)ptr; - } - break; - default: - ret = OPUS_UNIMPLEMENTED; - break; - } - - va_end(ap); - return ret; -} - -void opus_multistream_encoder_destroy(OpusMSEncoder *st) -{ - opus_free(st); -} - - -/* DECODER */ - -opus_int32 opus_multistream_decoder_get_size(int nb_streams, int nb_coupled_streams) -{ - int coupled_size; - int mono_size; - - if(nb_streams<1||nb_coupled_streams>nb_streams||nb_coupled_streams<0)return 0; - coupled_size = opus_decoder_get_size(2); - mono_size = opus_decoder_get_size(1); - return align(sizeof(OpusMSDecoder)) - + nb_coupled_streams * align(coupled_size) - + (nb_streams-nb_coupled_streams) * align(mono_size); -} - -int opus_multistream_decoder_init( - OpusMSDecoder *st, - opus_int32 Fs, - int channels, - int streams, - int coupled_streams, - const unsigned char *mapping -) -{ - int coupled_size; - int mono_size; - int i, ret; - char *ptr; - - if ((channels>255) || (channels<1) || (coupled_streams>streams) || - (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) - return OPUS_BAD_ARG; - - st->layout.nb_channels = channels; - st->layout.nb_streams = streams; - st->layout.nb_coupled_streams = coupled_streams; - - for (i=0;ilayout.nb_channels;i++) - st->layout.mapping[i] = mapping[i]; - if (!validate_layout(&st->layout)) - return OPUS_BAD_ARG; - - ptr = (char*)st + align(sizeof(OpusMSDecoder)); - coupled_size = opus_decoder_get_size(2); - mono_size = opus_decoder_get_size(1); - - for (i=0;ilayout.nb_coupled_streams;i++) - { - ret=opus_decoder_init((OpusDecoder*)ptr, Fs, 2); - if(ret!=OPUS_OK)return ret; - ptr += align(coupled_size); - } - for (;ilayout.nb_streams;i++) - { - ret=opus_decoder_init((OpusDecoder*)ptr, Fs, 1); - if(ret!=OPUS_OK)return ret; - ptr += align(mono_size); - } - return OPUS_OK; -} - - -OpusMSDecoder *opus_multistream_decoder_create( - opus_int32 Fs, - int channels, - int streams, - int coupled_streams, - const unsigned char *mapping, - int *error -) -{ - int ret; - OpusMSDecoder *st; - if ((channels>255) || (channels<1) || (coupled_streams>streams) || - (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) - { - if (error) - *error = OPUS_BAD_ARG; - return NULL; - } - st = (OpusMSDecoder *)opus_alloc(opus_multistream_decoder_get_size(streams, coupled_streams)); - if (st==NULL) - { - if (error) - *error = OPUS_ALLOC_FAIL; - return NULL; - } - ret = opus_multistream_decoder_init(st, Fs, channels, streams, coupled_streams, mapping); - if (error) - *error = ret; - if (ret != OPUS_OK) - { - opus_free(st); - st = NULL; - } - return st; -} - -typedef void (*opus_copy_channel_out_func)( - void *dst, - int dst_stride, - int dst_channel, - const opus_val16 *src, - int src_stride, - int frame_size -); - -static int opus_multistream_decode_native( - OpusMSDecoder *st, - const unsigned char *data, - opus_int32 len, - void *pcm, - opus_copy_channel_out_func copy_channel_out, - int frame_size, - int decode_fec -) -{ - opus_int32 Fs; - int coupled_size; - int mono_size; - int s, c; - char *ptr; - int do_plc=0; - VARDECL(opus_val16, buf); - ALLOC_STACK; - - /* Limit frame_size to avoid excessive stack allocations. */ - opus_multistream_decoder_ctl(st, OPUS_GET_SAMPLE_RATE(&Fs)); - frame_size = IMIN(frame_size, Fs/25*3); - ALLOC(buf, 2*frame_size, opus_val16); - ptr = (char*)st + align(sizeof(OpusMSDecoder)); - coupled_size = opus_decoder_get_size(2); - mono_size = opus_decoder_get_size(1); - - if (len==0) - do_plc = 1; - if (len < 0) - return OPUS_BAD_ARG; - if (!do_plc && len < 2*st->layout.nb_streams-1) - return OPUS_INVALID_PACKET; - for (s=0;slayout.nb_streams;s++) - { - OpusDecoder *dec; - int packet_offset, ret; - - dec = (OpusDecoder*)ptr; - ptr += (s < st->layout.nb_coupled_streams) ? align(coupled_size) : align(mono_size); - - if (!do_plc && len<=0) - { - RESTORE_STACK; - return OPUS_INVALID_PACKET; - } - packet_offset = 0; - ret = opus_decode_native(dec, data, len, buf, frame_size, decode_fec, s!=st->layout.nb_streams-1, &packet_offset); - data += packet_offset; - len -= packet_offset; - if (ret > frame_size) - { - RESTORE_STACK; - return OPUS_BUFFER_TOO_SMALL; - } - if (s>0 && ret != frame_size) - { - RESTORE_STACK; - return OPUS_INVALID_PACKET; - } - if (ret <= 0) - { - RESTORE_STACK; - return ret; - } - frame_size = ret; - if (s < st->layout.nb_coupled_streams) - { - int chan, prev; - prev = -1; - /* Copy "left" audio to the channel(s) where it belongs */ - while ( (chan = get_left_channel(&st->layout, s, prev)) != -1) - { - (*copy_channel_out)(pcm, st->layout.nb_channels, chan, - buf, 2, frame_size); - prev = chan; - } - prev = -1; - /* Copy "right" audio to the channel(s) where it belongs */ - while ( (chan = get_right_channel(&st->layout, s, prev)) != -1) - { - (*copy_channel_out)(pcm, st->layout.nb_channels, chan, - buf+1, 2, frame_size); - prev = chan; - } - } else { - int chan, prev; - prev = -1; - /* Copy audio to the channel(s) where it belongs */ - while ( (chan = get_mono_channel(&st->layout, s, prev)) != -1) - { - (*copy_channel_out)(pcm, st->layout.nb_channels, chan, - buf, 1, frame_size); - prev = chan; - } - } - } - /* Handle muted channels */ - for (c=0;clayout.nb_channels;c++) - { - if (st->layout.mapping[c] == 255) - { - (*copy_channel_out)(pcm, st->layout.nb_channels, c, - NULL, 0, frame_size); - } - } - RESTORE_STACK; - return frame_size; -} - -#if !defined(DISABLE_FLOAT_API) -static void opus_copy_channel_out_float( - void *dst, - int dst_stride, - int dst_channel, - const opus_val16 *src, - int src_stride, - int frame_size -) -{ - float *float_dst; - int i; - float_dst = (float*)dst; - if (src != NULL) - { - for (i=0;ilayout.nb_streams;s++) - { - OpusDecoder *dec; - dec = (OpusDecoder*)ptr; - if (s < st->layout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - ret = opus_decoder_ctl(dec, request, &tmp); - if (ret != OPUS_OK) break; - *value ^= tmp; - } - } - break; - case OPUS_RESET_STATE: - { - int s; - for (s=0;slayout.nb_streams;s++) - { - OpusDecoder *dec; - - dec = (OpusDecoder*)ptr; - if (s < st->layout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - ret = opus_decoder_ctl(dec, OPUS_RESET_STATE); - if (ret != OPUS_OK) - break; - } - } - break; - case OPUS_MULTISTREAM_GET_DECODER_STATE_REQUEST: - { - int s; - opus_int32 stream_id; - OpusDecoder **value; - stream_id = va_arg(ap, opus_int32); - if (stream_id<0 || stream_id >= st->layout.nb_streams) - ret = OPUS_BAD_ARG; - value = va_arg(ap, OpusDecoder**); - for (s=0;slayout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - } - *value = (OpusDecoder*)ptr; - } - break; - case OPUS_SET_GAIN_REQUEST: - { - int s; - /* This works for int32 params */ - opus_int32 value = va_arg(ap, opus_int32); - for (s=0;slayout.nb_streams;s++) - { - OpusDecoder *dec; - - dec = (OpusDecoder*)ptr; - if (s < st->layout.nb_coupled_streams) - ptr += align(coupled_size); - else - ptr += align(mono_size); - ret = opus_decoder_ctl(dec, request, value); - if (ret != OPUS_OK) - break; - } - } - break; - default: - ret = OPUS_UNIMPLEMENTED; - break; - } - - va_end(ap); - return ret; -} - - -void opus_multistream_decoder_destroy(OpusMSDecoder *st) -{ - opus_free(st); -} diff --git a/code/opus-1.0.2/celt/_kiss_fft_guts.h b/code/opus-1.1/celt/_kiss_fft_guts.h similarity index 97% rename from code/opus-1.0.2/celt/_kiss_fft_guts.h rename to code/opus-1.1/celt/_kiss_fft_guts.h index 33e62c6b..aefe490e 100644 --- a/code/opus-1.0.2/celt/_kiss_fft_guts.h +++ b/code/opus-1.1/celt/_kiss_fft_guts.h @@ -94,6 +94,14 @@ do {(res).r = ADD32((res).r,(a).r); (res).i = SUB32((res).i,(a).i); \ }while(0) +#if defined(OPUS_ARM_INLINE_ASM) +#include "arm/kiss_fft_armv4.h" +#endif + +#if defined(OPUS_ARM_INLINE_EDSP) +#include "arm/kiss_fft_armv5e.h" +#endif + #else /* not FIXED_POINT*/ # define S_MUL(a,b) ( (a)*(b) ) diff --git a/code/opus-1.0.2/celt/arch.h b/code/opus-1.1/celt/arch.h similarity index 94% rename from code/opus-1.0.2/celt/arch.h rename to code/opus-1.1/celt/arch.h index 03cda40f..3bbcd366 100644 --- a/code/opus-1.0.2/celt/arch.h +++ b/code/opus-1.1/celt/arch.h @@ -35,6 +35,7 @@ #define ARCH_H #include "opus_types.h" +#include "opus_defines.h" # if !defined(__GNUC_PREREQ) # if defined(__GNUC__)&&defined(__GNUC_MINOR__) @@ -54,7 +55,7 @@ #ifdef __GNUC__ __attribute__((noreturn)) #endif -static inline void _celt_fatal(const char *str, const char *file, int line) +static OPUS_INLINE void _celt_fatal(const char *str, const char *file, int line) { fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str); abort(); @@ -100,6 +101,7 @@ typedef opus_val32 celt_ener; #define DB_SHIFT 10 #define EPSILON 1 +#define VERY_SMALL 0 #define VERY_LARGE16 ((opus_val16)32767) #define Q15_ONE ((opus_val16)32767) @@ -112,10 +114,10 @@ typedef opus_val32 celt_ener; #include "fixed_generic.h" -#ifdef ARM5E_ASM -#include "fixed_arm5e.h" -#elif defined (ARM4_ASM) -#include "fixed_arm4.h" +#ifdef OPUS_ARM_INLINE_EDSP +#include "arm/fixed_armv5e.h" +#elif defined (OPUS_ARM_INLINE_ASM) +#include "arm/fixed_armv4.h" #elif defined (BFIN_ASM) #include "fixed_bfin.h" #elif defined (TI_C5X_ASM) @@ -140,6 +142,7 @@ typedef float celt_ener; #define NORM_SCALING 1.f #define EPSILON 1e-15f +#define VERY_SMALL 1e-30f #define VERY_LARGE16 1e15f #define Q15_ONE ((opus_val16)1.f) @@ -161,6 +164,7 @@ typedef float celt_ener; #define SHR(a,shift) (a) #define SHL(a,shift) (a) #define SATURATE(x,a) (x) +#define SATURATE16(x) (x) #define ROUND16(a,shift) (a) #define HALF16(x) (.5f*(x)) @@ -182,6 +186,7 @@ typedef float celt_ener; #define MAC16_32_Q15(c,a,b) ((c)+(a)*(b)) #define MULT16_16_Q11_32(a,b) ((a)*(b)) +#define MULT16_16_Q11(a,b) ((a)*(b)) #define MULT16_16_Q13(a,b) ((a)*(b)) #define MULT16_16_Q14(a,b) ((a)*(b)) #define MULT16_16_Q15(a,b) ((a)*(b)) diff --git a/code/opus-1.1/celt/arm/arm2gnu.pl b/code/opus-1.1/celt/arm/arm2gnu.pl new file mode 100755 index 00000000..eab42efa --- /dev/null +++ b/code/opus-1.1/celt/arm/arm2gnu.pl @@ -0,0 +1,316 @@ +#!/usr/bin/perl + +my $bigend; # little/big endian +my $nxstack; + +$nxstack = 0; + +eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}' + if $running_under_some_shell; + +while ($ARGV[0] =~ /^-/) { + $_ = shift; + last if /^--/; + if (/^-n/) { + $nflag++; + next; + } + die "I don't recognize this switch: $_\\n"; +} +$printit++ unless $nflag; + +$\ = "\n"; # automatically add newline on print +$n=0; + +$thumb = 0; # ARM mode by default, not Thumb. +@proc_stack = (); + +LINE: +while (<>) { + + # For ADRLs we need to add a new line after the substituted one. + $addPadding = 0; + + # First, we do not dare to touch *anything* inside double quotes, do we? + # Second, if you want a dollar character in the string, + # insert two of them -- that's how ARM C and assembler treat strings. + s/^([A-Za-z_]\w*)[ \t]+DCB[ \t]*\"/$1: .ascii \"/ && do { s/\$\$/\$/g; next }; + s/\bDCB\b[ \t]*\"/.ascii \"/ && do { s/\$\$/\$/g; next }; + s/^(\S+)\s+RN\s+(\S+)/$1 .req r$2/ && do { s/\$\$/\$/g; next }; + # If there's nothing on a line but a comment, don't try to apply any further + # substitutions (this is a cheap hack to avoid mucking up the license header) + s/^([ \t]*);/$1@/ && do { s/\$\$/\$/g; next }; + # If substituted -- leave immediately ! + + s/@/,:/; + s/;/@/; + while ( /@.*'/ ) { + s/(@.*)'/$1/g; + } + s/\{FALSE\}/0/g; + s/\{TRUE\}/1/g; + s/\{(\w\w\w\w+)\}/$1/g; + s/\bINCLUDE[ \t]*([^ \t\n]+)/.include \"$1\"/; + s/\bGET[ \t]*([^ \t\n]+)/.include \"${ my $x=$1; $x =~ s|\.s|-gnu.S|; \$x }\"/; + s/\bIMPORT\b/.extern/; + s/\bEXPORT\b/.global/; + s/^(\s+)\[/$1IF/; + s/^(\s+)\|/$1ELSE/; + s/^(\s+)\]/$1ENDIF/; + s/IF *:DEF:/ .ifdef/; + s/IF *:LNOT: *:DEF:/ .ifndef/; + s/ELSE/ .else/; + s/ENDIF/ .endif/; + + if( /\bIF\b/ ) { + s/\bIF\b/ .if/; + s/=/==/; + } + if ( $n == 2) { + s/\$/\\/g; + } + if ($n == 1) { + s/\$//g; + s/label//g; + $n = 2; + } + if ( /MACRO/ ) { + s/MACRO *\n/.macro/; + $n=1; + } + if ( /\bMEND\b/ ) { + s/\bMEND\b/.endm/; + $n=0; + } + + # ".rdata" doesn't work in 'as' version 2.13.2, as it is ".rodata" there. + # + if ( /\bAREA\b/ ) { + my $align; + $align = "2"; + if ( /ALIGN=(\d+)/ ) { + $align = $1; + } + if ( /CODE/ ) { + $nxstack = 1; + } + s/^(.+)CODE(.+)READONLY(.*)/ .text/; + s/^(.+)DATA(.+)READONLY(.*)/ .section .rdata/; + s/^(.+)\|\|\.data\|\|(.+)/ .data/; + s/^(.+)\|\|\.bss\|\|(.+)/ .bss/; + s/$/; .p2align $align/; + # Enable NEON instructions but don't produce a binary that requires + # ARMv7. RVCT does not have equivalent directives, so we just do this + # for all CODE areas. + if ( /.text/ ) { + # Separating .arch, .fpu, etc., by semicolons does not work (gas + # thinks the semicolon is part of the arch name, even when there's + # whitespace separating them). Sadly this means our line numbers + # won't match the original source file (we could use the .line + # directive, which is documented to be obsolete, but then gdb will + # show the wrong line in the translated source file). + s/$/; .arch armv7-a\n .fpu neon\n .object_arch armv4t/; + } + } + + s/\|\|\.constdata\$(\d+)\|\|/.L_CONST$1/; # ||.constdata$3|| + s/\|\|\.bss\$(\d+)\|\|/.L_BSS$1/; # ||.bss$2|| + s/\|\|\.data\$(\d+)\|\|/.L_DATA$1/; # ||.data$2|| + s/\|\|([a-zA-Z0-9_]+)\@([a-zA-Z0-9_]+)\|\|/@ $&/; + s/^(\s+)\%(\s)/ .space $1/; + + s/\|(.+)\.(\d+)\|/\.$1_$2/; # |L80.123| -> .L80_123 + s/\bCODE32\b/.code 32/ && do {$thumb = 0}; + s/\bCODE16\b/.code 16/ && do {$thumb = 1}; + if (/\bPROC\b/) + { + my $prefix; + my $proc; + /^([A-Za-z_\.]\w+)\b/; + $proc = $1; + $prefix = ""; + if ($proc) + { + $prefix = $prefix.sprintf("\t.type\t%s, %%function; ",$proc); + push(@proc_stack, $proc); + s/^[A-Za-z_\.]\w+/$&:/; + } + $prefix = $prefix."\t.thumb_func; " if ($thumb); + s/\bPROC\b/@ $&/; + $_ = $prefix.$_; + } + s/^(\s*)(S|Q|SH|U|UQ|UH)ASX\b/$1$2ADDSUBX/; + s/^(\s*)(S|Q|SH|U|UQ|UH)SAX\b/$1$2SUBADDX/; + if (/\bENDP\b/) + { + my $proc; + s/\bENDP\b/@ $&/; + $proc = pop(@proc_stack); + $_ = "\t.size $proc, .-$proc".$_ if ($proc); + } + s/\bSUBT\b/@ $&/; + s/\bDATA\b/@ $&/; # DATA directive is deprecated -- Asm guide, p.7-25 + s/\bKEEP\b/@ $&/; + s/\bEXPORTAS\b/@ $&/; + s/\|\|(.)+\bEQU\b/@ $&/; + s/\|\|([\w\$]+)\|\|/$1/; + s/\bENTRY\b/@ $&/; + s/\bASSERT\b/@ $&/; + s/\bGBLL\b/@ $&/; + s/\bGBLA\b/@ $&/; + s/^\W+OPT\b/@ $&/; + s/:OR:/|/g; + s/:SHL:/<>/g; + s/:AND:/&/g; + s/:LAND:/&&/g; + s/CPSR/cpsr/; + s/SPSR/spsr/; + s/ALIGN$/.balign 4/; + s/ALIGN\s+([0-9x]+)$/.balign $1/; + s/psr_cxsf/psr_all/; + s/LTORG/.ltorg/; + s/^([A-Za-z_]\w*)[ \t]+EQU/ .set $1,/; + s/^([A-Za-z_]\w*)[ \t]+SETL/ .set $1,/; + s/^([A-Za-z_]\w*)[ \t]+SETA/ .set $1,/; + s/^([A-Za-z_]\w*)[ \t]+\*/ .set $1,/; + + # {PC} + 0xdeadfeed --> . + 0xdeadfeed + s/\{PC\} \+/ \. +/; + + # Single hex constant on the line ! + # + # >>> NOTE <<< + # Double-precision floats in gcc are always mixed-endian, which means + # bytes in two words are little-endian, but words are big-endian. + # So, 0x0000deadfeed0000 would be stored as 0x0000dead at low address + # and 0xfeed0000 at high address. + # + s/\bDCFD\b[ \t]+0x([a-fA-F0-9]{8})([a-fA-F0-9]{8})/.long 0x$1, 0x$2/; + # Only decimal constants on the line, no hex ! + s/\bDCFD\b[ \t]+([0-9\.\-]+)/.double $1/; + + # Single hex constant on the line ! +# s/\bDCFS\b[ \t]+0x([a-f0-9]{8})([a-f0-9]{8})/.long 0x$1, 0x$2/; + # Only decimal constants on the line, no hex ! +# s/\bDCFS\b[ \t]+([0-9\.\-]+)/.double $1/; + s/\bDCFS[ \t]+0x/.word 0x/; + s/\bDCFS\b/.float/; + + s/^([A-Za-z_]\w*)[ \t]+DCD/$1 .word/; + s/\bDCD\b/.word/; + s/^([A-Za-z_]\w*)[ \t]+DCW/$1 .short/; + s/\bDCW\b/.short/; + s/^([A-Za-z_]\w*)[ \t]+DCB/$1 .byte/; + s/\bDCB\b/.byte/; + s/^([A-Za-z_]\w*)[ \t]+\%/.comm $1,/; + s/^[A-Za-z_\.]\w+/$&:/; + s/^(\d+)/$1:/; + s/\%(\d+)/$1b_or_f/; + s/\%[Bb](\d+)/$1b/; + s/\%[Ff](\d+)/$1f/; + s/\%[Ff][Tt](\d+)/$1f/; + s/&([\dA-Fa-f]+)/0x$1/; + if ( /\b2_[01]+\b/ ) { + s/\b2_([01]+)\b/conv$1&&&&/g; + while ( /[01][01][01][01]&&&&/ ) { + s/0000&&&&/&&&&0/g; + s/0001&&&&/&&&&1/g; + s/0010&&&&/&&&&2/g; + s/0011&&&&/&&&&3/g; + s/0100&&&&/&&&&4/g; + s/0101&&&&/&&&&5/g; + s/0110&&&&/&&&&6/g; + s/0111&&&&/&&&&7/g; + s/1000&&&&/&&&&8/g; + s/1001&&&&/&&&&9/g; + s/1010&&&&/&&&&A/g; + s/1011&&&&/&&&&B/g; + s/1100&&&&/&&&&C/g; + s/1101&&&&/&&&&D/g; + s/1110&&&&/&&&&E/g; + s/1111&&&&/&&&&F/g; + } + s/000&&&&/&&&&0/g; + s/001&&&&/&&&&1/g; + s/010&&&&/&&&&2/g; + s/011&&&&/&&&&3/g; + s/100&&&&/&&&&4/g; + s/101&&&&/&&&&5/g; + s/110&&&&/&&&&6/g; + s/111&&&&/&&&&7/g; + s/00&&&&/&&&&0/g; + s/01&&&&/&&&&1/g; + s/10&&&&/&&&&2/g; + s/11&&&&/&&&&3/g; + s/0&&&&/&&&&0/g; + s/1&&&&/&&&&1/g; + s/conv&&&&/0x/g; + } + + if ( /commandline/) + { + if( /-bigend/) + { + $bigend=1; + } + } + + if ( /\bDCDU\b/ ) + { + my $cmd=$_; + my $value; + my $prefix; + my $w1; + my $w2; + my $w3; + my $w4; + + s/\s+DCDU\b/@ $&/; + + $cmd =~ /\bDCDU\b\s+0x(\d+)/; + $value = $1; + $value =~ /(\w\w)(\w\w)(\w\w)(\w\w)/; + $w1 = $1; + $w2 = $2; + $w3 = $3; + $w4 = $4; + + if( $bigend ne "") + { + # big endian + $prefix = "\t.byte\t0x".$w1.";". + "\t.byte\t0x".$w2.";". + "\t.byte\t0x".$w3.";". + "\t.byte\t0x".$w4."; "; + } + else + { + # little endian + $prefix = "\t.byte\t0x".$w4.";". + "\t.byte\t0x".$w3.";". + "\t.byte\t0x".$w2.";". + "\t.byte\t0x".$w1."; "; + } + $_=$prefix.$_; + } + + if ( /\badrl\b/i ) + { + s/\badrl\s+(\w+)\s*,\s*(\w+)/ldr $1,=$2/i; + $addPadding = 1; + } + s/\bEND\b/@ END/; +} continue { + printf ("%s", $_) if $printit; + if ($addPadding != 0) + { + printf (" mov r0,r0\n"); + $addPadding = 0; + } +} +#If we had a code section, mark that this object doesn't need an executable +# stack. +if ($nxstack) { + printf (" .section\t.note.GNU-stack,\"\",\%\%progbits\n"); +} diff --git a/code/opus-1.1/celt/arm/arm_celt_map.c b/code/opus-1.1/celt/arm/arm_celt_map.c new file mode 100644 index 00000000..547a84d1 --- /dev/null +++ b/code/opus-1.1/celt/arm/arm_celt_map.c @@ -0,0 +1,49 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "pitch.h" + +#if defined(OPUS_HAVE_RTCD) + +# if defined(FIXED_POINT) +opus_val32 (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *, + const opus_val16 *, opus_val32 *, int , int) = { + celt_pitch_xcorr_c, /* ARMv4 */ + MAY_HAVE_EDSP(celt_pitch_xcorr), /* EDSP */ + MAY_HAVE_MEDIA(celt_pitch_xcorr), /* Media */ + MAY_HAVE_NEON(celt_pitch_xcorr) /* NEON */ +}; +# else +# error "Floating-point implementation is not supported by ARM asm yet." \ + "Reconfigure with --disable-rtcd or send patches." +# endif + +#endif diff --git a/code/opus-1.1/celt/arm/armcpu.c b/code/opus-1.1/celt/arm/armcpu.c new file mode 100644 index 00000000..17685258 --- /dev/null +++ b/code/opus-1.1/celt/arm/armcpu.c @@ -0,0 +1,174 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* Original code from libtheora modified to suit to Opus */ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#ifdef OPUS_HAVE_RTCD + +#include "armcpu.h" +#include "cpu_support.h" +#include "os_support.h" +#include "opus_types.h" + +#define OPUS_CPU_ARM_V4 (1) +#define OPUS_CPU_ARM_EDSP (1<<1) +#define OPUS_CPU_ARM_MEDIA (1<<2) +#define OPUS_CPU_ARM_NEON (1<<3) + +#if defined(_MSC_VER) +/*For GetExceptionCode() and EXCEPTION_ILLEGAL_INSTRUCTION.*/ +# define WIN32_LEAN_AND_MEAN +# define WIN32_EXTRA_LEAN +# include + +static OPUS_INLINE opus_uint32 opus_cpu_capabilities(void){ + opus_uint32 flags; + flags=0; + /* MSVC has no OPUS_INLINE __asm support for ARM, but it does let you __emit + * instructions via their assembled hex code. + * All of these instructions should be essentially nops. */ +# if defined(OPUS_ARM_MAY_HAVE_EDSP) + __try{ + /*PLD [r13]*/ + __emit(0xF5DDF000); + flags|=OPUS_CPU_ARM_EDSP; + } + __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ + /*Ignore exception.*/ + } +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) + __try{ + /*SHADD8 r3,r3,r3*/ + __emit(0xE6333F93); + flags|=OPUS_CPU_ARM_MEDIA; + } + __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ + /*Ignore exception.*/ + } +# if defined(OPUS_ARM_MAY_HAVE_NEON) + __try{ + /*VORR q0,q0,q0*/ + __emit(0xF2200150); + flags|=OPUS_CPU_ARM_NEON; + } + __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ + /*Ignore exception.*/ + } +# endif +# endif +# endif + return flags; +} + +#elif defined(__linux__) +/* Linux based */ +opus_uint32 opus_cpu_capabilities(void) +{ + opus_uint32 flags = 0; + FILE *cpuinfo; + + /* Reading /proc/self/auxv would be easier, but that doesn't work reliably on + * Android */ + cpuinfo = fopen("/proc/cpuinfo", "r"); + + if(cpuinfo != NULL) + { + /* 512 should be enough for anybody (it's even enough for all the flags that + * x86 has accumulated... so far). */ + char buf[512]; + + while(fgets(buf, 512, cpuinfo) != NULL) + { +# if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_NEON) + /* Search for edsp and neon flag */ + if(memcmp(buf, "Features", 8) == 0) + { + char *p; +# if defined(OPUS_ARM_MAY_HAVE_EDSP) + p = strstr(buf, " edsp"); + if(p != NULL && (p[5] == ' ' || p[5] == '\n')) + flags |= OPUS_CPU_ARM_EDSP; +# endif + +# if defined(OPUS_ARM_MAY_HAVE_NEON) + p = strstr(buf, " neon"); + if(p != NULL && (p[5] == ' ' || p[5] == '\n')) + flags |= OPUS_CPU_ARM_NEON; +# endif + } +# endif + +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) + /* Search for media capabilities (>= ARMv6) */ + if(memcmp(buf, "CPU architecture:", 17) == 0) + { + int version; + version = atoi(buf+17); + + if(version >= 6) + flags |= OPUS_CPU_ARM_MEDIA; + } +# endif + } + + fclose(cpuinfo); + } + return flags; +} +#else +/* The feature registers which can tell us what the processor supports are + * accessible in priveleged modes only, so we can't have a general user-space + * detection method like on x86.*/ +# error "Configured to use ARM asm but no CPU detection method available for " \ + "your platform. Reconfigure with --disable-rtcd (or send patches)." +#endif + +int opus_select_arch(void) +{ + opus_uint32 flags = opus_cpu_capabilities(); + int arch = 0; + + if(!(flags & OPUS_CPU_ARM_EDSP)) + return arch; + arch++; + + if(!(flags & OPUS_CPU_ARM_MEDIA)) + return arch; + arch++; + + if(!(flags & OPUS_CPU_ARM_NEON)) + return arch; + arch++; + + return arch; +} + +#endif diff --git a/code/opus-1.1/celt/arm/armcpu.h b/code/opus-1.1/celt/arm/armcpu.h new file mode 100644 index 00000000..ac574460 --- /dev/null +++ b/code/opus-1.1/celt/arm/armcpu.h @@ -0,0 +1,71 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(ARMCPU_H) +# define ARMCPU_H + +# if defined(OPUS_ARM_MAY_HAVE_EDSP) +# define MAY_HAVE_EDSP(name) name ## _edsp +# else +# define MAY_HAVE_EDSP(name) name ## _c +# endif + +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) +# define MAY_HAVE_MEDIA(name) name ## _media +# else +# define MAY_HAVE_MEDIA(name) MAY_HAVE_EDSP(name) +# endif + +# if defined(OPUS_ARM_MAY_HAVE_NEON) +# define MAY_HAVE_NEON(name) name ## _neon +# else +# define MAY_HAVE_NEON(name) MAY_HAVE_MEDIA(name) +# endif + +# if defined(OPUS_ARM_PRESUME_EDSP) +# define PRESUME_EDSP(name) name ## _edsp +# else +# define PRESUME_EDSP(name) name ## _c +# endif + +# if defined(OPUS_ARM_PRESUME_MEDIA) +# define PRESUME_MEDIA(name) name ## _media +# else +# define PRESUME_MEDIA(name) PRESUME_EDSP(name) +# endif + +# if defined(OPUS_ARM_PRESUME_NEON) +# define PRESUME_NEON(name) name ## _neon +# else +# define PRESUME_NEON(name) PRESUME_MEDIA(name) +# endif + +# if defined(OPUS_HAVE_RTCD) +int opus_select_arch(void); +# endif + +#endif diff --git a/code/opus-1.0.2/src/opus.c b/code/opus-1.1/celt/arm/armopts.s.in similarity index 75% rename from code/opus-1.0.2/src/opus.c rename to code/opus-1.1/celt/arm/armopts.s.in index d6ae7bab..3d8aaf27 100644 --- a/code/opus-1.0.2/src/opus.c +++ b/code/opus-1.1/celt/arm/armopts.s.in @@ -1,5 +1,4 @@ -/* Copyright (c) 2011 Xiph.Org Foundation, Skype Limited - Written by Jean-Marc Valin and Koen Vos */ +/* Copyright (C) 2013 Mozilla Corporation */ /* Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions @@ -25,23 +24,14 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ -#ifdef HAVE_CONFIG_H -#include "config.h" -#endif +; Set the following to 1 if we have EDSP instructions +; (LDRD/STRD, etc., ARMv5E and later). +OPUS_ARM_MAY_HAVE_EDSP * @OPUS_ARM_MAY_HAVE_EDSP@ -#include "opus.h" -#include "opus_private.h" +; Set the following to 1 if we have ARMv6 media instructions. +OPUS_ARM_MAY_HAVE_MEDIA * @OPUS_ARM_MAY_HAVE_MEDIA@ -int encode_size(int size, unsigned char *data) -{ - if (size < 252) - { - data[0] = size; - return 1; - } else { - data[0] = 252+(size&0x3); - data[1] = (size-(int)data[0])>>2; - return 2; - } -} +; Set the following to 1 if we have NEON (some ARMv7) +OPUS_ARM_MAY_HAVE_NEON * @OPUS_ARM_MAY_HAVE_NEON@ +END diff --git a/code/opus-1.1/celt/arm/celt_pitch_xcorr_arm.s b/code/opus-1.1/celt/arm/celt_pitch_xcorr_arm.s new file mode 100644 index 00000000..09917b16 --- /dev/null +++ b/code/opus-1.1/celt/arm/celt_pitch_xcorr_arm.s @@ -0,0 +1,545 @@ +; Copyright (c) 2007-2008 CSIRO +; Copyright (c) 2007-2009 Xiph.Org Foundation +; Copyright (c) 2013 Parrot +; Written by Aurélien Zanelli +; +; Redistribution and use in source and binary forms, with or without +; modification, are permitted provided that the following conditions +; are met: +; +; - Redistributions of source code must retain the above copyright +; notice, this list of conditions and the following disclaimer. +; +; - Redistributions in binary form must reproduce the above copyright +; notice, this list of conditions and the following disclaimer in the +; documentation and/or other materials provided with the distribution. +; +; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +; ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +; OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +; EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +; PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +; PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +; LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +; NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +; SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + AREA |.text|, CODE, READONLY + + GET celt/arm/armopts.s + +IF OPUS_ARM_MAY_HAVE_EDSP + EXPORT celt_pitch_xcorr_edsp +ENDIF + +IF OPUS_ARM_MAY_HAVE_NEON + EXPORT celt_pitch_xcorr_neon +ENDIF + +IF OPUS_ARM_MAY_HAVE_NEON + +; Compute sum[k]=sum(x[j]*y[j+k],j=0...len-1), k=0...3 +xcorr_kernel_neon PROC + ; input: + ; r3 = int len + ; r4 = opus_val16 *x + ; r5 = opus_val16 *y + ; q0 = opus_val32 sum[4] + ; output: + ; q0 = opus_val32 sum[4] + ; preserved: r0-r3, r6-r11, d2, q4-q7, q9-q15 + ; internal usage: + ; r12 = int j + ; d3 = y_3|y_2|y_1|y_0 + ; q2 = y_B|y_A|y_9|y_8|y_7|y_6|y_5|y_4 + ; q3 = x_7|x_6|x_5|x_4|x_3|x_2|x_1|x_0 + ; q8 = scratch + ; + ; Load y[0...3] + ; This requires len>0 to always be valid (which we assert in the C code). + VLD1.16 {d5}, [r5]! + SUBS r12, r3, #8 + BLE xcorr_kernel_neon_process4 +; Process 8 samples at a time. +; This loop loads one y value more than we actually need. Therefore we have to +; stop as soon as there are 8 or fewer samples left (instead of 7), to avoid +; reading past the end of the array. +xcorr_kernel_neon_process8 + ; This loop has 19 total instructions (10 cycles to issue, minimum), with + ; - 2 cycles of ARM insrtuctions, + ; - 10 cycles of load/store/byte permute instructions, and + ; - 9 cycles of data processing instructions. + ; On a Cortex A8, we dual-issue the maximum amount (9 cycles) between the + ; latter two categories, meaning the whole loop should run in 10 cycles per + ; iteration, barring cache misses. + ; + ; Load x[0...7] + VLD1.16 {d6, d7}, [r4]! + ; Unlike VMOV, VAND is a data processsing instruction (and doesn't get + ; assembled to VMOV, like VORR would), so it dual-issues with the prior VLD1. + VAND d3, d5, d5 + SUBS r12, r12, #8 + ; Load y[4...11] + VLD1.16 {d4, d5}, [r5]! + VMLAL.S16 q0, d3, d6[0] + VEXT.16 d16, d3, d4, #1 + VMLAL.S16 q0, d4, d7[0] + VEXT.16 d17, d4, d5, #1 + VMLAL.S16 q0, d16, d6[1] + VEXT.16 d16, d3, d4, #2 + VMLAL.S16 q0, d17, d7[1] + VEXT.16 d17, d4, d5, #2 + VMLAL.S16 q0, d16, d6[2] + VEXT.16 d16, d3, d4, #3 + VMLAL.S16 q0, d17, d7[2] + VEXT.16 d17, d4, d5, #3 + VMLAL.S16 q0, d16, d6[3] + VMLAL.S16 q0, d17, d7[3] + BGT xcorr_kernel_neon_process8 +; Process 4 samples here if we have > 4 left (still reading one extra y value). +xcorr_kernel_neon_process4 + ADDS r12, r12, #4 + BLE xcorr_kernel_neon_process2 + ; Load x[0...3] + VLD1.16 d6, [r4]! + ; Use VAND since it's a data processing instruction again. + VAND d4, d5, d5 + SUB r12, r12, #4 + ; Load y[4...7] + VLD1.16 d5, [r5]! + VMLAL.S16 q0, d4, d6[0] + VEXT.16 d16, d4, d5, #1 + VMLAL.S16 q0, d16, d6[1] + VEXT.16 d16, d4, d5, #2 + VMLAL.S16 q0, d16, d6[2] + VEXT.16 d16, d4, d5, #3 + VMLAL.S16 q0, d16, d6[3] +; Process 2 samples here if we have > 2 left (still reading one extra y value). +xcorr_kernel_neon_process2 + ADDS r12, r12, #2 + BLE xcorr_kernel_neon_process1 + ; Load x[0...1] + VLD2.16 {d6[],d7[]}, [r4]! + ; Use VAND since it's a data processing instruction again. + VAND d4, d5, d5 + SUB r12, r12, #2 + ; Load y[4...5] + VLD1.32 {d5[]}, [r5]! + VMLAL.S16 q0, d4, d6 + VEXT.16 d16, d4, d5, #1 + ; Replace bottom copy of {y5,y4} in d5 with {y3,y2} from d4, using VSRI + ; instead of VEXT, since it's a data-processing instruction. + VSRI.64 d5, d4, #32 + VMLAL.S16 q0, d16, d7 +; Process 1 sample using the extra y value we loaded above. +xcorr_kernel_neon_process1 + ; Load next *x + VLD1.16 {d6[]}, [r4]! + ADDS r12, r12, #1 + ; y[0...3] are left in d5 from prior iteration(s) (if any) + VMLAL.S16 q0, d5, d6 + MOVLE pc, lr +; Now process 1 last sample, not reading ahead. + ; Load last *y + VLD1.16 {d4[]}, [r5]! + VSRI.64 d4, d5, #16 + ; Load last *x + VLD1.16 {d6[]}, [r4]! + VMLAL.S16 q0, d4, d6 + MOV pc, lr + ENDP + +; opus_val32 celt_pitch_xcorr_neon(opus_val16 *_x, opus_val16 *_y, +; opus_val32 *xcorr, int len, int max_pitch) +celt_pitch_xcorr_neon PROC + ; input: + ; r0 = opus_val16 *_x + ; r1 = opus_val16 *_y + ; r2 = opus_val32 *xcorr + ; r3 = int len + ; output: + ; r0 = int maxcorr + ; internal usage: + ; r4 = opus_val16 *x (for xcorr_kernel_neon()) + ; r5 = opus_val16 *y (for xcorr_kernel_neon()) + ; r6 = int max_pitch + ; r12 = int j + ; q15 = int maxcorr[4] (q15 is not used by xcorr_kernel_neon()) + STMFD sp!, {r4-r6, lr} + LDR r6, [sp, #16] + VMOV.S32 q15, #1 + ; if (max_pitch < 4) goto celt_pitch_xcorr_neon_process4_done + SUBS r6, r6, #4 + BLT celt_pitch_xcorr_neon_process4_done +celt_pitch_xcorr_neon_process4 + ; xcorr_kernel_neon parameters: + ; r3 = len, r4 = _x, r5 = _y, q0 = {0, 0, 0, 0} + MOV r4, r0 + MOV r5, r1 + VEOR q0, q0, q0 + ; xcorr_kernel_neon only modifies r4, r5, r12, and q0...q3. + ; So we don't save/restore any other registers. + BL xcorr_kernel_neon + SUBS r6, r6, #4 + VST1.32 {q0}, [r2]! + ; _y += 4 + ADD r1, r1, #8 + VMAX.S32 q15, q15, q0 + ; if (max_pitch < 4) goto celt_pitch_xcorr_neon_process4_done + BGE celt_pitch_xcorr_neon_process4 +; We have less than 4 sums left to compute. +celt_pitch_xcorr_neon_process4_done + ADDS r6, r6, #4 + ; Reduce maxcorr to a single value + VMAX.S32 d30, d30, d31 + VPMAX.S32 d30, d30, d30 + ; if (max_pitch <= 0) goto celt_pitch_xcorr_neon_done + BLE celt_pitch_xcorr_neon_done +; Now compute each remaining sum one at a time. +celt_pitch_xcorr_neon_process_remaining + MOV r4, r0 + MOV r5, r1 + VMOV.I32 q0, #0 + SUBS r12, r3, #8 + BLT celt_pitch_xcorr_neon_process_remaining4 +; Sum terms 8 at a time. +celt_pitch_xcorr_neon_process_remaining_loop8 + ; Load x[0...7] + VLD1.16 {q1}, [r4]! + ; Load y[0...7] + VLD1.16 {q2}, [r5]! + SUBS r12, r12, #8 + VMLAL.S16 q0, d4, d2 + VMLAL.S16 q0, d5, d3 + BGE celt_pitch_xcorr_neon_process_remaining_loop8 +; Sum terms 4 at a time. +celt_pitch_xcorr_neon_process_remaining4 + ADDS r12, r12, #4 + BLT celt_pitch_xcorr_neon_process_remaining4_done + ; Load x[0...3] + VLD1.16 {d2}, [r4]! + ; Load y[0...3] + VLD1.16 {d3}, [r5]! + SUB r12, r12, #4 + VMLAL.S16 q0, d3, d2 +celt_pitch_xcorr_neon_process_remaining4_done + ; Reduce the sum to a single value. + VADD.S32 d0, d0, d1 + VPADDL.S32 d0, d0 + ADDS r12, r12, #4 + BLE celt_pitch_xcorr_neon_process_remaining_loop_done +; Sum terms 1 at a time. +celt_pitch_xcorr_neon_process_remaining_loop1 + VLD1.16 {d2[]}, [r4]! + VLD1.16 {d3[]}, [r5]! + SUBS r12, r12, #1 + VMLAL.S16 q0, d2, d3 + BGT celt_pitch_xcorr_neon_process_remaining_loop1 +celt_pitch_xcorr_neon_process_remaining_loop_done + VST1.32 {d0[0]}, [r2]! + VMAX.S32 d30, d30, d0 + SUBS r6, r6, #1 + ; _y++ + ADD r1, r1, #2 + ; if (--max_pitch > 0) goto celt_pitch_xcorr_neon_process_remaining + BGT celt_pitch_xcorr_neon_process_remaining +celt_pitch_xcorr_neon_done + VMOV.32 r0, d30[0] + LDMFD sp!, {r4-r6, pc} + ENDP + +ENDIF + +IF OPUS_ARM_MAY_HAVE_EDSP + +; This will get used on ARMv7 devices without NEON, so it has been optimized +; to take advantage of dual-issuing where possible. +xcorr_kernel_edsp PROC + ; input: + ; r3 = int len + ; r4 = opus_val16 *_x (must be 32-bit aligned) + ; r5 = opus_val16 *_y (must be 32-bit aligned) + ; r6...r9 = opus_val32 sum[4] + ; output: + ; r6...r9 = opus_val32 sum[4] + ; preserved: r0-r5 + ; internal usage + ; r2 = int j + ; r12,r14 = opus_val16 x[4] + ; r10,r11 = opus_val16 y[4] + STMFD sp!, {r2,r4,r5,lr} + LDR r10, [r5], #4 ; Load y[0...1] + SUBS r2, r3, #4 ; j = len-4 + LDR r11, [r5], #4 ; Load y[2...3] + BLE xcorr_kernel_edsp_process4_done + LDR r12, [r4], #4 ; Load x[0...1] + ; Stall +xcorr_kernel_edsp_process4 + ; The multiplies must issue from pipeline 0, and can't dual-issue with each + ; other. Every other instruction here dual-issues with a multiply, and is + ; thus "free". There should be no stalls in the body of the loop. + SMLABB r6, r12, r10, r6 ; sum[0] = MAC16_16(sum[0],x_0,y_0) + LDR r14, [r4], #4 ; Load x[2...3] + SMLABT r7, r12, r10, r7 ; sum[1] = MAC16_16(sum[1],x_0,y_1) + SUBS r2, r2, #4 ; j-=4 + SMLABB r8, r12, r11, r8 ; sum[2] = MAC16_16(sum[2],x_0,y_2) + SMLABT r9, r12, r11, r9 ; sum[3] = MAC16_16(sum[3],x_0,y_3) + SMLATT r6, r12, r10, r6 ; sum[0] = MAC16_16(sum[0],x_1,y_1) + LDR r10, [r5], #4 ; Load y[4...5] + SMLATB r7, r12, r11, r7 ; sum[1] = MAC16_16(sum[1],x_1,y_2) + SMLATT r8, r12, r11, r8 ; sum[2] = MAC16_16(sum[2],x_1,y_3) + SMLATB r9, r12, r10, r9 ; sum[3] = MAC16_16(sum[3],x_1,y_4) + LDRGT r12, [r4], #4 ; Load x[0...1] + SMLABB r6, r14, r11, r6 ; sum[0] = MAC16_16(sum[0],x_2,y_2) + SMLABT r7, r14, r11, r7 ; sum[1] = MAC16_16(sum[1],x_2,y_3) + SMLABB r8, r14, r10, r8 ; sum[2] = MAC16_16(sum[2],x_2,y_4) + SMLABT r9, r14, r10, r9 ; sum[3] = MAC16_16(sum[3],x_2,y_5) + SMLATT r6, r14, r11, r6 ; sum[0] = MAC16_16(sum[0],x_3,y_3) + LDR r11, [r5], #4 ; Load y[6...7] + SMLATB r7, r14, r10, r7 ; sum[1] = MAC16_16(sum[1],x_3,y_4) + SMLATT r8, r14, r10, r8 ; sum[2] = MAC16_16(sum[2],x_3,y_5) + SMLATB r9, r14, r11, r9 ; sum[3] = MAC16_16(sum[3],x_3,y_6) + BGT xcorr_kernel_edsp_process4 +xcorr_kernel_edsp_process4_done + ADDS r2, r2, #4 + BLE xcorr_kernel_edsp_done + LDRH r12, [r4], #2 ; r12 = *x++ + SUBS r2, r2, #1 ; j-- + ; Stall + SMLABB r6, r12, r10, r6 ; sum[0] = MAC16_16(sum[0],x,y_0) + LDRGTH r14, [r4], #2 ; r14 = *x++ + SMLABT r7, r12, r10, r7 ; sum[1] = MAC16_16(sum[1],x,y_1) + SMLABB r8, r12, r11, r8 ; sum[2] = MAC16_16(sum[2],x,y_2) + SMLABT r9, r12, r11, r9 ; sum[3] = MAC16_16(sum[3],x,y_3) + BLE xcorr_kernel_edsp_done + SMLABT r6, r14, r10, r6 ; sum[0] = MAC16_16(sum[0],x,y_1) + SUBS r2, r2, #1 ; j-- + SMLABB r7, r14, r11, r7 ; sum[1] = MAC16_16(sum[1],x,y_2) + LDRH r10, [r5], #2 ; r10 = y_4 = *y++ + SMLABT r8, r14, r11, r8 ; sum[2] = MAC16_16(sum[2],x,y_3) + LDRGTH r12, [r4], #2 ; r12 = *x++ + SMLABB r9, r14, r10, r9 ; sum[3] = MAC16_16(sum[3],x,y_4) + BLE xcorr_kernel_edsp_done + SMLABB r6, r12, r11, r6 ; sum[0] = MAC16_16(sum[0],tmp,y_2) + CMP r2, #1 ; j-- + SMLABT r7, r12, r11, r7 ; sum[1] = MAC16_16(sum[1],tmp,y_3) + LDRH r2, [r5], #2 ; r2 = y_5 = *y++ + SMLABB r8, r12, r10, r8 ; sum[2] = MAC16_16(sum[2],tmp,y_4) + LDRGTH r14, [r4] ; r14 = *x + SMLABB r9, r12, r2, r9 ; sum[3] = MAC16_16(sum[3],tmp,y_5) + BLE xcorr_kernel_edsp_done + SMLABT r6, r14, r11, r6 ; sum[0] = MAC16_16(sum[0],tmp,y_3) + LDRH r11, [r5] ; r11 = y_6 = *y + SMLABB r7, r14, r10, r7 ; sum[1] = MAC16_16(sum[1],tmp,y_4) + SMLABB r8, r14, r2, r8 ; sum[2] = MAC16_16(sum[2],tmp,y_5) + SMLABB r9, r14, r11, r9 ; sum[3] = MAC16_16(sum[3],tmp,y_6) +xcorr_kernel_edsp_done + LDMFD sp!, {r2,r4,r5,pc} + ENDP + +celt_pitch_xcorr_edsp PROC + ; input: + ; r0 = opus_val16 *_x (must be 32-bit aligned) + ; r1 = opus_val16 *_y (only needs to be 16-bit aligned) + ; r2 = opus_val32 *xcorr + ; r3 = int len + ; output: + ; r0 = maxcorr + ; internal usage + ; r4 = opus_val16 *x + ; r5 = opus_val16 *y + ; r6 = opus_val32 sum0 + ; r7 = opus_val32 sum1 + ; r8 = opus_val32 sum2 + ; r9 = opus_val32 sum3 + ; r1 = int max_pitch + ; r12 = int j + STMFD sp!, {r4-r11, lr} + MOV r5, r1 + LDR r1, [sp, #36] + MOV r4, r0 + TST r5, #3 + ; maxcorr = 1 + MOV r0, #1 + BEQ celt_pitch_xcorr_edsp_process1u_done +; Compute one sum at the start to make y 32-bit aligned. + SUBS r12, r3, #4 + ; r14 = sum = 0 + MOV r14, #0 + LDRH r8, [r5], #2 + BLE celt_pitch_xcorr_edsp_process1u_loop4_done + LDR r6, [r4], #4 + MOV r8, r8, LSL #16 +celt_pitch_xcorr_edsp_process1u_loop4 + LDR r9, [r5], #4 + SMLABT r14, r6, r8, r14 ; sum = MAC16_16(sum, x_0, y_0) + LDR r7, [r4], #4 + SMLATB r14, r6, r9, r14 ; sum = MAC16_16(sum, x_1, y_1) + LDR r8, [r5], #4 + SMLABT r14, r7, r9, r14 ; sum = MAC16_16(sum, x_2, y_2) + SUBS r12, r12, #4 ; j-=4 + SMLATB r14, r7, r8, r14 ; sum = MAC16_16(sum, x_3, y_3) + LDRGT r6, [r4], #4 + BGT celt_pitch_xcorr_edsp_process1u_loop4 + MOV r8, r8, LSR #16 +celt_pitch_xcorr_edsp_process1u_loop4_done + ADDS r12, r12, #4 +celt_pitch_xcorr_edsp_process1u_loop1 + LDRGEH r6, [r4], #2 + ; Stall + SMLABBGE r14, r6, r8, r14 ; sum = MAC16_16(sum, *x, *y) + SUBGES r12, r12, #1 + LDRGTH r8, [r5], #2 + BGT celt_pitch_xcorr_edsp_process1u_loop1 + ; Restore _x + SUB r4, r4, r3, LSL #1 + ; Restore and advance _y + SUB r5, r5, r3, LSL #1 + ; maxcorr = max(maxcorr, sum) + CMP r0, r14 + ADD r5, r5, #2 + MOVLT r0, r14 + SUBS r1, r1, #1 + ; xcorr[i] = sum + STR r14, [r2], #4 + BLE celt_pitch_xcorr_edsp_done +celt_pitch_xcorr_edsp_process1u_done + ; if (max_pitch < 4) goto celt_pitch_xcorr_edsp_process2 + SUBS r1, r1, #4 + BLT celt_pitch_xcorr_edsp_process2 +celt_pitch_xcorr_edsp_process4 + ; xcorr_kernel_edsp parameters: + ; r3 = len, r4 = _x, r5 = _y, r6...r9 = sum[4] = {0, 0, 0, 0} + MOV r6, #0 + MOV r7, #0 + MOV r8, #0 + MOV r9, #0 + BL xcorr_kernel_edsp ; xcorr_kernel_edsp(_x, _y+i, xcorr+i, len) + ; maxcorr = max(maxcorr, sum0, sum1, sum2, sum3) + CMP r0, r6 + ; _y+=4 + ADD r5, r5, #8 + MOVLT r0, r6 + CMP r0, r7 + MOVLT r0, r7 + CMP r0, r8 + MOVLT r0, r8 + CMP r0, r9 + MOVLT r0, r9 + STMIA r2!, {r6-r9} + SUBS r1, r1, #4 + BGE celt_pitch_xcorr_edsp_process4 +celt_pitch_xcorr_edsp_process2 + ADDS r1, r1, #2 + BLT celt_pitch_xcorr_edsp_process1a + SUBS r12, r3, #4 + ; {r10, r11} = {sum0, sum1} = {0, 0} + MOV r10, #0 + MOV r11, #0 + LDR r8, [r5], #4 + BLE celt_pitch_xcorr_edsp_process2_loop_done + LDR r6, [r4], #4 + LDR r9, [r5], #4 +celt_pitch_xcorr_edsp_process2_loop4 + SMLABB r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_0) + LDR r7, [r4], #4 + SMLABT r11, r6, r8, r11 ; sum1 = MAC16_16(sum1, x_0, y_1) + SUBS r12, r12, #4 ; j-=4 + SMLATT r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_1, y_1) + LDR r8, [r5], #4 + SMLATB r11, r6, r9, r11 ; sum1 = MAC16_16(sum1, x_1, y_2) + LDRGT r6, [r4], #4 + SMLABB r10, r7, r9, r10 ; sum0 = MAC16_16(sum0, x_2, y_2) + SMLABT r11, r7, r9, r11 ; sum1 = MAC16_16(sum1, x_2, y_3) + SMLATT r10, r7, r9, r10 ; sum0 = MAC16_16(sum0, x_3, y_3) + LDRGT r9, [r5], #4 + SMLATB r11, r7, r8, r11 ; sum1 = MAC16_16(sum1, x_3, y_4) + BGT celt_pitch_xcorr_edsp_process2_loop4 +celt_pitch_xcorr_edsp_process2_loop_done + ADDS r12, r12, #2 + BLE celt_pitch_xcorr_edsp_process2_1 + LDR r6, [r4], #4 + ; Stall + SMLABB r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_0) + LDR r9, [r5], #4 + SMLABT r11, r6, r8, r11 ; sum1 = MAC16_16(sum1, x_0, y_1) + SUB r12, r12, #2 + SMLATT r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_1, y_1) + MOV r8, r9 + SMLATB r11, r6, r9, r11 ; sum1 = MAC16_16(sum1, x_1, y_2) +celt_pitch_xcorr_edsp_process2_1 + LDRH r6, [r4], #2 + ADDS r12, r12, #1 + ; Stall + SMLABB r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_0) + LDRGTH r7, [r4], #2 + SMLABT r11, r6, r8, r11 ; sum1 = MAC16_16(sum1, x_0, y_1) + BLE celt_pitch_xcorr_edsp_process2_done + LDRH r9, [r5], #2 + SMLABT r10, r7, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_1) + SMLABB r11, r7, r9, r11 ; sum1 = MAC16_16(sum1, x_0, y_2) +celt_pitch_xcorr_edsp_process2_done + ; Restore _x + SUB r4, r4, r3, LSL #1 + ; Restore and advance _y + SUB r5, r5, r3, LSL #1 + ; maxcorr = max(maxcorr, sum0) + CMP r0, r10 + ADD r5, r5, #2 + MOVLT r0, r10 + SUB r1, r1, #2 + ; maxcorr = max(maxcorr, sum1) + CMP r0, r11 + ; xcorr[i] = sum + STR r10, [r2], #4 + MOVLT r0, r11 + STR r11, [r2], #4 +celt_pitch_xcorr_edsp_process1a + ADDS r1, r1, #1 + BLT celt_pitch_xcorr_edsp_done + SUBS r12, r3, #4 + ; r14 = sum = 0 + MOV r14, #0 + BLT celt_pitch_xcorr_edsp_process1a_loop_done + LDR r6, [r4], #4 + LDR r8, [r5], #4 + LDR r7, [r4], #4 + LDR r9, [r5], #4 +celt_pitch_xcorr_edsp_process1a_loop4 + SMLABB r14, r6, r8, r14 ; sum = MAC16_16(sum, x_0, y_0) + SUBS r12, r12, #4 ; j-=4 + SMLATT r14, r6, r8, r14 ; sum = MAC16_16(sum, x_1, y_1) + LDRGE r6, [r4], #4 + SMLABB r14, r7, r9, r14 ; sum = MAC16_16(sum, x_2, y_2) + LDRGE r8, [r5], #4 + SMLATT r14, r7, r9, r14 ; sum = MAC16_16(sum, x_3, y_3) + LDRGE r7, [r4], #4 + LDRGE r9, [r5], #4 + BGE celt_pitch_xcorr_edsp_process1a_loop4 +celt_pitch_xcorr_edsp_process1a_loop_done + ADDS r12, r12, #2 + LDRGE r6, [r4], #4 + LDRGE r8, [r5], #4 + ; Stall + SMLABBGE r14, r6, r8, r14 ; sum = MAC16_16(sum, x_0, y_0) + SUBGE r12, r12, #2 + SMLATTGE r14, r6, r8, r14 ; sum = MAC16_16(sum, x_1, y_1) + ADDS r12, r12, #1 + LDRGEH r6, [r4], #2 + LDRGEH r8, [r5], #2 + ; Stall + SMLABBGE r14, r6, r8, r14 ; sum = MAC16_16(sum, *x, *y) + ; maxcorr = max(maxcorr, sum) + CMP r0, r14 + ; xcorr[i] = sum + STR r14, [r2], #4 + MOVLT r0, r14 +celt_pitch_xcorr_edsp_done + LDMFD sp!, {r4-r11, pc} + ENDP + +ENDIF + +END diff --git a/code/opus-1.1/celt/arm/fixed_armv4.h b/code/opus-1.1/celt/arm/fixed_armv4.h new file mode 100644 index 00000000..b690bc8c --- /dev/null +++ b/code/opus-1.1/celt/arm/fixed_armv4.h @@ -0,0 +1,76 @@ +/* Copyright (C) 2013 Xiph.Org Foundation and contributors */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_ARMv4_H +#define FIXED_ARMv4_H + +/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q16 +static OPUS_INLINE opus_val32 MULT16_32_Q16_armv4(opus_val16 a, opus_val32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#MULT16_32_Q16\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(b),"r"(a<<16) + ); + return rd_hi; +} +#define MULT16_32_Q16(a, b) (MULT16_32_Q16_armv4(a, b)) + + +/** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q15 +static OPUS_INLINE opus_val32 MULT16_32_Q15_armv4(opus_val16 a, opus_val32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#MULT16_32_Q15\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(b), "r"(a<<16) + ); + /*We intentionally don't OR in the high bit of rd_lo for speed.*/ + return rd_hi<<1; +} +#define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv4(a, b)) + + +/** 16x32 multiply, followed by a 15-bit shift right and 32-bit add. + b must fit in 31 bits. + Result fits in 32 bits. */ +#undef MAC16_32_Q15 +#define MAC16_32_Q15(c, a, b) ADD32(c, MULT16_32_Q15(a, b)) + + +/** 32x32 multiplication, followed by a 31-bit shift right. Results fits in 32 bits */ +#undef MULT32_32_Q31 +#define MULT32_32_Q31(a,b) (opus_val32)((((opus_int64)(a)) * ((opus_int64)(b)))>>31) + +#endif diff --git a/code/opus-1.1/celt/arm/fixed_armv5e.h b/code/opus-1.1/celt/arm/fixed_armv5e.h new file mode 100644 index 00000000..1194a7d3 --- /dev/null +++ b/code/opus-1.1/celt/arm/fixed_armv5e.h @@ -0,0 +1,116 @@ +/* Copyright (C) 2007-2009 Xiph.Org Foundation + Copyright (C) 2003-2008 Jean-Marc Valin + Copyright (C) 2007-2008 CSIRO + Copyright (C) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_ARMv5E_H +#define FIXED_ARMv5E_H + +#include "fixed_armv4.h" + +/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q16 +static OPUS_INLINE opus_val32 MULT16_32_Q16_armv5e(opus_val16 a, opus_val32 b) +{ + int res; + __asm__( + "#MULT16_32_Q16\n\t" + "smulwb %0, %1, %2\n\t" + : "=r"(res) + : "r"(b),"r"(a) + ); + return res; +} +#define MULT16_32_Q16(a, b) (MULT16_32_Q16_armv5e(a, b)) + + +/** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q15 +static OPUS_INLINE opus_val32 MULT16_32_Q15_armv5e(opus_val16 a, opus_val32 b) +{ + int res; + __asm__( + "#MULT16_32_Q15\n\t" + "smulwb %0, %1, %2\n\t" + : "=r"(res) + : "r"(b), "r"(a) + ); + return res<<1; +} +#define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv5e(a, b)) + + +/** 16x32 multiply, followed by a 15-bit shift right and 32-bit add. + b must fit in 31 bits. + Result fits in 32 bits. */ +#undef MAC16_32_Q15 +static OPUS_INLINE opus_val32 MAC16_32_Q15_armv5e(opus_val32 c, opus_val16 a, + opus_val32 b) +{ + int res; + __asm__( + "#MAC16_32_Q15\n\t" + "smlawb %0, %1, %2, %3;\n" + : "=r"(res) + : "r"(b<<1), "r"(a), "r"(c) + ); + return res; +} +#define MAC16_32_Q15(c, a, b) (MAC16_32_Q15_armv5e(c, a, b)) + +/** 16x16 multiply-add where the result fits in 32 bits */ +#undef MAC16_16 +static OPUS_INLINE opus_val32 MAC16_16_armv5e(opus_val32 c, opus_val16 a, + opus_val16 b) +{ + int res; + __asm__( + "#MAC16_16\n\t" + "smlabb %0, %1, %2, %3;\n" + : "=r"(res) + : "r"(a), "r"(b), "r"(c) + ); + return res; +} +#define MAC16_16(c, a, b) (MAC16_16_armv5e(c, a, b)) + +/** 16x16 multiplication where the result fits in 32 bits */ +#undef MULT16_16 +static OPUS_INLINE opus_val32 MULT16_16_armv5e(opus_val16 a, opus_val16 b) +{ + int res; + __asm__( + "#MULT16_16\n\t" + "smulbb %0, %1, %2;\n" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define MULT16_16(a, b) (MULT16_16_armv5e(a, b)) + +#endif diff --git a/code/opus-1.1/celt/arm/kiss_fft_armv4.h b/code/opus-1.1/celt/arm/kiss_fft_armv4.h new file mode 100644 index 00000000..e4faad6f --- /dev/null +++ b/code/opus-1.1/celt/arm/kiss_fft_armv4.h @@ -0,0 +1,121 @@ +/*Copyright (c) 2013, Xiph.Org Foundation and contributors. + + All rights reserved. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE.*/ + +#ifndef KISS_FFT_ARMv4_H +#define KISS_FFT_ARMv4_H + +#if !defined(KISS_FFT_GUTS_H) +#error "This file should only be included from _kiss_fft_guts.h" +#endif + +#ifdef FIXED_POINT + +#undef C_MUL +#define C_MUL(m,a,b) \ + do{ \ + int br__; \ + int bi__; \ + int tt__; \ + __asm__ __volatile__( \ + "#C_MUL\n\t" \ + "ldrsh %[br], [%[bp], #0]\n\t" \ + "ldm %[ap], {r0,r1}\n\t" \ + "ldrsh %[bi], [%[bp], #2]\n\t" \ + "smull %[tt], %[mi], r1, %[br]\n\t" \ + "smlal %[tt], %[mi], r0, %[bi]\n\t" \ + "rsb %[bi], %[bi], #0\n\t" \ + "smull %[br], %[mr], r0, %[br]\n\t" \ + "mov %[tt], %[tt], lsr #15\n\t" \ + "smlal %[br], %[mr], r1, %[bi]\n\t" \ + "orr %[mi], %[tt], %[mi], lsl #17\n\t" \ + "mov %[br], %[br], lsr #15\n\t" \ + "orr %[mr], %[br], %[mr], lsl #17\n\t" \ + : [mr]"=r"((m).r), [mi]"=r"((m).i), \ + [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \ + : [ap]"r"(&(a)), [bp]"r"(&(b)) \ + : "r0", "r1" \ + ); \ + } \ + while(0) + +#undef C_MUL4 +#define C_MUL4(m,a,b) \ + do{ \ + int br__; \ + int bi__; \ + int tt__; \ + __asm__ __volatile__( \ + "#C_MUL4\n\t" \ + "ldrsh %[br], [%[bp], #0]\n\t" \ + "ldm %[ap], {r0,r1}\n\t" \ + "ldrsh %[bi], [%[bp], #2]\n\t" \ + "smull %[tt], %[mi], r1, %[br]\n\t" \ + "smlal %[tt], %[mi], r0, %[bi]\n\t" \ + "rsb %[bi], %[bi], #0\n\t" \ + "smull %[br], %[mr], r0, %[br]\n\t" \ + "mov %[tt], %[tt], lsr #17\n\t" \ + "smlal %[br], %[mr], r1, %[bi]\n\t" \ + "orr %[mi], %[tt], %[mi], lsl #15\n\t" \ + "mov %[br], %[br], lsr #17\n\t" \ + "orr %[mr], %[br], %[mr], lsl #15\n\t" \ + : [mr]"=r"((m).r), [mi]"=r"((m).i), \ + [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \ + : [ap]"r"(&(a)), [bp]"r"(&(b)) \ + : "r0", "r1" \ + ); \ + } \ + while(0) + +#undef C_MULC +#define C_MULC(m,a,b) \ + do{ \ + int br__; \ + int bi__; \ + int tt__; \ + __asm__ __volatile__( \ + "#C_MULC\n\t" \ + "ldrsh %[br], [%[bp], #0]\n\t" \ + "ldm %[ap], {r0,r1}\n\t" \ + "ldrsh %[bi], [%[bp], #2]\n\t" \ + "smull %[tt], %[mr], r0, %[br]\n\t" \ + "smlal %[tt], %[mr], r1, %[bi]\n\t" \ + "rsb %[bi], %[bi], #0\n\t" \ + "smull %[br], %[mi], r1, %[br]\n\t" \ + "mov %[tt], %[tt], lsr #15\n\t" \ + "smlal %[br], %[mi], r0, %[bi]\n\t" \ + "orr %[mr], %[tt], %[mr], lsl #17\n\t" \ + "mov %[br], %[br], lsr #15\n\t" \ + "orr %[mi], %[br], %[mi], lsl #17\n\t" \ + : [mr]"=r"((m).r), [mi]"=r"((m).i), \ + [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \ + : [ap]"r"(&(a)), [bp]"r"(&(b)) \ + : "r0", "r1" \ + ); \ + } \ + while(0) + +#endif /* FIXED_POINT */ + +#endif /* KISS_FFT_ARMv4_H */ diff --git a/code/opus-1.1/celt/arm/kiss_fft_armv5e.h b/code/opus-1.1/celt/arm/kiss_fft_armv5e.h new file mode 100644 index 00000000..9eca183d --- /dev/null +++ b/code/opus-1.1/celt/arm/kiss_fft_armv5e.h @@ -0,0 +1,118 @@ +/*Copyright (c) 2013, Xiph.Org Foundation and contributors. + + All rights reserved. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE.*/ + +#ifndef KISS_FFT_ARMv5E_H +#define KISS_FFT_ARMv5E_H + +#if !defined(KISS_FFT_GUTS_H) +#error "This file should only be included from _kiss_fft_guts.h" +#endif + +#ifdef FIXED_POINT + +#if defined(__thumb__)||defined(__thumb2__) +#define LDRD_CONS "Q" +#else +#define LDRD_CONS "Uq" +#endif + +#undef C_MUL +#define C_MUL(m,a,b) \ + do{ \ + int mr1__; \ + int mr2__; \ + int mi__; \ + long long aval__; \ + int bval__; \ + __asm__( \ + "#C_MUL\n\t" \ + "ldrd %[aval], %H[aval], %[ap]\n\t" \ + "ldr %[bval], %[bp]\n\t" \ + "smulwb %[mi], %H[aval], %[bval]\n\t" \ + "smulwb %[mr1], %[aval], %[bval]\n\t" \ + "smulwt %[mr2], %H[aval], %[bval]\n\t" \ + "smlawt %[mi], %[aval], %[bval], %[mi]\n\t" \ + : [mr1]"=r"(mr1__), [mr2]"=r"(mr2__), [mi]"=r"(mi__), \ + [aval]"=&r"(aval__), [bval]"=r"(bval__) \ + : [ap]LDRD_CONS(a), [bp]"m"(b) \ + ); \ + (m).r = SHL32(SUB32(mr1__, mr2__), 1); \ + (m).i = SHL32(mi__, 1); \ + } \ + while(0) + +#undef C_MUL4 +#define C_MUL4(m,a,b) \ + do{ \ + int mr1__; \ + int mr2__; \ + int mi__; \ + long long aval__; \ + int bval__; \ + __asm__( \ + "#C_MUL4\n\t" \ + "ldrd %[aval], %H[aval], %[ap]\n\t" \ + "ldr %[bval], %[bp]\n\t" \ + "smulwb %[mi], %H[aval], %[bval]\n\t" \ + "smulwb %[mr1], %[aval], %[bval]\n\t" \ + "smulwt %[mr2], %H[aval], %[bval]\n\t" \ + "smlawt %[mi], %[aval], %[bval], %[mi]\n\t" \ + : [mr1]"=r"(mr1__), [mr2]"=r"(mr2__), [mi]"=r"(mi__), \ + [aval]"=&r"(aval__), [bval]"=r"(bval__) \ + : [ap]LDRD_CONS(a), [bp]"m"(b) \ + ); \ + (m).r = SHR32(SUB32(mr1__, mr2__), 1); \ + (m).i = SHR32(mi__, 1); \ + } \ + while(0) + +#undef C_MULC +#define C_MULC(m,a,b) \ + do{ \ + int mr__; \ + int mi1__; \ + int mi2__; \ + long long aval__; \ + int bval__; \ + __asm__( \ + "#C_MULC\n\t" \ + "ldrd %[aval], %H[aval], %[ap]\n\t" \ + "ldr %[bval], %[bp]\n\t" \ + "smulwb %[mr], %[aval], %[bval]\n\t" \ + "smulwb %[mi1], %H[aval], %[bval]\n\t" \ + "smulwt %[mi2], %[aval], %[bval]\n\t" \ + "smlawt %[mr], %H[aval], %[bval], %[mr]\n\t" \ + : [mr]"=r"(mr__), [mi1]"=r"(mi1__), [mi2]"=r"(mi2__), \ + [aval]"=&r"(aval__), [bval]"=r"(bval__) \ + : [ap]LDRD_CONS(a), [bp]"m"(b) \ + ); \ + (m).r = SHL32(mr__, 1); \ + (m).i = SHL32(SUB32(mi1__, mi2__), 1); \ + } \ + while(0) + +#endif /* FIXED_POINT */ + +#endif /* KISS_FFT_GUTS_H */ diff --git a/code/opus-1.1/celt/arm/pitch_arm.h b/code/opus-1.1/celt/arm/pitch_arm.h new file mode 100644 index 00000000..a07f8ac2 --- /dev/null +++ b/code/opus-1.1/celt/arm/pitch_arm.h @@ -0,0 +1,57 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(PITCH_ARM_H) +# define PITCH_ARM_H + +# include "armcpu.h" + +# if defined(FIXED_POINT) + +# if defined(OPUS_ARM_MAY_HAVE_NEON) +opus_val32 celt_pitch_xcorr_neon(const opus_val16 *_x, const opus_val16 *_y, + opus_val32 *xcorr, int len, int max_pitch); +# endif + +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) +# define celt_pitch_xcorr_media MAY_HAVE_EDSP(celt_pitch_xcorr) +# endif + +# if defined(OPUS_ARM_MAY_HAVE_EDSP) +opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y, + opus_val32 *xcorr, int len, int max_pitch); +# endif + +# if !defined(OPUS_HAVE_RTCD) +# define OVERRIDE_PITCH_XCORR (1) +# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \ + ((void)(arch),PRESUME_NEON(celt_pitch_xcorr)(_x, _y, xcorr, len, max_pitch)) +# endif + +# endif + +#endif diff --git a/code/opus-1.0.2/celt/bands.c b/code/opus-1.1/celt/bands.c similarity index 54% rename from code/opus-1.0.2/celt/bands.c rename to code/opus-1.1/celt/bands.c index 3be543c3..cce56e2f 100644 --- a/code/opus-1.0.2/celt/bands.c +++ b/code/opus-1.1/celt/bands.c @@ -40,6 +40,23 @@ #include "os_support.h" #include "mathops.h" #include "rate.h" +#include "quant_bands.h" +#include "pitch.h" + +int hysteresis_decision(opus_val16 val, const opus_val16 *thresholds, const opus_val16 *hysteresis, int N, int prev) +{ + int i; + for (i=0;iprev && val < thresholds[prev]+hysteresis[prev]) + i=prev; + if (i thresholds[prev-1]-hysteresis[prev-1]) + i=prev; + return i; +} opus_uint32 celt_lcg_rand(opus_uint32 seed) { @@ -172,7 +189,8 @@ void normalise_bands(const CELTMode *m, const celt_sig * OPUS_RESTRICT freq, cel #endif /* FIXED_POINT */ /* De-normalise the energy to produce the synthesis from the unit-energy bands */ -void denormalise_bands(const CELTMode *m, const celt_norm * OPUS_RESTRICT X, celt_sig * OPUS_RESTRICT freq, const celt_ener *bandE, int end, int C, int M) +void denormalise_bands(const CELTMode *m, const celt_norm * OPUS_RESTRICT X, + celt_sig * OPUS_RESTRICT freq, const opus_val16 *bandLogE, int start, int end, int C, int M) { int i, c, N; const opus_int16 *eBands = m->eBands; @@ -182,18 +200,55 @@ void denormalise_bands(const CELTMode *m, const celt_norm * OPUS_RESTRICT X, cel celt_sig * OPUS_RESTRICT f; const celt_norm * OPUS_RESTRICT x; f = freq+c*N; - x = X+c*N; - for (i=0;inbEBands],1); + opus_val16 g; + opus_val16 lg; +#ifdef FIXED_POINT + int shift; +#endif j=M*eBands[i]; band_end = M*eBands[i+1]; + lg = ADD16(bandLogE[i+c*m->nbEBands], SHL16((opus_val16)eMeans[i],6)); +#ifndef FIXED_POINT + g = celt_exp2(lg); +#else + /* Handle the integer part of the log energy */ + shift = 16-(lg>>DB_SHIFT); + if (shift>31) + { + shift=0; + g=0; + } else { + /* Handle the fractional part. */ + g = celt_exp2_frac(lg&((1<0); /*M*(eBands[end]-eBands[end-1]) <= 8 assures this*/ + celt_assert(nbBands>0); /* end has to be non-zero */ sum /= nbBands; /* Recursive averaging */ sum = (sum+*average)>>1; @@ -483,50 +534,6 @@ int spreading_decision(const CELTMode *m, celt_norm *X, int *average, return decision; } -#ifdef MEASURE_NORM_MSE - -float MSE[30] = {0}; -int nbMSEBands = 0; -int MSECount[30] = {0}; - -void dump_norm_mse(void) -{ - int i; - for (i=0;inbEBands;i++) - { - int j; - int c; - float g; - if (bandE0[i]<10 || (C==2 && bandE0[i+m->nbEBands]<1)) - continue; - c=0; do { - g = bandE[i+c*m->nbEBands]/(1e-15+bandE0[i+c*m->nbEBands]); - for (j=M*m->eBands[i];jeBands[i+1];j++) - MSE[i] += (g*X[j+c*N]-X0[j+c*N])*(g*X[j+c*N]-X0[j+c*N]); - } while (++cnbEBands; -} - -#endif - /* Indexing table for converting from natural Hadamard to ordery Hadamard This is essentially a bit-reversed Gray, on top of which we've added an inversion of the order because we want the DC at the end rather than @@ -629,289 +636,301 @@ static int compute_qn(int N, int b, int offset, int pulse_cap, int stereo) return qn; } -/* This function is responsible for encoding and decoding a band for both - the mono and stereo case. Even in the mono case, it can split the band - in two and transmit the energy difference with the two half-bands. It - can be called recursively so bands can end up being split in 8 parts. */ -static unsigned quant_band(int encode, const CELTMode *m, int i, celt_norm *X, celt_norm *Y, - int N, int b, int spread, int B, int intensity, int tf_change, celt_norm *lowband, ec_ctx *ec, - opus_int32 *remaining_bits, int LM, celt_norm *lowband_out, const celt_ener *bandE, int level, - opus_uint32 *seed, opus_val16 gain, celt_norm *lowband_scratch, int fill) +struct band_ctx { + int encode; + const CELTMode *m; + int i; + int intensity; + int spread; + int tf_change; + ec_ctx *ec; + opus_int32 remaining_bits; + const celt_ener *bandE; + opus_uint32 seed; +}; + +struct split_ctx { + int inv; + int imid; + int iside; + int delta; + int itheta; + int qalloc; +}; + +static void compute_theta(struct band_ctx *ctx, struct split_ctx *sctx, + celt_norm *X, celt_norm *Y, int N, int *b, int B, int B0, + int LM, + int stereo, int *fill) +{ + int qn; + int itheta=0; + int delta; + int imid, iside; + int qalloc; + int pulse_cap; + int offset; + opus_int32 tell; + int inv=0; + int encode; + const CELTMode *m; + int i; + int intensity; + ec_ctx *ec; + const celt_ener *bandE; + + encode = ctx->encode; + m = ctx->m; + i = ctx->i; + intensity = ctx->intensity; + ec = ctx->ec; + bandE = ctx->bandE; + + /* Decide on the resolution to give to the split parameter theta */ + pulse_cap = m->logN[i]+LM*(1<>1) - (stereo&&N==2 ? QTHETA_OFFSET_TWOPHASE : QTHETA_OFFSET); + qn = compute_qn(N, *b, offset, pulse_cap, stereo); + if (stereo && i>=intensity) + qn = 1; + if (encode) + { + /* theta is the atan() of the ratio between the (normalized) + side and mid. With just that parameter, we can re-scale both + mid and side because we know that 1) they have unit norm and + 2) they are orthogonal. */ + itheta = stereo_itheta(X, Y, stereo, N); + } + tell = ec_tell_frac(ec); + if (qn!=1) + { + if (encode) + itheta = (itheta*qn+8192)>>14; + + /* Entropy coding of the angle. We use a uniform pdf for the + time split, a step for stereo, and a triangular one for the rest. */ + if (stereo && N>2) + { + int p0 = 3; + int x = itheta; + int x0 = qn/2; + int ft = p0*(x0+1) + x0; + /* Use a probability of p0 up to itheta=8192 and then use 1 after */ + if (encode) + { + ec_encode(ec,x<=x0?p0*x:(x-1-x0)+(x0+1)*p0,x<=x0?p0*(x+1):(x-x0)+(x0+1)*p0,ft); + } else { + int fs; + fs=ec_decode(ec,ft); + if (fs<(x0+1)*p0) + x=fs/p0; + else + x=x0+1+(fs-(x0+1)*p0); + ec_dec_update(ec,x<=x0?p0*x:(x-1-x0)+(x0+1)*p0,x<=x0?p0*(x+1):(x-x0)+(x0+1)*p0,ft); + itheta = x; + } + } else if (B0>1 || stereo) { + /* Uniform pdf */ + if (encode) + ec_enc_uint(ec, itheta, qn+1); + else + itheta = ec_dec_uint(ec, qn+1); + } else { + int fs=1, ft; + ft = ((qn>>1)+1)*((qn>>1)+1); + if (encode) + { + int fl; + + fs = itheta <= (qn>>1) ? itheta + 1 : qn + 1 - itheta; + fl = itheta <= (qn>>1) ? itheta*(itheta + 1)>>1 : + ft - ((qn + 1 - itheta)*(qn + 2 - itheta)>>1); + + ec_encode(ec, fl, fl+fs, ft); + } else { + /* Triangular pdf */ + int fl=0; + int fm; + fm = ec_decode(ec, ft); + + if (fm < ((qn>>1)*((qn>>1) + 1)>>1)) + { + itheta = (isqrt32(8*(opus_uint32)fm + 1) - 1)>>1; + fs = itheta + 1; + fl = itheta*(itheta + 1)>>1; + } + else + { + itheta = (2*(qn + 1) + - isqrt32(8*(opus_uint32)(ft - fm - 1) + 1))>>1; + fs = qn + 1 - itheta; + fl = ft - ((qn + 1 - itheta)*(qn + 2 - itheta)>>1); + } + + ec_dec_update(ec, fl, fl+fs, ft); + } + } + itheta = (opus_int32)itheta*16384/qn; + if (encode && stereo) + { + if (itheta==0) + intensity_stereo(m, X, Y, bandE, i, N); + else + stereo_split(X, Y, N); + } + /* NOTE: Renormalising X and Y *may* help fixed-point a bit at very high rate. + Let's do that at higher complexity */ + } else if (stereo) { + if (encode) + { + inv = itheta > 8192; + if (inv) + { + int j; + for (j=0;j2<remaining_bits > 2<inv = inv; + sctx->imid = imid; + sctx->iside = iside; + sctx->delta = delta; + sctx->itheta = itheta; + sctx->qalloc = qalloc; +} +static unsigned quant_band_n1(struct band_ctx *ctx, celt_norm *X, celt_norm *Y, int b, + celt_norm *lowband_out) +{ +#ifdef RESYNTH + int resynth = 1; +#else + int resynth = !ctx->encode; +#endif + int c; + int stereo; + celt_norm *x = X; + int encode; + ec_ctx *ec; + + encode = ctx->encode; + ec = ctx->ec; + + stereo = Y != NULL; + c=0; do { + int sign=0; + if (ctx->remaining_bits>=1<remaining_bits -= 1<encode; #endif + celt_norm *Y=NULL; + int encode; + const CELTMode *m; + int i; + int spread; + ec_ctx *ec; - longBlocks = B0==1; - - N_B /= B; - N_B0 = N_B; - - split = stereo = Y != NULL; - - /* Special case for one sample */ - if (N==1) - { - int c; - celt_norm *x = X; - c=0; do { - int sign=0; - if (*remaining_bits>=1<0) - recombine = tf_change; - /* Band recombining to increase frequency resolution */ - - if (lowband && (recombine || ((N_B&1) == 0 && tf_change<0) || B0>1)) - { - int j; - for (j=0;j>k, 1<>k, 1<>4]<<2; - } - B>>=recombine; - N_B<<=recombine; - - /* Increasing the time resolution */ - while ((N_B&1) == 0 && tf_change<0) - { - if (encode) - haar1(X, N_B, B); - if (lowband) - haar1(lowband, N_B, B); - fill |= fill<>= 1; - time_divide++; - tf_change++; - } - B0=B; - N_B0 = N_B; - - /* Reorganize the samples in time order instead of frequency order */ - if (B0>1) - { - if (encode) - deinterleave_hadamard(X, N_B>>recombine, B0<>recombine, B0<encode; + m = ctx->m; + i = ctx->i; + spread = ctx->spread; + ec = ctx->ec; /* If we need 1.5 more bit than we can produce, split the band in two. */ cache = m->cache.bits + m->cache.index[(LM+1)*m->nbEBands+i]; - if (!stereo && LM != -1 && b > cache[cache[0]]+12 && N>2) + if (LM != -1 && b > cache[cache[0]]+12 && N>2) { + int mbits, sbits, delta; + int itheta; + int qalloc; + struct split_ctx sctx; + celt_norm *next_lowband2=NULL; + opus_int32 rebalance; + N >>= 1; Y = X+N; - split = 1; LM -= 1; if (B==1) fill = (fill&1)|(fill<<1); B = (B+1)>>1; - } - - if (split) - { - int qn; - int itheta=0; - int mbits, sbits, delta; - int qalloc; - int pulse_cap; - int offset; - int orig_fill; - opus_int32 tell; - - /* Decide on the resolution to give to the split parameter theta */ - pulse_cap = m->logN[i]+LM*(1<>1) - (stereo&&N==2 ? QTHETA_OFFSET_TWOPHASE : QTHETA_OFFSET); - qn = compute_qn(N, b, offset, pulse_cap, stereo); - if (stereo && i>=intensity) - qn = 1; - if (encode) - { - /* theta is the atan() of the ratio between the (normalized) - side and mid. With just that parameter, we can re-scale both - mid and side because we know that 1) they have unit norm and - 2) they are orthogonal. */ - itheta = stereo_itheta(X, Y, stereo, N); - } - tell = ec_tell_frac(ec); - if (qn!=1) - { - if (encode) - itheta = (itheta*qn+8192)>>14; - - /* Entropy coding of the angle. We use a uniform pdf for the - time split, a step for stereo, and a triangular one for the rest. */ - if (stereo && N>2) - { - int p0 = 3; - int x = itheta; - int x0 = qn/2; - int ft = p0*(x0+1) + x0; - /* Use a probability of p0 up to itheta=8192 and then use 1 after */ - if (encode) - { - ec_encode(ec,x<=x0?p0*x:(x-1-x0)+(x0+1)*p0,x<=x0?p0*(x+1):(x-x0)+(x0+1)*p0,ft); - } else { - int fs; - fs=ec_decode(ec,ft); - if (fs<(x0+1)*p0) - x=fs/p0; - else - x=x0+1+(fs-(x0+1)*p0); - ec_dec_update(ec,x<=x0?p0*x:(x-1-x0)+(x0+1)*p0,x<=x0?p0*(x+1):(x-x0)+(x0+1)*p0,ft); - itheta = x; - } - } else if (B0>1 || stereo) { - /* Uniform pdf */ - if (encode) - ec_enc_uint(ec, itheta, qn+1); - else - itheta = ec_dec_uint(ec, qn+1); - } else { - int fs=1, ft; - ft = ((qn>>1)+1)*((qn>>1)+1); - if (encode) - { - int fl; - - fs = itheta <= (qn>>1) ? itheta + 1 : qn + 1 - itheta; - fl = itheta <= (qn>>1) ? itheta*(itheta + 1)>>1 : - ft - ((qn + 1 - itheta)*(qn + 2 - itheta)>>1); - - ec_encode(ec, fl, fl+fs, ft); - } else { - /* Triangular pdf */ - int fl=0; - int fm; - fm = ec_decode(ec, ft); - - if (fm < ((qn>>1)*((qn>>1) + 1)>>1)) - { - itheta = (isqrt32(8*(opus_uint32)fm + 1) - 1)>>1; - fs = itheta + 1; - fl = itheta*(itheta + 1)>>1; - } - else - { - itheta = (2*(qn + 1) - - isqrt32(8*(opus_uint32)(ft - fm - 1) + 1))>>1; - fs = qn + 1 - itheta; - fl = ft - ((qn + 1 - itheta)*(qn + 2 - itheta)>>1); - } - - ec_dec_update(ec, fl, fl+fs, ft); - } - } - itheta = (opus_int32)itheta*16384/qn; - if (encode && stereo) - { - if (itheta==0) - intensity_stereo(m, X, Y, bandE, i, N); - else - stereo_split(X, Y, N); - } - /* NOTE: Renormalising X and Y *may* help fixed-point a bit at very high rate. - Let's do that at higher complexity */ - } else if (stereo) { - if (encode) - { - inv = itheta > 8192; - if (inv) - { - int j; - for (j=0;j2< 2<1 && (itheta&0x3fff)) { - int c; - int sign=0; - celt_norm *x2, *y2; - mbits = b; - sbits = 0; - /* Only need one bit for the side */ - if (itheta != 0 && itheta != 16384) - sbits = 1< 8192; - *remaining_bits -= qalloc+sbits; - - x2 = c ? Y : X; - y2 = c ? X : Y; - if (sbits) - { - if (encode) - { - /* Here we only need to encode a sign for the side */ - sign = x2[0]*y2[1] - x2[1]*y2[0] < 0; - ec_enc_bits(ec, sign, 1); - } else { - sign = ec_dec_bits(ec, 1); - } - } - sign = 1-2*sign; - /* We use orig_fill here because we want to fold the side, but if - itheta==16384, we'll have cleared the low bits of fill. */ - cm = quant_band(encode, m, i, x2, NULL, N, mbits, spread, B, intensity, tf_change, lowband, ec, remaining_bits, LM, lowband_out, NULL, level, seed, gain, lowband_scratch, orig_fill); - /* We don't split N=2 bands, so cm is either 1 or 0 (for a fold-collapse), - and there's no need to worry about mixing with the other channel. */ - y2[0] = -sign*x2[1]; - y2[1] = sign*x2[0]; - if (resynth) - { - celt_norm tmp; - X[0] = MULT16_16_Q15(mid, X[0]); - X[1] = MULT16_16_Q15(mid, X[1]); - Y[0] = MULT16_16_Q15(side, Y[0]); - Y[1] = MULT16_16_Q15(side, Y[1]); - tmp = X[0]; - X[0] = SUB16(tmp,Y[0]); - Y[0] = ADD16(tmp,Y[0]); - tmp = X[1]; - X[1] = SUB16(tmp,Y[1]); - Y[1] = ADD16(tmp,Y[1]); - } - } else { - /* "Normal" split code */ - celt_norm *next_lowband2=NULL; - celt_norm *next_lowband_out1=NULL; - int next_level=0; - opus_int32 rebalance; - - /* Give more bits to low-energy MDCTs than they would otherwise deserve */ - if (B0>1 && !stereo && (itheta&0x3fff)) - { - if (itheta > 8192) - /* Rough approximation for pre-echo masking */ - delta -= delta>>(4-LM); - else - /* Corresponds to a forward-masking slope of 1.5 dB per 10 ms */ - delta = IMIN(0, delta + (N<>(5-LM))); - } - mbits = IMAX(0, IMIN(b, (b-delta)/2)); - sbits = b-mbits; - *remaining_bits -= qalloc; - - if (lowband && !stereo) - next_lowband2 = lowband+N; /* >32-bit split case */ - - /* Only stereo needs to pass on lowband_out. Otherwise, it's - handled at the end */ - if (stereo) - next_lowband_out1 = lowband_out; + if (itheta > 8192) + /* Rough approximation for pre-echo masking */ + delta -= delta>>(4-LM); else - next_level = level+1; - - rebalance = *remaining_bits; - if (mbits >= sbits) - { - /* In stereo mode, we do not apply a scaling to the mid because we need the normalized - mid for folding later */ - cm = quant_band(encode, m, i, X, NULL, N, mbits, spread, B, intensity, tf_change, - lowband, ec, remaining_bits, LM, next_lowband_out1, - NULL, next_level, seed, stereo ? Q15ONE : MULT16_16_P15(gain,mid), lowband_scratch, fill); - rebalance = mbits - (rebalance-*remaining_bits); - if (rebalance > 3<>B)<<((B0>>1)&(stereo-1)); - } else { - /* For a stereo split, the high bits of fill are always zero, so no - folding will be done to the side. */ - cm = quant_band(encode, m, i, Y, NULL, N, sbits, spread, B, intensity, tf_change, - next_lowband2, ec, remaining_bits, LM, NULL, - NULL, next_level, seed, MULT16_16_P15(gain,side), NULL, fill>>B)<<((B0>>1)&(stereo-1)); - rebalance = sbits - (rebalance-*remaining_bits); - if (rebalance > 3<>(5-LM))); } + mbits = IMAX(0, IMIN(b, (b-delta)/2)); + sbits = b-mbits; + ctx->remaining_bits -= qalloc; + if (lowband) + next_lowband2 = lowband+N; /* >32-bit split case */ + + rebalance = ctx->remaining_bits; + if (mbits >= sbits) + { + cm = quant_partition(ctx, X, N, mbits, B, + lowband, LM, + MULT16_16_P15(gain,mid), fill); + rebalance = mbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<>B)<<(B0>>1); + } else { + cm = quant_partition(ctx, Y, N, sbits, B, + next_lowband2, LM, + MULT16_16_P15(gain,side), fill>>B)<<(B0>>1); + rebalance = sbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<remaining_bits -= curr_bits; /* Ensures we can never bust the budget */ - while (*remaining_bits < 0 && q > 0) + while (ctx->remaining_bits < 0 && q > 0) { - *remaining_bits += curr_bits; + ctx->remaining_bits += curr_bits; q--; curr_bits = pulses2bits(m, i, LM, q); - *remaining_bits -= curr_bits; + ctx->remaining_bits -= curr_bits; } if (q!=0) @@ -1073,7 +1015,7 @@ static unsigned quant_band(int encode, const CELTMode *m, int i, celt_norm *X, c if (resynth) { unsigned cm_mask; - /*B can be as large as 16, so this shift might overflow an int on a + /* B can be as large as 16, so this shift might overflow an int on a 16-bit platform; use a long to get defined behavior.*/ cm_mask = (unsigned)(1UL<>20); + ctx->seed = celt_lcg_rand(ctx->seed); + X[j] = (celt_norm)((opus_int32)ctx->seed>>20); } cm = cm_mask; } else { @@ -1096,10 +1038,10 @@ static unsigned quant_band(int encode, const CELTMode *m, int i, celt_norm *X, c for (j=0;jseed = celt_lcg_rand(ctx->seed); /* About 48 dB below the "normal" folding level */ tmp = QCONST16(1.0f/256, 10); - tmp = (*seed)&0x8000 ? tmp : -tmp; + tmp = (ctx->seed)&0x8000 ? tmp : -tmp; X[j] = lowband[j]+tmp; } cm = fill; @@ -1110,64 +1052,306 @@ static unsigned quant_band(int encode, const CELTMode *m, int i, celt_norm *X, c } } + return cm; +} + + +/* This function is responsible for encoding and decoding a band for the mono case. */ +static unsigned quant_band(struct band_ctx *ctx, celt_norm *X, + int N, int b, int B, celt_norm *lowband, + int LM, celt_norm *lowband_out, + opus_val16 gain, celt_norm *lowband_scratch, int fill) +{ + int N0=N; + int N_B=N; + int N_B0; + int B0=B; + int time_divide=0; + int recombine=0; + int longBlocks; + unsigned cm=0; +#ifdef RESYNTH + int resynth = 1; +#else + int resynth = !ctx->encode; +#endif + int k; + int encode; + int tf_change; + + encode = ctx->encode; + tf_change = ctx->tf_change; + + longBlocks = B0==1; + + N_B /= B; + + /* Special case for one sample */ + if (N==1) + { + return quant_band_n1(ctx, X, NULL, b, lowband_out); + } + + if (tf_change>0) + recombine = tf_change; + /* Band recombining to increase frequency resolution */ + + if (lowband_scratch && lowband && (recombine || ((N_B&1) == 0 && tf_change<0) || B0>1)) + { + int j; + for (j=0;j>k, 1<>k, 1<>4]<<2; + } + B>>=recombine; + N_B<<=recombine; + + /* Increasing the time resolution */ + while ((N_B&1) == 0 && tf_change<0) + { + if (encode) + haar1(X, N_B, B); + if (lowband) + haar1(lowband, N_B, B); + fill |= fill<>= 1; + time_divide++; + tf_change++; + } + B0=B; + N_B0 = N_B; + + /* Reorganize the samples in time order instead of frequency order */ + if (B0>1) + { + if (encode) + deinterleave_hadamard(X, N_B>>recombine, B0<>recombine, B0<1) + interleave_hadamard(X, N_B>>recombine, B0<>= 1; + N_B <<= 1; + cm |= cm>>B; + haar1(X, N_B, B); + } + + for (k=0;k>k, 1<1) - interleave_hadamard(X, N_B>>recombine, B0<encode; +#endif + int mbits, sbits, delta; + int itheta; + int qalloc; + struct split_ctx sctx; + int orig_fill; + int encode; + ec_ctx *ec; + + encode = ctx->encode; + ec = ctx->ec; + + /* Special case for one sample */ + if (N==1) + { + return quant_band_n1(ctx, X, Y, b, lowband_out); + } + + orig_fill = fill; + + compute_theta(ctx, &sctx, X, Y, N, &b, B, B, + LM, 1, &fill); + inv = sctx.inv; + imid = sctx.imid; + iside = sctx.iside; + delta = sctx.delta; + itheta = sctx.itheta; + qalloc = sctx.qalloc; +#ifdef FIXED_POINT + mid = imid; + side = iside; +#else + mid = (1.f/32768)*imid; + side = (1.f/32768)*iside; +#endif + + /* This is a special case for N=2 that only works for stereo and takes + advantage of the fact that mid and side are orthogonal to encode + the side with just one bit. */ + if (N==2) + { + int c; + int sign=0; + celt_norm *x2, *y2; + mbits = b; + sbits = 0; + /* Only need one bit for the side. */ + if (itheta != 0 && itheta != 16384) + sbits = 1< 8192; + ctx->remaining_bits -= qalloc+sbits; + + x2 = c ? Y : X; + y2 = c ? X : Y; + if (sbits) + { + if (encode) { - B >>= 1; - N_B <<= 1; - cm |= cm>>B; - haar1(X, N_B, B); + /* Here we only need to encode a sign for the side. */ + sign = x2[0]*y2[1] - x2[1]*y2[0] < 0; + ec_enc_bits(ec, sign, 1); + } else { + sign = ec_dec_bits(ec, 1); } + } + sign = 1-2*sign; + /* We use orig_fill here because we want to fold the side, but if + itheta==16384, we'll have cleared the low bits of fill. */ + cm = quant_band(ctx, x2, N, mbits, B, lowband, + LM, lowband_out, Q15ONE, lowband_scratch, orig_fill); + /* We don't split N=2 bands, so cm is either 1 or 0 (for a fold-collapse), + and there's no need to worry about mixing with the other channel. */ + y2[0] = -sign*x2[1]; + y2[1] = sign*x2[0]; + if (resynth) + { + celt_norm tmp; + X[0] = MULT16_16_Q15(mid, X[0]); + X[1] = MULT16_16_Q15(mid, X[1]); + Y[0] = MULT16_16_Q15(side, Y[0]); + Y[1] = MULT16_16_Q15(side, Y[1]); + tmp = X[0]; + X[0] = SUB16(tmp,Y[0]); + Y[0] = ADD16(tmp,Y[0]); + tmp = X[1]; + X[1] = SUB16(tmp,Y[1]); + Y[1] = ADD16(tmp,Y[1]); + } + } else { + /* "Normal" split code */ + opus_int32 rebalance; - for (k=0;k>k, 1<remaining_bits -= qalloc; - /* Scale output for later folding */ - if (lowband_out) - { - int j; - opus_val16 n; - n = celt_sqrt(SHL32(EXTEND32(N0),22)); - for (j=0;jremaining_bits; + if (mbits >= sbits) + { + /* In stereo mode, we do not apply a scaling to the mid because we need the normalized + mid for folding later. */ + cm = quant_band(ctx, X, N, mbits, B, + lowband, LM, lowband_out, + Q15ONE, lowband_scratch, fill); + rebalance = mbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<>B); + } else { + /* For a stereo split, the high bits of fill are always zero, so no + folding will be done to the side. */ + cm = quant_band(ctx, Y, N, sbits, B, + NULL, LM, NULL, + side, NULL, fill>>B); + rebalance = sbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<eBands; celt_norm * OPUS_RESTRICT norm, * OPUS_RESTRICT norm2; VARDECL(celt_norm, _norm); - VARDECL(celt_norm, lowband_scratch); + celt_norm *lowband_scratch; int B; int M; int lowband_offset; int update_lowband = 1; int C = Y_ != NULL ? 2 : 1; + int norm_offset; #ifdef RESYNTH int resynth = 1; #else int resynth = !encode; #endif + struct band_ctx ctx; SAVE_STACK; M = 1<nbEBands], celt_norm); - ALLOC(lowband_scratch, M*(eBands[m->nbEBands]-eBands[m->nbEBands-1]), celt_norm); + norm_offset = M*eBands[start]; + /* No need to allocate norm for the last band because we don't need an + output in that band. */ + ALLOC(_norm, C*(M*eBands[m->nbEBands-1]-norm_offset), celt_norm); norm = _norm; - norm2 = norm + M*eBands[m->nbEBands]; + norm2 = norm + M*eBands[m->nbEBands-1]-norm_offset; + /* We can use the last band as scratch space because we don't need that + scratch space for the last band. */ + lowband_scratch = X_+M*eBands[m->nbEBands-1]; lowband_offset = 0; + ctx.bandE = bandE; + ctx.ec = ec; + ctx.encode = encode; + ctx.intensity = intensity; + ctx.m = m; + ctx.seed = *seed; + ctx.spread = spread; for (i=start;i=m->effEBands) { X=norm; if (Y_!=NULL) Y = norm; + lowband_scratch = NULL; } + if (i==end-1) + lowband_scratch = NULL; /* Get a conservative estimate of the collapse_mask's for the bands we're - going to be folding from. */ + going to be folding from. */ if (lowband_offset != 0 && (spread!=SPREAD_AGGRESSIVE || B>1 || tf_change<0)) { int fold_start; int fold_end; int fold_i; /* This ensures we never repeat spectral content within one band */ - effective_lowband = IMAX(M*eBands[start], M*eBands[lowband_offset]-N); + effective_lowband = IMAX(0, M*eBands[lowband_offset]-norm_offset-N); fold_start = lowband_offset; - while(M*eBands[--fold_start] > effective_lowband); + while(M*eBands[--fold_start] > effective_lowband+norm_offset); fold_end = lowband_offset-1; - while(M*eBands[++fold_end] < effective_lowband+N); + while(M*eBands[++fold_end] < effective_lowband+norm_offset+N); x_cm = y_cm = 0; fold_i = fold_start; do { x_cm |= collapse_masks[fold_i*C+0]; @@ -1262,7 +1469,7 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end, } while (++fold_i(N< +#include "celt.h" +#include "pitch.h" +#include "bands.h" +#include "modes.h" +#include "entcode.h" +#include "quant_bands.h" +#include "rate.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "float_cast.h" +#include +#include "celt_lpc.h" +#include "vq.h" + +#ifndef PACKAGE_VERSION +#define PACKAGE_VERSION "unknown" +#endif + + +int resampling_factor(opus_int32 rate) +{ + int ret; + switch (rate) + { + case 48000: + ret = 1; + break; + case 24000: + ret = 2; + break; + case 16000: + ret = 3; + break; + case 12000: + ret = 4; + break; + case 8000: + ret = 6; + break; + default: +#ifndef CUSTOM_MODES + celt_assert(0); +#endif + ret = 0; + break; + } + return ret; +} + +#ifndef OVERRIDE_COMB_FILTER_CONST +static void comb_filter_const(opus_val32 *y, opus_val32 *x, int T, int N, + opus_val16 g10, opus_val16 g11, opus_val16 g12) +{ + opus_val32 x0, x1, x2, x3, x4; + int i; + x4 = x[-T-2]; + x3 = x[-T-1]; + x2 = x[-T]; + x1 = x[-T+1]; + for (i=0;inbEBands;i++) + { + int N; + N=(m->eBands[i+1]-m->eBands[i])<cache.caps[m->nbEBands*(2*LM+C-1)+i]+64)*C*N>>2; + } +} + + + +const char *opus_strerror(int error) +{ + static const char * const error_strings[8] = { + "success", + "invalid argument", + "buffer too small", + "internal error", + "corrupted stream", + "request not implemented", + "invalid state", + "memory allocation failed" + }; + if (error > 0 || error < -7) + return "unknown error"; + else + return error_strings[-error]; +} + +const char *opus_get_version_string(void) +{ + return "libopus " PACKAGE_VERSION +#ifdef FIXED_POINT + "-fixed" +#endif +#ifdef FUZZING + "-fuzzing" +#endif + ; +} diff --git a/code/opus-1.0.2/celt/celt.h b/code/opus-1.1/celt/celt.h similarity index 53% rename from code/opus-1.0.2/celt/celt.h rename to code/opus-1.1/celt/celt.h index 218cd883..5deea1f0 100644 --- a/code/opus-1.0.2/celt/celt.h +++ b/code/opus-1.1/celt/celt.h @@ -50,10 +50,26 @@ extern "C" { #define CELTDecoder OpusCustomDecoder #define CELTMode OpusCustomMode -#define _celt_check_mode_ptr_ptr(ptr) ((ptr) + ((ptr) - (const CELTMode**)(ptr))) +typedef struct { + int valid; + float tonality; + float tonality_slope; + float noisiness; + float activity; + float music_prob; + int bandwidth; +}AnalysisInfo; + +#define __celt_check_mode_ptr_ptr(ptr) ((ptr) + ((ptr) - (const CELTMode**)(ptr))) + +#define __celt_check_analysis_ptr(ptr) ((ptr) + ((ptr) - (const AnalysisInfo*)(ptr))) /* Encoder/decoder Requests */ +/* Expose this option again when variable framesize actually works */ +#define OPUS_FRAMESIZE_VARIABLE 5010 /**< Optimize the frame size dynamically */ + + #define CELT_SET_PREDICTION_REQUEST 10002 /** Controls the use of interframe prediction. 0=Independent frames @@ -81,12 +97,24 @@ extern "C" { #define CELT_GET_MODE_REQUEST 10015 /** Get the CELTMode used by an encoder or decoder */ -#define CELT_GET_MODE(x) CELT_GET_MODE_REQUEST, _celt_check_mode_ptr_ptr(x) +#define CELT_GET_MODE(x) CELT_GET_MODE_REQUEST, __celt_check_mode_ptr_ptr(x) #define CELT_SET_SIGNALLING_REQUEST 10016 #define CELT_SET_SIGNALLING(x) CELT_SET_SIGNALLING_REQUEST, __opus_check_int(x) +#define CELT_SET_TONALITY_REQUEST 10018 +#define CELT_SET_TONALITY(x) CELT_SET_TONALITY_REQUEST, __opus_check_int(x) +#define CELT_SET_TONALITY_SLOPE_REQUEST 10020 +#define CELT_SET_TONALITY_SLOPE(x) CELT_SET_TONALITY_SLOPE_REQUEST, __opus_check_int(x) +#define CELT_SET_ANALYSIS_REQUEST 10022 +#define CELT_SET_ANALYSIS(x) CELT_SET_ANALYSIS_REQUEST, __celt_check_analysis_ptr(x) + +#define OPUS_SET_LFE_REQUEST 10024 +#define OPUS_SET_LFE(x) OPUS_SET_LFE_REQUEST, __opus_check_int(x) + +#define OPUS_SET_ENERGY_MASK_REQUEST 10026 +#define OPUS_SET_ENERGY_MASK(x) OPUS_SET_ENERGY_MASK_REQUEST, __opus_check_val16_ptr(x) /* Encoder stuff */ @@ -94,7 +122,8 @@ int celt_encoder_get_size(int channels); int celt_encode_with_ec(OpusCustomEncoder * OPUS_RESTRICT st, const opus_val16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes, ec_enc *enc); -int celt_encoder_init(CELTEncoder *st, opus_int32 sampling_rate, int channels); +int celt_encoder_init(CELTEncoder *st, opus_int32 sampling_rate, int channels, + int arch); @@ -110,6 +139,78 @@ int celt_decode_with_ec(OpusCustomDecoder * OPUS_RESTRICT st, const unsigned cha #define celt_encoder_ctl opus_custom_encoder_ctl #define celt_decoder_ctl opus_custom_decoder_ctl + +#ifdef CUSTOM_MODES +#define OPUS_CUSTOM_NOSTATIC +#else +#define OPUS_CUSTOM_NOSTATIC static OPUS_INLINE +#endif + +static const unsigned char trim_icdf[11] = {126, 124, 119, 109, 87, 41, 19, 9, 4, 2, 0}; +/* Probs: NONE: 21.875%, LIGHT: 6.25%, NORMAL: 65.625%, AGGRESSIVE: 6.25% */ +static const unsigned char spread_icdf[4] = {25, 23, 2, 0}; + +static const unsigned char tapset_icdf[3]={2,1,0}; + +#ifdef CUSTOM_MODES +static const unsigned char toOpusTable[20] = { + 0xE0, 0xE8, 0xF0, 0xF8, + 0xC0, 0xC8, 0xD0, 0xD8, + 0xA0, 0xA8, 0xB0, 0xB8, + 0x00, 0x00, 0x00, 0x00, + 0x80, 0x88, 0x90, 0x98, +}; + +static const unsigned char fromOpusTable[16] = { + 0x80, 0x88, 0x90, 0x98, + 0x40, 0x48, 0x50, 0x58, + 0x20, 0x28, 0x30, 0x38, + 0x00, 0x08, 0x10, 0x18 +}; + +static OPUS_INLINE int toOpus(unsigned char c) +{ + int ret=0; + if (c<0xA0) + ret = toOpusTable[c>>3]; + if (ret == 0) + return -1; + else + return ret|(c&0x7); +} + +static OPUS_INLINE int fromOpus(unsigned char c) +{ + if (c<0x80) + return -1; + else + return fromOpusTable[(c>>3)-16] | (c&0x7); +} +#endif /* CUSTOM_MODES */ + +#define COMBFILTER_MAXPERIOD 1024 +#define COMBFILTER_MINPERIOD 15 + +extern const signed char tf_select_table[4][8]; + +int resampling_factor(opus_int32 rate); + +void celt_preemphasis(const opus_val16 * OPUS_RESTRICT pcmp, celt_sig * OPUS_RESTRICT inp, + int N, int CC, int upsample, const opus_val16 *coef, celt_sig *mem, int clip); + +void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N, + opus_val16 g0, opus_val16 g1, int tapset0, int tapset1, + const opus_val16 *window, int overlap); + +void init_caps(const CELTMode *m,int *cap,int LM,int C); + +#ifdef RESYNTH +void deemphasis(celt_sig *in[], opus_val16 *pcm, int N, int C, int downsample, const opus_val16 *coef, celt_sig *mem, celt_sig * OPUS_RESTRICT scratch); + +void compute_inv_mdcts(const CELTMode *mode, int shortBlocks, celt_sig *X, + celt_sig * OPUS_RESTRICT out_mem[], int C, int LM); +#endif + #ifdef __cplusplus } #endif diff --git a/code/opus-1.1/celt/celt_decoder.c b/code/opus-1.1/celt/celt_decoder.c new file mode 100644 index 00000000..830398ee --- /dev/null +++ b/code/opus-1.1/celt/celt_decoder.c @@ -0,0 +1,1195 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2010 Xiph.Org Foundation + Copyright (c) 2008 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#define CELT_DECODER_C + +#include "cpu_support.h" +#include "os_support.h" +#include "mdct.h" +#include +#include "celt.h" +#include "pitch.h" +#include "bands.h" +#include "modes.h" +#include "entcode.h" +#include "quant_bands.h" +#include "rate.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "float_cast.h" +#include +#include "celt_lpc.h" +#include "vq.h" + +/**********************************************************************/ +/* */ +/* DECODER */ +/* */ +/**********************************************************************/ +#define DECODE_BUFFER_SIZE 2048 + +/** Decoder state + @brief Decoder state + */ +struct OpusCustomDecoder { + const OpusCustomMode *mode; + int overlap; + int channels; + int stream_channels; + + int downsample; + int start, end; + int signalling; + int arch; + + /* Everything beyond this point gets cleared on a reset */ +#define DECODER_RESET_START rng + + opus_uint32 rng; + int error; + int last_pitch_index; + int loss_count; + int postfilter_period; + int postfilter_period_old; + opus_val16 postfilter_gain; + opus_val16 postfilter_gain_old; + int postfilter_tapset; + int postfilter_tapset_old; + + celt_sig preemph_memD[2]; + + celt_sig _decode_mem[1]; /* Size = channels*(DECODE_BUFFER_SIZE+mode->overlap) */ + /* opus_val16 lpc[], Size = channels*LPC_ORDER */ + /* opus_val16 oldEBands[], Size = 2*mode->nbEBands */ + /* opus_val16 oldLogE[], Size = 2*mode->nbEBands */ + /* opus_val16 oldLogE2[], Size = 2*mode->nbEBands */ + /* opus_val16 backgroundLogE[], Size = 2*mode->nbEBands */ +}; + +int celt_decoder_get_size(int channels) +{ + const CELTMode *mode = opus_custom_mode_create(48000, 960, NULL); + return opus_custom_decoder_get_size(mode, channels); +} + +OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_get_size(const CELTMode *mode, int channels) +{ + int size = sizeof(struct CELTDecoder) + + (channels*(DECODE_BUFFER_SIZE+mode->overlap)-1)*sizeof(celt_sig) + + channels*LPC_ORDER*sizeof(opus_val16) + + 4*2*mode->nbEBands*sizeof(opus_val16); + return size; +} + +#ifdef CUSTOM_MODES +CELTDecoder *opus_custom_decoder_create(const CELTMode *mode, int channels, int *error) +{ + int ret; + CELTDecoder *st = (CELTDecoder *)opus_alloc(opus_custom_decoder_get_size(mode, channels)); + ret = opus_custom_decoder_init(st, mode, channels); + if (ret != OPUS_OK) + { + opus_custom_decoder_destroy(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} +#endif /* CUSTOM_MODES */ + +int celt_decoder_init(CELTDecoder *st, opus_int32 sampling_rate, int channels) +{ + int ret; + ret = opus_custom_decoder_init(st, opus_custom_mode_create(48000, 960, NULL), channels); + if (ret != OPUS_OK) + return ret; + st->downsample = resampling_factor(sampling_rate); + if (st->downsample==0) + return OPUS_BAD_ARG; + else + return OPUS_OK; +} + +OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_init(CELTDecoder *st, const CELTMode *mode, int channels) +{ + if (channels < 0 || channels > 2) + return OPUS_BAD_ARG; + + if (st==NULL) + return OPUS_ALLOC_FAIL; + + OPUS_CLEAR((char*)st, opus_custom_decoder_get_size(mode, channels)); + + st->mode = mode; + st->overlap = mode->overlap; + st->stream_channels = st->channels = channels; + + st->downsample = 1; + st->start = 0; + st->end = st->mode->effEBands; + st->signalling = 1; + st->arch = opus_select_arch(); + + st->loss_count = 0; + + opus_custom_decoder_ctl(st, OPUS_RESET_STATE); + + return OPUS_OK; +} + +#ifdef CUSTOM_MODES +void opus_custom_decoder_destroy(CELTDecoder *st) +{ + opus_free(st); +} +#endif /* CUSTOM_MODES */ + +static OPUS_INLINE opus_val16 SIG2WORD16(celt_sig x) +{ +#ifdef FIXED_POINT + x = PSHR32(x, SIG_SHIFT); + x = MAX32(x, -32768); + x = MIN32(x, 32767); + return EXTRACT16(x); +#else + return (opus_val16)x; +#endif +} + +#ifndef RESYNTH +static +#endif +void deemphasis(celt_sig *in[], opus_val16 *pcm, int N, int C, int downsample, const opus_val16 *coef, celt_sig *mem, celt_sig * OPUS_RESTRICT scratch) +{ + int c; + int Nd; + int apply_downsampling=0; + opus_val16 coef0; + + coef0 = coef[0]; + Nd = N/downsample; + c=0; do { + int j; + celt_sig * OPUS_RESTRICT x; + opus_val16 * OPUS_RESTRICT y; + celt_sig m = mem[c]; + x =in[c]; + y = pcm+c; +#ifdef CUSTOM_MODES + if (coef[1] != 0) + { + opus_val16 coef1 = coef[1]; + opus_val16 coef3 = coef[3]; + for (j=0;j1) + { + /* Shortcut for the standard (non-custom modes) case */ + for (j=0;jshortMdctSize; + shift = mode->maxLM; + } else { + B = 1; + N = mode->shortMdctSize<maxLM-LM; + } + c=0; do { + /* IMDCT on the interleaved the sub-frames, overlap-add is performed by the IMDCT */ + for (b=0;bmdct, &X[b+c*N*B], out_mem[c]+N*b, mode->window, overlap, shift, B); + } while (++cstorage*8; + tell = ec_tell(dec); + logp = isTransient ? 2 : 4; + tf_select_rsv = LM>0 && tell+logp+1<=budget; + budget -= tf_select_rsv; + tf_changed = curr = 0; + for (i=start;ichannels; + celt_sig *decode_mem[2]; + celt_sig *out_syn[2]; + opus_val16 *lpc; + opus_val16 *oldBandE, *oldLogE, *oldLogE2, *backgroundLogE; + const OpusCustomMode *mode; + int nbEBands; + int overlap; + int start; + int downsample; + int loss_count; + int noise_based; + const opus_int16 *eBands; + VARDECL(celt_sig, scratch); + SAVE_STACK; + + mode = st->mode; + nbEBands = mode->nbEBands; + overlap = mode->overlap; + eBands = mode->eBands; + + c=0; do { + decode_mem[c] = st->_decode_mem + c*(DECODE_BUFFER_SIZE+overlap); + out_syn[c] = decode_mem[c]+DECODE_BUFFER_SIZE-N; + } while (++c_decode_mem+(DECODE_BUFFER_SIZE+overlap)*C); + oldBandE = lpc+C*LPC_ORDER; + oldLogE = oldBandE + 2*nbEBands; + oldLogE2 = oldLogE + 2*nbEBands; + backgroundLogE = oldLogE2 + 2*nbEBands; + + loss_count = st->loss_count; + start = st->start; + downsample = st->downsample; + noise_based = loss_count >= 5 || start != 0; + ALLOC(scratch, noise_based?N*C:N, celt_sig); + if (noise_based) + { + /* Noise-based PLC/CNG */ + celt_sig *freq; + VARDECL(celt_norm, X); + opus_uint32 seed; + opus_val16 *plcLogE; + int end; + int effEnd; + + end = st->end; + effEnd = IMAX(start, IMIN(end, mode->effEBands)); + + /* Share the interleaved signal MDCT coefficient buffer with the + deemphasis scratch buffer. */ + freq = scratch; + ALLOC(X, C*N, celt_norm); /**< Interleaved normalised MDCTs */ + + if (loss_count >= 5) + plcLogE = backgroundLogE; + else { + /* Energy decay */ + opus_val16 decay = loss_count==0 ? + QCONST16(1.5f, DB_SHIFT) : QCONST16(.5f, DB_SHIFT); + c=0; do + { + for (i=start;irng; + for (c=0;c>20); + } + renormalise_vector(X+boffs, blen, Q15ONE); + } + } + st->rng = seed; + + denormalise_bands(mode, X, freq, plcLogE, start, effEnd, C, 1<>1)); + } while (++c>1, opus_val16 ); + pitch_downsample(decode_mem, lp_pitch_buf, + DECODE_BUFFER_SIZE, C, st->arch); + pitch_search(lp_pitch_buf+(PLC_PITCH_LAG_MAX>>1), lp_pitch_buf, + DECODE_BUFFER_SIZE-PLC_PITCH_LAG_MAX, + PLC_PITCH_LAG_MAX-PLC_PITCH_LAG_MIN, &pitch_index, st->arch); + pitch_index = PLC_PITCH_LAG_MAX-pitch_index; + st->last_pitch_index = pitch_index; + } else { + pitch_index = st->last_pitch_index; + fade = QCONST16(.8f,15); + } + + ALLOC(etmp, overlap, opus_val32); + ALLOC(exc, MAX_PERIOD, opus_val16); + window = mode->window; + c=0; do { + opus_val16 decay; + opus_val16 attenuation; + opus_val32 S1=0; + celt_sig *buf; + int extrapolation_offset; + int extrapolation_len; + int exc_length; + int j; + + buf = decode_mem[c]; + for (i=0;iarch); + /* Add a noise floor of -40 dB. */ +#ifdef FIXED_POINT + ac[0] += SHR32(ac[0],13); +#else + ac[0] *= 1.0001f; +#endif + /* Use lag windowing to stabilize the Levinson-Durbin recursion. */ + for (i=1;i<=LPC_ORDER;i++) + { + /*ac[i] *= exp(-.5*(2*M_PI*.002*i)*(2*M_PI*.002*i));*/ +#ifdef FIXED_POINT + ac[i] -= MULT16_32_Q15(2*i*i, ac[i]); +#else + ac[i] -= ac[i]*(0.008f*0.008f)*i*i; +#endif + } + _celt_lpc(lpc+c*LPC_ORDER, ac, LPC_ORDER); + } + /* We want the excitation for 2 pitch periods in order to look for a + decaying signal, but we can't get more than MAX_PERIOD. */ + exc_length = IMIN(2*pitch_index, MAX_PERIOD); + /* Initialize the LPC history with the samples just before the start + of the region for which we're computing the excitation. */ + { + opus_val16 lpc_mem[LPC_ORDER]; + for (i=0;i>1; + for (i=0;i= pitch_index) { + j -= pitch_index; + attenuation = MULT16_16_Q15(attenuation, decay); + } + buf[DECODE_BUFFER_SIZE-N+i] = + SHL32(EXTEND32(MULT16_16_Q15(attenuation, + exc[extrapolation_offset+j])), SIG_SHIFT); + /* Compute the energy of the previously decoded signal whose + excitation we're copying. */ + tmp = ROUND16( + buf[DECODE_BUFFER_SIZE-MAX_PERIOD-N+extrapolation_offset+j], + SIG_SHIFT); + S1 += SHR32(MULT16_16(tmp, tmp), 8); + } + + { + opus_val16 lpc_mem[LPC_ORDER]; + /* Copy the last decoded samples (prior to the overlap region) to + synthesis filter memory so we can have a continuous signal. */ + for (i=0;i SHR32(S2,2))) +#else + /* The float test is written this way to catch NaNs in the output + of the IIR filter at the same time. */ + if (!(S1 > 0.2f*S2)) +#endif + { + for (i=0;ipostfilter_period, st->postfilter_period, overlap, + -st->postfilter_gain, -st->postfilter_gain, + st->postfilter_tapset, st->postfilter_tapset, NULL, 0); + + /* Simulate TDAC on the concealed audio so that it blends with the + MDCT of the next frame. */ + for (i=0;ipreemph, st->preemph_memD, scratch); + + st->loss_count = loss_count+1; + + RESTORE_STACK; +} + +int celt_decode_with_ec(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, opus_val16 * OPUS_RESTRICT pcm, int frame_size, ec_dec *dec) +{ + int c, i, N; + int spread_decision; + opus_int32 bits; + ec_dec _dec; + VARDECL(celt_sig, freq); + VARDECL(celt_norm, X); + VARDECL(int, fine_quant); + VARDECL(int, pulses); + VARDECL(int, cap); + VARDECL(int, offsets); + VARDECL(int, fine_priority); + VARDECL(int, tf_res); + VARDECL(unsigned char, collapse_masks); + celt_sig *decode_mem[2]; + celt_sig *out_syn[2]; + opus_val16 *lpc; + opus_val16 *oldBandE, *oldLogE, *oldLogE2, *backgroundLogE; + + int shortBlocks; + int isTransient; + int intra_ener; + const int CC = st->channels; + int LM, M; + int effEnd; + int codedBands; + int alloc_trim; + int postfilter_pitch; + opus_val16 postfilter_gain; + int intensity=0; + int dual_stereo=0; + opus_int32 total_bits; + opus_int32 balance; + opus_int32 tell; + int dynalloc_logp; + int postfilter_tapset; + int anti_collapse_rsv; + int anti_collapse_on=0; + int silence; + int C = st->stream_channels; + const OpusCustomMode *mode; + int nbEBands; + int overlap; + const opus_int16 *eBands; + ALLOC_STACK; + + mode = st->mode; + nbEBands = mode->nbEBands; + overlap = mode->overlap; + eBands = mode->eBands; + frame_size *= st->downsample; + + c=0; do { + decode_mem[c] = st->_decode_mem + c*(DECODE_BUFFER_SIZE+overlap); + } while (++c_decode_mem+(DECODE_BUFFER_SIZE+overlap)*CC); + oldBandE = lpc+CC*LPC_ORDER; + oldLogE = oldBandE + 2*nbEBands; + oldLogE2 = oldLogE + 2*nbEBands; + backgroundLogE = oldLogE2 + 2*nbEBands; + +#ifdef CUSTOM_MODES + if (st->signalling && data!=NULL) + { + int data0=data[0]; + /* Convert "standard mode" to Opus header */ + if (mode->Fs==48000 && mode->shortMdctSize==120) + { + data0 = fromOpus(data0); + if (data0<0) + return OPUS_INVALID_PACKET; + } + st->end = IMAX(1, mode->effEBands-2*(data0>>5)); + LM = (data0>>3)&0x3; + C = 1 + ((data0>>2)&0x1); + data++; + len--; + if (LM>mode->maxLM) + return OPUS_INVALID_PACKET; + if (frame_size < mode->shortMdctSize<shortMdctSize<maxLM;LM++) + if (mode->shortMdctSize<mode->maxLM) + return OPUS_BAD_ARG; + } + M=1<1275 || pcm==NULL) + return OPUS_BAD_ARG; + + N = M*mode->shortMdctSize; + + effEnd = st->end; + if (effEnd > mode->effEBands) + effEnd = mode->effEBands; + + if (data == NULL || len<=1) + { + celt_decode_lost(st, pcm, N, LM); + RESTORE_STACK; + return frame_size/st->downsample; + } + + if (dec == NULL) + { + ec_dec_init(&_dec,(unsigned char*)data,len); + dec = &_dec; + } + + if (C==1) + { + for (i=0;i= total_bits) + silence = 1; + else if (tell==1) + silence = ec_dec_bit_logp(dec, 15); + else + silence = 0; + if (silence) + { + /* Pretend we've read all the remaining bits */ + tell = len*8; + dec->nbits_total+=tell-ec_tell(dec); + } + + postfilter_gain = 0; + postfilter_pitch = 0; + postfilter_tapset = 0; + if (st->start==0 && tell+16 <= total_bits) + { + if(ec_dec_bit_logp(dec, 1)) + { + int qg, octave; + octave = ec_dec_uint(dec, 6); + postfilter_pitch = (16< 0 && tell+3 <= total_bits) + { + isTransient = ec_dec_bit_logp(dec, 3); + tell = ec_tell(dec); + } + else + isTransient = 0; + + if (isTransient) + shortBlocks = M; + else + shortBlocks = 0; + + /* Decode the global flags (first symbols in the stream) */ + intra_ener = tell+3<=total_bits ? ec_dec_bit_logp(dec, 3) : 0; + /* Get band energies */ + unquant_coarse_energy(mode, st->start, st->end, oldBandE, + intra_ener, dec, C, LM); + + ALLOC(tf_res, nbEBands, int); + tf_decode(st->start, st->end, isTransient, tf_res, LM, dec); + + tell = ec_tell(dec); + spread_decision = SPREAD_NORMAL; + if (tell+4 <= total_bits) + spread_decision = ec_dec_icdf(dec, spread_icdf, 5); + + ALLOC(cap, nbEBands, int); + + init_caps(mode,cap,LM,C); + + ALLOC(offsets, nbEBands, int); + + dynalloc_logp = 6; + total_bits<<=BITRES; + tell = ec_tell_frac(dec); + for (i=st->start;iend;i++) + { + int width, quanta; + int dynalloc_loop_logp; + int boost; + width = C*(eBands[i+1]-eBands[i])<0) + dynalloc_logp = IMAX(2, dynalloc_logp-1); + } + + ALLOC(fine_quant, nbEBands, int); + alloc_trim = tell+(6<=2&&bits>=((LM+2)<start, st->end, offsets, cap, + alloc_trim, &intensity, &dual_stereo, bits, &balance, pulses, + fine_quant, fine_priority, C, LM, dec, 0, 0, 0); + + unquant_fine_energy(mode, st->start, st->end, oldBandE, fine_quant, dec, C); + + /* Decode fixed codebook */ + ALLOC(collapse_masks, C*nbEBands, unsigned char); + ALLOC(X, C*N, celt_norm); /**< Interleaved normalised MDCTs */ + + quant_all_bands(0, mode, st->start, st->end, X, C==2 ? X+N : NULL, collapse_masks, + NULL, pulses, shortBlocks, spread_decision, dual_stereo, intensity, tf_res, + len*(8<rng); + + if (anti_collapse_rsv > 0) + { + anti_collapse_on = ec_dec_bits(dec, 1); + } + + unquant_energy_finalise(mode, st->start, st->end, oldBandE, + fine_quant, fine_priority, len*8-ec_tell(dec), dec, C); + + if (anti_collapse_on) + anti_collapse(mode, X, collapse_masks, LM, C, N, + st->start, st->end, oldBandE, oldLogE, oldLogE2, pulses, st->rng); + + ALLOC(freq, IMAX(CC,C)*N, celt_sig); /**< Interleaved signal MDCTs */ + + if (silence) + { + for (i=0;istart, effEnd, C, M); + } + c=0; do { + OPUS_MOVE(decode_mem[c], decode_mem[c]+N, DECODE_BUFFER_SIZE-N+overlap/2); + } while (++cdownsample!=1) + bound = IMIN(bound, N/st->downsample); + for (i=bound;ipostfilter_period=IMAX(st->postfilter_period, COMBFILTER_MINPERIOD); + st->postfilter_period_old=IMAX(st->postfilter_period_old, COMBFILTER_MINPERIOD); + comb_filter(out_syn[c], out_syn[c], st->postfilter_period_old, st->postfilter_period, mode->shortMdctSize, + st->postfilter_gain_old, st->postfilter_gain, st->postfilter_tapset_old, st->postfilter_tapset, + mode->window, overlap); + if (LM!=0) + comb_filter(out_syn[c]+mode->shortMdctSize, out_syn[c]+mode->shortMdctSize, st->postfilter_period, postfilter_pitch, N-mode->shortMdctSize, + st->postfilter_gain, postfilter_gain, st->postfilter_tapset, postfilter_tapset, + mode->window, overlap); + + } while (++cpostfilter_period_old = st->postfilter_period; + st->postfilter_gain_old = st->postfilter_gain; + st->postfilter_tapset_old = st->postfilter_tapset; + st->postfilter_period = postfilter_pitch; + st->postfilter_gain = postfilter_gain; + st->postfilter_tapset = postfilter_tapset; + if (LM!=0) + { + st->postfilter_period_old = st->postfilter_period; + st->postfilter_gain_old = st->postfilter_gain; + st->postfilter_tapset_old = st->postfilter_tapset; + } + + if (C==1) { + for (i=0;istart;i++) + { + oldBandE[c*nbEBands+i]=0; + oldLogE[c*nbEBands+i]=oldLogE2[c*nbEBands+i]=-QCONST16(28.f,DB_SHIFT); + } + for (i=st->end;irng = dec->rng; + + /* We reuse freq[] as scratch space for the de-emphasis */ + deemphasis(out_syn, pcm, N, CC, st->downsample, mode->preemph, st->preemph_memD, freq); + st->loss_count = 0; + RESTORE_STACK; + if (ec_tell(dec) > 8*len) + return OPUS_INTERNAL_ERROR; + if(ec_get_error(dec)) + st->error = 1; + return frame_size/st->downsample; +} + + +#ifdef CUSTOM_MODES + +#ifdef FIXED_POINT +int opus_custom_decode(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, opus_int16 * OPUS_RESTRICT pcm, int frame_size) +{ + return celt_decode_with_ec(st, data, len, pcm, frame_size, NULL); +} + +#ifndef DISABLE_FLOAT_API +int opus_custom_decode_float(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, float * OPUS_RESTRICT pcm, int frame_size) +{ + int j, ret, C, N; + VARDECL(opus_int16, out); + ALLOC_STACK; + + if (pcm==NULL) + return OPUS_BAD_ARG; + + C = st->channels; + N = frame_size; + + ALLOC(out, C*N, opus_int16); + ret=celt_decode_with_ec(st, data, len, out, frame_size, NULL); + if (ret>0) + for (j=0;jchannels; + N = frame_size; + ALLOC(out, C*N, celt_sig); + + ret=celt_decode_with_ec(st, data, len, out, frame_size, NULL); + + if (ret>0) + for (j=0;j=st->mode->nbEBands) + goto bad_arg; + st->start = value; + } + break; + case CELT_SET_END_BAND_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>st->mode->nbEBands) + goto bad_arg; + st->end = value; + } + break; + case CELT_SET_CHANNELS_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>2) + goto bad_arg; + st->stream_channels = value; + } + break; + case CELT_GET_AND_CLEAR_ERROR_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (value==NULL) + goto bad_arg; + *value=st->error; + st->error = 0; + } + break; + case OPUS_GET_LOOKAHEAD_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (value==NULL) + goto bad_arg; + *value = st->overlap/st->downsample; + } + break; + case OPUS_RESET_STATE: + { + int i; + opus_val16 *lpc, *oldBandE, *oldLogE, *oldLogE2; + lpc = (opus_val16*)(st->_decode_mem+(DECODE_BUFFER_SIZE+st->overlap)*st->channels); + oldBandE = lpc+st->channels*LPC_ORDER; + oldLogE = oldBandE + 2*st->mode->nbEBands; + oldLogE2 = oldLogE + 2*st->mode->nbEBands; + OPUS_CLEAR((char*)&st->DECODER_RESET_START, + opus_custom_decoder_get_size(st->mode, st->channels)- + ((char*)&st->DECODER_RESET_START - (char*)st)); + for (i=0;i<2*st->mode->nbEBands;i++) + oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT); + } + break; + case OPUS_GET_PITCH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (value==NULL) + goto bad_arg; + *value = st->postfilter_period; + } + break; + case CELT_GET_MODE_REQUEST: + { + const CELTMode ** value = va_arg(ap, const CELTMode**); + if (value==0) + goto bad_arg; + *value=st->mode; + } + break; + case CELT_SET_SIGNALLING_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->signalling = value; + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + opus_uint32 * value = va_arg(ap, opus_uint32 *); + if (value==0) + goto bad_arg; + *value=st->rng; + } + break; + default: + goto bad_request; + } + va_end(ap); + return OPUS_OK; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +bad_request: + va_end(ap); + return OPUS_UNIMPLEMENTED; +} diff --git a/code/opus-1.1/celt/celt_encoder.c b/code/opus-1.1/celt/celt_encoder.c new file mode 100644 index 00000000..ffff0775 --- /dev/null +++ b/code/opus-1.1/celt/celt_encoder.c @@ -0,0 +1,2353 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2010 Xiph.Org Foundation + Copyright (c) 2008 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#define CELT_ENCODER_C + +#include "cpu_support.h" +#include "os_support.h" +#include "mdct.h" +#include +#include "celt.h" +#include "pitch.h" +#include "bands.h" +#include "modes.h" +#include "entcode.h" +#include "quant_bands.h" +#include "rate.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "float_cast.h" +#include +#include "celt_lpc.h" +#include "vq.h" + + +/** Encoder state + @brief Encoder state + */ +struct OpusCustomEncoder { + const OpusCustomMode *mode; /**< Mode used by the encoder */ + int overlap; + int channels; + int stream_channels; + + int force_intra; + int clip; + int disable_pf; + int complexity; + int upsample; + int start, end; + + opus_int32 bitrate; + int vbr; + int signalling; + int constrained_vbr; /* If zero, VBR can do whatever it likes with the rate */ + int loss_rate; + int lsb_depth; + int variable_duration; + int lfe; + int arch; + + /* Everything beyond this point gets cleared on a reset */ +#define ENCODER_RESET_START rng + + opus_uint32 rng; + int spread_decision; + opus_val32 delayedIntra; + int tonal_average; + int lastCodedBands; + int hf_average; + int tapset_decision; + + int prefilter_period; + opus_val16 prefilter_gain; + int prefilter_tapset; +#ifdef RESYNTH + int prefilter_period_old; + opus_val16 prefilter_gain_old; + int prefilter_tapset_old; +#endif + int consec_transient; + AnalysisInfo analysis; + + opus_val32 preemph_memE[2]; + opus_val32 preemph_memD[2]; + + /* VBR-related parameters */ + opus_int32 vbr_reservoir; + opus_int32 vbr_drift; + opus_int32 vbr_offset; + opus_int32 vbr_count; + opus_val32 overlap_max; + opus_val16 stereo_saving; + int intensity; + opus_val16 *energy_mask; + opus_val16 spec_avg; + +#ifdef RESYNTH + /* +MAX_PERIOD/2 to make space for overlap */ + celt_sig syn_mem[2][2*MAX_PERIOD+MAX_PERIOD/2]; +#endif + + celt_sig in_mem[1]; /* Size = channels*mode->overlap */ + /* celt_sig prefilter_mem[], Size = channels*COMBFILTER_MAXPERIOD */ + /* opus_val16 oldBandE[], Size = channels*mode->nbEBands */ + /* opus_val16 oldLogE[], Size = channels*mode->nbEBands */ + /* opus_val16 oldLogE2[], Size = channels*mode->nbEBands */ +}; + +int celt_encoder_get_size(int channels) +{ + CELTMode *mode = opus_custom_mode_create(48000, 960, NULL); + return opus_custom_encoder_get_size(mode, channels); +} + +OPUS_CUSTOM_NOSTATIC int opus_custom_encoder_get_size(const CELTMode *mode, int channels) +{ + int size = sizeof(struct CELTEncoder) + + (channels*mode->overlap-1)*sizeof(celt_sig) /* celt_sig in_mem[channels*mode->overlap]; */ + + channels*COMBFILTER_MAXPERIOD*sizeof(celt_sig) /* celt_sig prefilter_mem[channels*COMBFILTER_MAXPERIOD]; */ + + 3*channels*mode->nbEBands*sizeof(opus_val16); /* opus_val16 oldBandE[channels*mode->nbEBands]; */ + /* opus_val16 oldLogE[channels*mode->nbEBands]; */ + /* opus_val16 oldLogE2[channels*mode->nbEBands]; */ + return size; +} + +#ifdef CUSTOM_MODES +CELTEncoder *opus_custom_encoder_create(const CELTMode *mode, int channels, int *error) +{ + int ret; + CELTEncoder *st = (CELTEncoder *)opus_alloc(opus_custom_encoder_get_size(mode, channels)); + /* init will handle the NULL case */ + ret = opus_custom_encoder_init(st, mode, channels); + if (ret != OPUS_OK) + { + opus_custom_encoder_destroy(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} +#endif /* CUSTOM_MODES */ + +static int opus_custom_encoder_init_arch(CELTEncoder *st, const CELTMode *mode, + int channels, int arch) +{ + if (channels < 0 || channels > 2) + return OPUS_BAD_ARG; + + if (st==NULL || mode==NULL) + return OPUS_ALLOC_FAIL; + + OPUS_CLEAR((char*)st, opus_custom_encoder_get_size(mode, channels)); + + st->mode = mode; + st->overlap = mode->overlap; + st->stream_channels = st->channels = channels; + + st->upsample = 1; + st->start = 0; + st->end = st->mode->effEBands; + st->signalling = 1; + + st->arch = arch; + + st->constrained_vbr = 1; + st->clip = 1; + + st->bitrate = OPUS_BITRATE_MAX; + st->vbr = 0; + st->force_intra = 0; + st->complexity = 5; + st->lsb_depth=24; + + opus_custom_encoder_ctl(st, OPUS_RESET_STATE); + + return OPUS_OK; +} + +#ifdef CUSTOM_MODES +int opus_custom_encoder_init(CELTEncoder *st, const CELTMode *mode, int channels) +{ + return opus_custom_encoder_init_arch(st, mode, channels, opus_select_arch()); +} +#endif + +int celt_encoder_init(CELTEncoder *st, opus_int32 sampling_rate, int channels, + int arch) +{ + int ret; + ret = opus_custom_encoder_init_arch(st, + opus_custom_mode_create(48000, 960, NULL), channels, arch); + if (ret != OPUS_OK) + return ret; + st->upsample = resampling_factor(sampling_rate); + return OPUS_OK; +} + +#ifdef CUSTOM_MODES +void opus_custom_encoder_destroy(CELTEncoder *st) +{ + opus_free(st); +} +#endif /* CUSTOM_MODES */ + + +static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int C, + opus_val16 *tf_estimate, int *tf_chan) +{ + int i; + VARDECL(opus_val16, tmp); + opus_val32 mem0,mem1; + int is_transient = 0; + opus_int32 mask_metric = 0; + int c; + opus_val16 tf_max; + int len2; + /* Table of 6*64/x, trained on real data to minimize the average error */ + static const unsigned char inv_table[128] = { + 255,255,156,110, 86, 70, 59, 51, 45, 40, 37, 33, 31, 28, 26, 25, + 23, 22, 21, 20, 19, 18, 17, 16, 16, 15, 15, 14, 13, 13, 12, 12, + 12, 12, 11, 11, 11, 10, 10, 10, 9, 9, 9, 9, 9, 9, 8, 8, + 8, 8, 8, 7, 7, 7, 7, 7, 7, 6, 6, 6, 6, 6, 6, 6, + 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5, 5, 5, 5, + 5, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, + 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, + }; + SAVE_STACK; + ALLOC(tmp, len, opus_val16); + + len2=len/2; + for (c=0;c=0;i--) + { +#ifdef FIXED_POINT + /* FIXME: Use PSHR16() instead */ + tmp[i] = mem0 + PSHR32(tmp[i]-mem0,3); +#else + tmp[i] = mem0 + MULT16_16_P15(QCONST16(0.125f,15),tmp[i]-mem0); +#endif + mem0 = tmp[i]; + maxE = MAX16(maxE, mem0); + } + /*for (i=0;i>1))); +#else + mean = celt_sqrt(mean * maxE*.5*len2); +#endif + /* Inverse of the mean energy in Q15+6 */ + norm = SHL32(EXTEND32(len2),6+14)/ADD32(EPSILON,SHR32(mean,1)); + /* Compute harmonic mean discarding the unreliable boundaries + The data is smooth, so we only take 1/4th of the samples */ + unmask=0; + for (i=12;imask_metric) + { + *tf_chan = c; + mask_metric = unmask; + } + } + is_transient = mask_metric>200; + + /* Arbitrary metric for VBR boost */ + tf_max = MAX16(0,celt_sqrt(27*mask_metric)-42); + /* *tf_estimate = 1 + MIN16(1, sqrt(MAX16(0, tf_max-30))/20); */ + *tf_estimate = celt_sqrt(MAX16(0, SHL32(MULT16_16(QCONST16(0.0069,14),MIN16(163,tf_max)),14)-QCONST32(0.139,28))); + /*printf("%d %f\n", tf_max, mask_metric);*/ + RESTORE_STACK; +#ifdef FUZZING + is_transient = rand()&0x1; +#endif + /*printf("%d %f %d\n", is_transient, (float)*tf_estimate, tf_max);*/ + return is_transient; +} + +/* Looks for sudden increases of energy to decide whether we need to patch + the transient decision */ +int patch_transient_decision(opus_val16 *newE, opus_val16 *oldE, int nbEBands, + int end, int C) +{ + int i, c; + opus_val32 mean_diff=0; + opus_val16 spread_old[26]; + /* Apply an aggressive (-6 dB/Bark) spreading function to the old frame to + avoid false detection caused by irrelevant bands */ + if (C==1) + { + spread_old[0] = oldE[0]; + for (i=1;i=0;i--) + spread_old[i] = MAX16(spread_old[i], spread_old[i+1]-QCONST16(1.0f, DB_SHIFT)); + /* Compute mean increase */ + c=0; do { + for (i=2;i QCONST16(1.f, DB_SHIFT); +} + +/** Apply window and compute the MDCT for all sub-frames and + all channels in a frame */ +static void compute_mdcts(const CELTMode *mode, int shortBlocks, celt_sig * OPUS_RESTRICT in, + celt_sig * OPUS_RESTRICT out, int C, int CC, int LM, int upsample) +{ + const int overlap = OVERLAP(mode); + int N; + int B; + int shift; + int i, b, c; + if (shortBlocks) + { + B = shortBlocks; + N = mode->shortMdctSize; + shift = mode->maxLM; + } else { + B = 1; + N = mode->shortMdctSize<maxLM-LM; + } + c=0; do { + for (b=0;bmdct, in+c*(B*N+overlap)+b*N, &out[b+c*N*B], mode->window, overlap, shift, B); + } + } while (++ceBands[len]-m->eBands[len-1])<eBands[len]-m->eBands[len-1])<eBands[i+1]-m->eBands[i])<eBands[i+1]-m->eBands[i])==1; + for (j=0;jeBands[i]<eBands[i]<>LM, 1<>k, 1<=0;i--) + { + if (tf_res[i+1] == 1) + tf_res[i] = path1[i+1]; + else + tf_res[i] = path0[i+1]; + } + /*printf("%d %f\n", *tf_sum, tf_estimate);*/ + RESTORE_STACK; +#ifdef FUZZING + tf_select = rand()&0x1; + tf_res[0] = rand()&0x1; + for (i=1;istorage*8; + tell = ec_tell(enc); + logp = isTransient ? 2 : 4; + /* Reserve space to code the tf_select decision. */ + tf_select_rsv = LM>0 && tell+logp+1 <= budget; + budget -= tf_select_rsv; + curr = tf_changed = 0; + for (i=start;ieBands[i]<eBands[i+1]<eBands[i]<eBands[i+1]< QCONST16(.995f,10)) + trim_index-=4; + else if (sum > QCONST16(.92f,10)) + trim_index-=3; + else if (sum > QCONST16(.85f,10)) + trim_index-=2; + else if (sum > QCONST16(.8f,10)) + trim_index-=1; + /* mid-side savings estimations based on the LF average*/ + logXC = celt_log2(QCONST32(1.001f, 20)-MULT16_16(sum, sum)); + /* mid-side savings estimations based on min correlation */ + logXC2 = MAX16(HALF16(logXC), celt_log2(QCONST32(1.001f, 20)-MULT16_16(minXC, minXC))); +#ifdef FIXED_POINT + /* Compensate for Q20 vs Q14 input and convert output to Q8 */ + logXC = PSHR32(logXC-QCONST16(6.f, DB_SHIFT),DB_SHIFT-8); + logXC2 = PSHR32(logXC2-QCONST16(6.f, DB_SHIFT),DB_SHIFT-8); +#endif + + trim += MAX16(-QCONST16(4.f, 8), MULT16_16_Q15(QCONST16(.75f,15),logXC)); + *stereo_saving = MIN16(*stereo_saving + QCONST16(0.25f, 8), -HALF16(logXC2)); + } + + /* Estimate spectral tilt */ + c=0; do { + for (i=0;inbEBands]*(opus_int32)(2+2*i-end); + } + } while (++c QCONST16(2.f, DB_SHIFT)) + trim_index--; + if (diff > QCONST16(8.f, DB_SHIFT)) + trim_index--; + if (diff < -QCONST16(4.f, DB_SHIFT)) + trim_index++; + if (diff < -QCONST16(10.f, DB_SHIFT)) + trim_index++; + trim -= MAX16(-QCONST16(2.f, 8), MIN16(QCONST16(2.f, 8), SHR16(diff+QCONST16(1.f, DB_SHIFT),DB_SHIFT-8)/6 )); + trim -= SHR16(surround_trim, DB_SHIFT-8); + trim -= 2*SHR16(tf_estimate, 14-8); +#ifndef DISABLE_FLOAT_API + if (analysis->valid) + { + trim -= MAX16(-QCONST16(2.f, 8), MIN16(QCONST16(2.f, 8), + (opus_val16)(QCONST16(2.f, 8)*(analysis->tonality_slope+.05f)))); + } +#endif + +#ifdef FIXED_POINT + trim_index = PSHR32(trim, 8); +#else + trim_index = (int)floor(.5f+trim); +#endif + if (trim_index<0) + trim_index = 0; + if (trim_index>10) + trim_index = 10; + /*printf("%d\n", trim_index);*/ +#ifdef FUZZING + trim_index = rand()%11; +#endif + return trim_index; +} + +static int stereo_analysis(const CELTMode *m, const celt_norm *X, + int LM, int N0) +{ + int i; + int thetas; + opus_val32 sumLR = EPSILON, sumMS = EPSILON; + + /* Use the L1 norm to model the entropy of the L/R signal vs the M/S signal */ + for (i=0;i<13;i++) + { + int j; + for (j=m->eBands[i]<eBands[i+1]<eBands[13]<<(LM+1))+thetas, sumMS) + > MULT16_32_Q15(m->eBands[13]<<(LM+1), sumLR); +} + +static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16 *bandLogE2, + int nbEBands, int start, int end, int C, int *offsets, int lsb_depth, const opus_int16 *logN, + int isTransient, int vbr, int constrained_vbr, const opus_int16 *eBands, int LM, + int effectiveBytes, opus_int32 *tot_boost_, int lfe, opus_val16 *surround_dynalloc) +{ + int i, c; + opus_int32 tot_boost=0; + opus_val16 maxDepth; + VARDECL(opus_val16, follower); + VARDECL(opus_val16, noise_floor); + SAVE_STACK; + ALLOC(follower, C*nbEBands, opus_val16); + ALLOC(noise_floor, C*nbEBands, opus_val16); + for (i=0;i 50 && LM>=1 && !lfe) + { + int last=0; + c=0;do + { + follower[c*nbEBands] = bandLogE2[c*nbEBands]; + for (i=1;i bandLogE2[c*nbEBands+i-1]+QCONST16(.5f,DB_SHIFT)) + last=i; + follower[c*nbEBands+i] = MIN16(follower[c*nbEBands+i-1]+QCONST16(1.5f,DB_SHIFT), bandLogE2[c*nbEBands+i]); + } + for (i=last-1;i>=0;i--) + follower[c*nbEBands+i] = MIN16(follower[c*nbEBands+i], MIN16(follower[c*nbEBands+i+1]+QCONST16(2.f,DB_SHIFT), bandLogE2[c*nbEBands+i])); + for (i=0;i=12) + follower[i] = HALF16(follower[i]); + follower[i] = MIN16(follower[i], QCONST16(4, DB_SHIFT)); + + width = C*(eBands[i+1]-eBands[i])< 48) { + boost = (int)SHR32(EXTEND32(follower[i])*8,DB_SHIFT); + boost_bits = (boost*width<>BITRES>>3 > effectiveBytes/4) + { + opus_int32 cap = ((effectiveBytes/4)<mode; + ALLOC(_pre, CC*(N+COMBFILTER_MAXPERIOD), celt_sig); + + pre[0] = _pre; + pre[1] = _pre + (N+COMBFILTER_MAXPERIOD); + + + c=0; do { + OPUS_COPY(pre[c], prefilter_mem+c*COMBFILTER_MAXPERIOD, COMBFILTER_MAXPERIOD); + OPUS_COPY(pre[c]+COMBFILTER_MAXPERIOD, in+c*(N+st->overlap)+st->overlap, N); + } while (++c>1, opus_val16); + + pitch_downsample(pre, pitch_buf, COMBFILTER_MAXPERIOD+N, CC, st->arch); + /* Don't search for the fir last 1.5 octave of the range because + there's too many false-positives due to short-term correlation */ + pitch_search(pitch_buf+(COMBFILTER_MAXPERIOD>>1), pitch_buf, N, + COMBFILTER_MAXPERIOD-3*COMBFILTER_MINPERIOD, &pitch_index, + st->arch); + pitch_index = COMBFILTER_MAXPERIOD-pitch_index; + + gain1 = remove_doubling(pitch_buf, COMBFILTER_MAXPERIOD, COMBFILTER_MINPERIOD, + N, &pitch_index, st->prefilter_period, st->prefilter_gain); + if (pitch_index > COMBFILTER_MAXPERIOD-2) + pitch_index = COMBFILTER_MAXPERIOD-2; + gain1 = MULT16_16_Q15(QCONST16(.7f,15),gain1); + /*printf("%d %d %f %f\n", pitch_change, pitch_index, gain1, st->analysis.tonality);*/ + if (st->loss_rate>2) + gain1 = HALF32(gain1); + if (st->loss_rate>4) + gain1 = HALF32(gain1); + if (st->loss_rate>8) + gain1 = 0; + } else { + gain1 = 0; + pitch_index = COMBFILTER_MINPERIOD; + } + + /* Gain threshold for enabling the prefilter/postfilter */ + pf_threshold = QCONST16(.2f,15); + + /* Adjusting the threshold based on rate and continuity */ + if (abs(pitch_index-st->prefilter_period)*10>pitch_index) + pf_threshold += QCONST16(.2f,15); + if (nbAvailableBytes<25) + pf_threshold += QCONST16(.1f,15); + if (nbAvailableBytes<35) + pf_threshold += QCONST16(.1f,15); + if (st->prefilter_gain > QCONST16(.4f,15)) + pf_threshold -= QCONST16(.1f,15); + if (st->prefilter_gain > QCONST16(.55f,15)) + pf_threshold -= QCONST16(.1f,15); + + /* Hard threshold at 0.2 */ + pf_threshold = MAX16(pf_threshold, QCONST16(.2f,15)); + if (gain1prefilter_gain)prefilter_gain; + +#ifdef FIXED_POINT + qg = ((gain1+1536)>>10)/3-1; +#else + qg = (int)floor(.5f+gain1*32/3)-1; +#endif + qg = IMAX(0, IMIN(7, qg)); + gain1 = QCONST16(0.09375f,15)*(qg+1); + pf_on = 1; + } + /*printf("%d %f\n", pitch_index, gain1);*/ + + c=0; do { + int offset = mode->shortMdctSize-st->overlap; + st->prefilter_period=IMAX(st->prefilter_period, COMBFILTER_MINPERIOD); + OPUS_COPY(in+c*(N+st->overlap), st->in_mem+c*(st->overlap), st->overlap); + if (offset) + comb_filter(in+c*(N+st->overlap)+st->overlap, pre[c]+COMBFILTER_MAXPERIOD, + st->prefilter_period, st->prefilter_period, offset, -st->prefilter_gain, -st->prefilter_gain, + st->prefilter_tapset, st->prefilter_tapset, NULL, 0); + + comb_filter(in+c*(N+st->overlap)+st->overlap+offset, pre[c]+COMBFILTER_MAXPERIOD+offset, + st->prefilter_period, pitch_index, N-offset, -st->prefilter_gain, -gain1, + st->prefilter_tapset, prefilter_tapset, mode->window, st->overlap); + OPUS_COPY(st->in_mem+c*(st->overlap), in+c*(N+st->overlap)+N, st->overlap); + + if (N>COMBFILTER_MAXPERIOD) + { + OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, pre[c]+N, COMBFILTER_MAXPERIOD); + } else { + OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, prefilter_mem+c*COMBFILTER_MAXPERIOD+N, COMBFILTER_MAXPERIOD-N); + OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD+COMBFILTER_MAXPERIOD-N, pre[c]+COMBFILTER_MAXPERIOD, N); + } + } while (++cnbEBands; + eBands = mode->eBands; + + coded_bands = lastCodedBands ? lastCodedBands : nbEBands; + coded_bins = eBands[coded_bands]<analysis.activity, st->analysis.tonality, tf_estimate, st->stereo_saving, tot_boost, coded_bands);*/ +#ifndef DISABLE_FLOAT_API + if (analysis->valid && analysis->activity<.4) + target -= (opus_int32)((coded_bins<activity)); +#endif + /* Stereo savings */ + if (C==2) + { + int coded_stereo_bands; + int coded_stereo_dof; + opus_val16 max_frac; + coded_stereo_bands = IMIN(intensity, coded_bands); + coded_stereo_dof = (eBands[coded_stereo_bands]<valid && !lfe) + { + opus_int32 tonal_target; + float tonal; + + /* Tonality boost (compensating for the average). */ + tonal = MAX16(0.f,analysis->tonality-.15f)-0.09f; + tonal_target = target + (opus_int32)((coded_bins<tonality, tonal);*/ + target = tonal_target; + } +#endif + + if (has_surround_mask&&!lfe) + { + opus_int32 surround_target = target + (opus_int32)SHR32(MULT16_16(surround_masking,coded_bins<end, st->intensity, surround_target, target, st->bitrate);*/ + target = IMAX(target/4, surround_target); + } + + { + opus_int32 floor_depth; + int bins; + bins = eBands[nbEBands-2]<>2); + target = IMIN(target, floor_depth); + /*printf("%f %d\n", maxDepth, floor_depth);*/ + } + + if ((!has_surround_mask||lfe) && (constrained_vbr || bitrate<64000)) + { + opus_val16 rate_factor; +#ifdef FIXED_POINT + rate_factor = MAX16(0,(bitrate-32000)); +#else + rate_factor = MAX16(0,(1.f/32768)*(bitrate-32000)); +#endif + if (constrained_vbr) + rate_factor = MIN16(rate_factor, QCONST16(0.67f, 15)); + target = base_target + (opus_int32)MULT16_32_Q15(rate_factor, target-base_target); + + } + + if (!has_surround_mask && tf_estimate < QCONST16(.2f, 14)) + { + opus_val16 amount; + opus_val16 tvbr_factor; + amount = MULT16_16_Q15(QCONST16(.0000031f, 30), IMAX(0, IMIN(32000, 96000-bitrate))); + tvbr_factor = SHR32(MULT16_16(temporal_vbr, amount), DB_SHIFT); + target += (opus_int32)MULT16_32_Q15(tvbr_factor, target); + } + + /* Don't allow more than doubling the rate */ + target = IMIN(2*base_target, target); + + return target; +} + +int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes, ec_enc *enc) +{ + int i, c, N; + opus_int32 bits; + ec_enc _enc; + VARDECL(celt_sig, in); + VARDECL(celt_sig, freq); + VARDECL(celt_norm, X); + VARDECL(celt_ener, bandE); + VARDECL(opus_val16, bandLogE); + VARDECL(opus_val16, bandLogE2); + VARDECL(int, fine_quant); + VARDECL(opus_val16, error); + VARDECL(int, pulses); + VARDECL(int, cap); + VARDECL(int, offsets); + VARDECL(int, fine_priority); + VARDECL(int, tf_res); + VARDECL(unsigned char, collapse_masks); + celt_sig *prefilter_mem; + opus_val16 *oldBandE, *oldLogE, *oldLogE2; + int shortBlocks=0; + int isTransient=0; + const int CC = st->channels; + const int C = st->stream_channels; + int LM, M; + int tf_select; + int nbFilledBytes, nbAvailableBytes; + int effEnd; + int codedBands; + int tf_sum; + int alloc_trim; + int pitch_index=COMBFILTER_MINPERIOD; + opus_val16 gain1 = 0; + int dual_stereo=0; + int effectiveBytes; + int dynalloc_logp; + opus_int32 vbr_rate; + opus_int32 total_bits; + opus_int32 total_boost; + opus_int32 balance; + opus_int32 tell; + int prefilter_tapset=0; + int pf_on; + int anti_collapse_rsv; + int anti_collapse_on=0; + int silence=0; + int tf_chan = 0; + opus_val16 tf_estimate; + int pitch_change=0; + opus_int32 tot_boost; + opus_val32 sample_max; + opus_val16 maxDepth; + const OpusCustomMode *mode; + int nbEBands; + int overlap; + const opus_int16 *eBands; + int secondMdct; + int signalBandwidth; + int transient_got_disabled=0; + opus_val16 surround_masking=0; + opus_val16 temporal_vbr=0; + opus_val16 surround_trim = 0; + opus_int32 equiv_rate = 510000; + VARDECL(opus_val16, surround_dynalloc); + ALLOC_STACK; + + mode = st->mode; + nbEBands = mode->nbEBands; + overlap = mode->overlap; + eBands = mode->eBands; + tf_estimate = 0; + if (nbCompressedBytes<2 || pcm==NULL) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + + frame_size *= st->upsample; + for (LM=0;LM<=mode->maxLM;LM++) + if (mode->shortMdctSize<mode->maxLM) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + M=1<shortMdctSize; + + prefilter_mem = st->in_mem+CC*(st->overlap); + oldBandE = (opus_val16*)(st->in_mem+CC*(st->overlap+COMBFILTER_MAXPERIOD)); + oldLogE = oldBandE + CC*nbEBands; + oldLogE2 = oldLogE + CC*nbEBands; + + if (enc==NULL) + { + tell=1; + nbFilledBytes=0; + } else { + tell=ec_tell(enc); + nbFilledBytes=(tell+4)>>3; + } + +#ifdef CUSTOM_MODES + if (st->signalling && enc==NULL) + { + int tmp = (mode->effEBands-st->end)>>1; + st->end = IMAX(1, mode->effEBands-tmp); + compressed[0] = tmp<<5; + compressed[0] |= LM<<3; + compressed[0] |= (C==2)<<2; + /* Convert "standard mode" to Opus header */ + if (mode->Fs==48000 && mode->shortMdctSize==120) + { + int c0 = toOpus(compressed[0]); + if (c0<0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + compressed[0] = c0; + } + compressed++; + nbCompressedBytes--; + } +#else + celt_assert(st->signalling==0); +#endif + + /* Can't produce more than 1275 output bytes */ + nbCompressedBytes = IMIN(nbCompressedBytes,1275); + nbAvailableBytes = nbCompressedBytes - nbFilledBytes; + + if (st->vbr && st->bitrate!=OPUS_BITRATE_MAX) + { + opus_int32 den=mode->Fs>>BITRES; + vbr_rate=(st->bitrate*frame_size+(den>>1))/den; +#ifdef CUSTOM_MODES + if (st->signalling) + vbr_rate -= 8<>(3+BITRES); + } else { + opus_int32 tmp; + vbr_rate = 0; + tmp = st->bitrate*frame_size; + if (tell>1) + tmp += tell; + if (st->bitrate!=OPUS_BITRATE_MAX) + nbCompressedBytes = IMAX(2, IMIN(nbCompressedBytes, + (tmp+4*mode->Fs)/(8*mode->Fs)-!!st->signalling)); + effectiveBytes = nbCompressedBytes; + } + if (st->bitrate != OPUS_BITRATE_MAX) + equiv_rate = st->bitrate - (40*C+20)*((400>>LM) - 50); + + if (enc==NULL) + { + ec_enc_init(&_enc, compressed, nbCompressedBytes); + enc = &_enc; + } + + if (vbr_rate>0) + { + /* Computes the max bit-rate allowed in VBR mode to avoid violating the + target rate and buffering. + We must do this up front so that bust-prevention logic triggers + correctly if we don't have enough bits. */ + if (st->constrained_vbr) + { + opus_int32 vbr_bound; + opus_int32 max_allowed; + /* We could use any multiple of vbr_rate as bound (depending on the + delay). + This is clamped to ensure we use at least two bytes if the encoder + was entirely empty, but to allow 0 in hybrid mode. */ + vbr_bound = vbr_rate; + max_allowed = IMIN(IMAX(tell==1?2:0, + (vbr_rate+vbr_bound-st->vbr_reservoir)>>(BITRES+3)), + nbAvailableBytes); + if(max_allowed < nbAvailableBytes) + { + nbCompressedBytes = nbFilledBytes+max_allowed; + nbAvailableBytes = max_allowed; + ec_enc_shrink(enc, nbCompressedBytes); + } + } + } + total_bits = nbCompressedBytes*8; + + effEnd = st->end; + if (effEnd > mode->effEBands) + effEnd = mode->effEBands; + + ALLOC(in, CC*(N+st->overlap), celt_sig); + + sample_max=MAX32(st->overlap_max, celt_maxabs16(pcm, C*(N-overlap)/st->upsample)); + st->overlap_max=celt_maxabs16(pcm+C*(N-overlap)/st->upsample, C*overlap/st->upsample); + sample_max=MAX32(sample_max, st->overlap_max); +#ifdef FIXED_POINT + silence = (sample_max==0); +#else + silence = (sample_max <= (opus_val16)1/(1<lsb_depth)); +#endif +#ifdef FUZZING + if ((rand()&0x3F)==0) + silence = 1; +#endif + if (tell==1) + ec_enc_bit_logp(enc, silence, 15); + else + silence=0; + if (silence) + { + /*In VBR mode there is no need to send more than the minimum. */ + if (vbr_rate>0) + { + effectiveBytes=nbCompressedBytes=IMIN(nbCompressedBytes, nbFilledBytes+2); + total_bits=nbCompressedBytes*8; + nbAvailableBytes=2; + ec_enc_shrink(enc, nbCompressedBytes); + } + /* Pretend we've filled all the remaining bits with zeros + (that's what the initialiser did anyway) */ + tell = nbCompressedBytes*8; + enc->nbits_total+=tell-ec_tell(enc); + } + c=0; do { + celt_preemphasis(pcm+c, in+c*(N+st->overlap)+st->overlap, N, CC, st->upsample, + mode->preemph, st->preemph_memE+c, st->clip); + } while (++clfe&&nbAvailableBytes>3) || nbAvailableBytes>12*C) && st->start==0 && !silence && !st->disable_pf + && st->complexity >= 5 && !(st->consec_transient && LM!=3 && st->variable_duration==OPUS_FRAMESIZE_VARIABLE); + + prefilter_tapset = st->tapset_decision; + pf_on = run_prefilter(st, in, prefilter_mem, CC, N, prefilter_tapset, &pitch_index, &gain1, &qg, enabled, nbAvailableBytes); + if ((gain1 > QCONST16(.4f,15) || st->prefilter_gain > QCONST16(.4f,15)) && (!st->analysis.valid || st->analysis.tonality > .3) + && (pitch_index > 1.26*st->prefilter_period || pitch_index < .79*st->prefilter_period)) + pitch_change = 1; + if (pf_on==0) + { + if(st->start==0 && tell+16<=total_bits) + ec_enc_bit_logp(enc, 0, 1); + } else { + /*This block is not gated by a total bits check only because + of the nbAvailableBytes check above.*/ + int octave; + ec_enc_bit_logp(enc, 1, 1); + pitch_index += 1; + octave = EC_ILOG(pitch_index)-5; + ec_enc_uint(enc, octave, 6); + ec_enc_bits(enc, pitch_index-(16<complexity >= 1 && !st->lfe) + { + isTransient = transient_analysis(in, N+st->overlap, CC, + &tf_estimate, &tf_chan); + } + if (LM>0 && ec_tell(enc)+3<=total_bits) + { + if (isTransient) + shortBlocks = M; + } else { + isTransient = 0; + transient_got_disabled=1; + } + + ALLOC(freq, CC*N, celt_sig); /**< Interleaved signal MDCTs */ + ALLOC(bandE,nbEBands*CC, celt_ener); + ALLOC(bandLogE,nbEBands*CC, opus_val16); + + secondMdct = shortBlocks && st->complexity>=8; + ALLOC(bandLogE2, C*nbEBands, opus_val16); + if (secondMdct) + { + compute_mdcts(mode, 0, in, freq, C, CC, LM, st->upsample); + compute_band_energies(mode, freq, bandE, effEnd, C, M); + amp2Log2(mode, effEnd, st->end, bandE, bandLogE2, C); + for (i=0;iupsample); + if (CC==2&&C==1) + tf_chan = 0; + compute_band_energies(mode, freq, bandE, effEnd, C, M); + + if (st->lfe) + { + for (i=2;iend;i++) + { + bandE[i] = IMIN(bandE[i], MULT16_32_Q15(QCONST16(1e-4f,15),bandE[0])); + bandE[i] = MAX32(bandE[i], EPSILON); + } + } + amp2Log2(mode, effEnd, st->end, bandE, bandLogE, C); + + ALLOC(surround_dynalloc, C*nbEBands, opus_val16); + for(i=0;iend;i++) + surround_dynalloc[i] = 0; + /* This computes how much masking takes place between surround channels */ + if (st->start==0&&st->energy_mask&&!st->lfe) + { + int mask_end; + int midband; + int count_dynalloc; + opus_val32 mask_avg=0; + opus_val32 diff=0; + int count=0; + mask_end = IMAX(2,st->lastCodedBands); + for (c=0;cenergy_mask[nbEBands*c+i], + QCONST16(.25f, DB_SHIFT)), -QCONST16(2.0f, DB_SHIFT)); + if (mask > 0) + mask = HALF16(mask); + mask_avg += MULT16_16(mask, eBands[i+1]-eBands[i]); + count += eBands[i+1]-eBands[i]; + diff += MULT16_16(mask, 1+2*i-mask_end); + } + } + mask_avg = DIV32_16(mask_avg,count); + mask_avg += QCONST16(.2f, DB_SHIFT); + diff = diff*6/(C*(mask_end-1)*(mask_end+1)*mask_end); + /* Again, being conservative */ + diff = HALF32(diff); + diff = MAX32(MIN32(diff, QCONST32(.031f, DB_SHIFT)), -QCONST32(.031f, DB_SHIFT)); + /* Find the band that's in the middle of the coded spectrum */ + for (midband=0;eBands[midband+1] < eBands[mask_end]/2;midband++); + count_dynalloc=0; + for(i=0;ienergy_mask[i], st->energy_mask[nbEBands+i]); + else + unmask = st->energy_mask[i]; + unmask = MIN16(unmask, QCONST16(.0f, DB_SHIFT)); + unmask -= lin; + if (unmask > QCONST16(.25f, DB_SHIFT)) + { + surround_dynalloc[i] = unmask - QCONST16(.25f, DB_SHIFT); + count_dynalloc++; + } + } + if (count_dynalloc>=3) + { + /* If we need dynalloc in many bands, it's probably because our + initial masking rate was too low. */ + mask_avg += QCONST16(.25f, DB_SHIFT); + if (mask_avg>0) + { + /* Something went really wrong in the original calculations, + disabling masking. */ + mask_avg = 0; + diff = 0; + for(i=0;ilfe) + { + opus_val16 follow=-QCONST16(10.0f,DB_SHIFT); + opus_val32 frame_avg=0; + opus_val16 offset = shortBlocks?HALF16(SHL16(LM, DB_SHIFT)):0; + for(i=st->start;iend;i++) + { + follow = MAX16(follow-QCONST16(1.f, DB_SHIFT), bandLogE[i]-offset); + if (C==2) + follow = MAX16(follow, bandLogE[i+nbEBands]-offset); + frame_avg += follow; + } + frame_avg /= (st->end-st->start); + temporal_vbr = SUB16(frame_avg,st->spec_avg); + temporal_vbr = MIN16(QCONST16(3.f, DB_SHIFT), MAX16(-QCONST16(1.5f, DB_SHIFT), temporal_vbr)); + st->spec_avg += MULT16_16_Q15(QCONST16(.02f, 15), temporal_vbr); + } + /*for (i=0;i<21;i++) + printf("%f ", bandLogE[i]); + printf("\n");*/ + + if (!secondMdct) + { + for (i=0;i0 && ec_tell(enc)+3<=total_bits && !isTransient && st->complexity>=5 && !st->lfe) + { + if (patch_transient_decision(bandLogE, oldBandE, nbEBands, st->end, C)) + { + isTransient = 1; + shortBlocks = M; + compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample); + compute_band_energies(mode, freq, bandE, effEnd, C, M); + amp2Log2(mode, effEnd, st->end, bandE, bandLogE, C); + /* Compensate for the scaling of short vs long mdcts */ + for (i=0;i0 && ec_tell(enc)+3<=total_bits) + ec_enc_bit_logp(enc, isTransient, 3); + + ALLOC(X, C*N, celt_norm); /**< Interleaved normalised MDCTs */ + + /* Band normalisation */ + normalise_bands(mode, freq, X, bandE, effEnd, C, M); + + ALLOC(tf_res, nbEBands, int); + /* Disable variable tf resolution for hybrid and at very low bitrate */ + if (effectiveBytes>=15*C && st->start==0 && st->complexity>=2 && !st->lfe) + { + int lambda; + if (effectiveBytes<40) + lambda = 12; + else if (effectiveBytes<60) + lambda = 6; + else if (effectiveBytes<100) + lambda = 4; + else + lambda = 3; + lambda*=2; + tf_select = tf_analysis(mode, effEnd, isTransient, tf_res, lambda, X, N, LM, &tf_sum, tf_estimate, tf_chan); + for (i=effEnd;iend;i++) + tf_res[i] = tf_res[effEnd-1]; + } else { + tf_sum = 0; + for (i=0;iend;i++) + tf_res[i] = isTransient; + tf_select=0; + } + + ALLOC(error, C*nbEBands, opus_val16); + quant_coarse_energy(mode, st->start, st->end, effEnd, bandLogE, + oldBandE, total_bits, error, enc, + C, LM, nbAvailableBytes, st->force_intra, + &st->delayedIntra, st->complexity >= 4, st->loss_rate, st->lfe); + + tf_encode(st->start, st->end, isTransient, tf_res, LM, tf_select, enc); + + if (ec_tell(enc)+4<=total_bits) + { + if (st->lfe) + { + st->tapset_decision = 0; + st->spread_decision = SPREAD_NORMAL; + } else if (shortBlocks || st->complexity < 3 || nbAvailableBytes < 10*C || st->start != 0) + { + if (st->complexity == 0) + st->spread_decision = SPREAD_NONE; + else + st->spread_decision = SPREAD_NORMAL; + } else { + /* Disable new spreading+tapset estimator until we can show it works + better than the old one. So far it seems like spreading_decision() + works best. */ +#if 0 + if (st->analysis.valid) + { + static const opus_val16 spread_thresholds[3] = {-QCONST16(.6f, 15), -QCONST16(.2f, 15), -QCONST16(.07f, 15)}; + static const opus_val16 spread_histeresis[3] = {QCONST16(.15f, 15), QCONST16(.07f, 15), QCONST16(.02f, 15)}; + static const opus_val16 tapset_thresholds[2] = {QCONST16(.0f, 15), QCONST16(.15f, 15)}; + static const opus_val16 tapset_histeresis[2] = {QCONST16(.1f, 15), QCONST16(.05f, 15)}; + st->spread_decision = hysteresis_decision(-st->analysis.tonality, spread_thresholds, spread_histeresis, 3, st->spread_decision); + st->tapset_decision = hysteresis_decision(st->analysis.tonality_slope, tapset_thresholds, tapset_histeresis, 2, st->tapset_decision); + } else +#endif + { + st->spread_decision = spreading_decision(mode, X, + &st->tonal_average, st->spread_decision, &st->hf_average, + &st->tapset_decision, pf_on&&!shortBlocks, effEnd, C, M); + } + /*printf("%d %d\n", st->tapset_decision, st->spread_decision);*/ + /*printf("%f %d %f %d\n\n", st->analysis.tonality, st->spread_decision, st->analysis.tonality_slope, st->tapset_decision);*/ + } + ec_enc_icdf(enc, st->spread_decision, spread_icdf, 5); + } + + ALLOC(offsets, nbEBands, int); + + maxDepth = dynalloc_analysis(bandLogE, bandLogE2, nbEBands, st->start, st->end, C, offsets, + st->lsb_depth, mode->logN, isTransient, st->vbr, st->constrained_vbr, + eBands, LM, effectiveBytes, &tot_boost, st->lfe, surround_dynalloc); + /* For LFE, everything interesting is in the first band */ + if (st->lfe) + offsets[0] = IMIN(8, effectiveBytes/3); + ALLOC(cap, nbEBands, int); + init_caps(mode,cap,LM,C); + + dynalloc_logp = 6; + total_bits<<=BITRES; + total_boost = 0; + tell = ec_tell_frac(enc); + for (i=st->start;iend;i++) + { + int width, quanta; + int dynalloc_loop_logp; + int boost; + int j; + width = C*(eBands[i+1]-eBands[i])<intensity = hysteresis_decision((opus_val16)(equiv_rate/1000), + intensity_thresholds, intensity_histeresis, 21, st->intensity); + st->intensity = IMIN(st->end,IMAX(st->start, st->intensity)); + } + + alloc_trim = 5; + if (tell+(6<lfe) + alloc_trim = 5; + else + alloc_trim = alloc_trim_analysis(mode, X, bandLogE, + st->end, LM, C, N, &st->analysis, &st->stereo_saving, tf_estimate, st->intensity, surround_trim); + ec_enc_icdf(enc, alloc_trim, trim_icdf, 7); + tell = ec_tell_frac(enc); + } + + /* Variable bitrate */ + if (vbr_rate>0) + { + opus_val16 alpha; + opus_int32 delta; + /* The target rate in 8th bits per frame */ + opus_int32 target, base_target; + opus_int32 min_allowed; + int lm_diff = mode->maxLM - LM; + + /* Don't attempt to use more than 510 kb/s, even for frames smaller than 20 ms. + The CELT allocator will just not be able to use more than that anyway. */ + nbCompressedBytes = IMIN(nbCompressedBytes,1275>>(3-LM)); + base_target = vbr_rate - ((40*C+20)<constrained_vbr) + base_target += (st->vbr_offset>>lm_diff); + + target = compute_vbr(mode, &st->analysis, base_target, LM, equiv_rate, + st->lastCodedBands, C, st->intensity, st->constrained_vbr, + st->stereo_saving, tot_boost, tf_estimate, pitch_change, maxDepth, + st->variable_duration, st->lfe, st->energy_mask!=NULL, surround_masking, + temporal_vbr); + + /* The current offset is removed from the target and the space used + so far is added*/ + target=target+tell; + /* In VBR mode the frame size must not be reduced so much that it would + result in the encoder running out of bits. + The margin of 2 bytes ensures that none of the bust-prevention logic + in the decoder will have triggered so far. */ + min_allowed = ((tell+total_boost+(1<<(BITRES+3))-1)>>(BITRES+3)) + 2 - nbFilledBytes; + + nbAvailableBytes = (target+(1<<(BITRES+2)))>>(BITRES+3); + nbAvailableBytes = IMAX(min_allowed,nbAvailableBytes); + nbAvailableBytes = IMIN(nbCompressedBytes,nbAvailableBytes+nbFilledBytes) - nbFilledBytes; + + /* By how much did we "miss" the target on that frame */ + delta = target - vbr_rate; + + target=nbAvailableBytes<<(BITRES+3); + + /*If the frame is silent we don't adjust our drift, otherwise + the encoder will shoot to very high rates after hitting a + span of silence, but we do allow the bitres to refill. + This means that we'll undershoot our target in CVBR/VBR modes + on files with lots of silence. */ + if(silence) + { + nbAvailableBytes = 2; + target = 2*8<vbr_count < 970) + { + st->vbr_count++; + alpha = celt_rcp(SHL32(EXTEND32(st->vbr_count+20),16)); + } else + alpha = QCONST16(.001f,15); + /* How many bits have we used in excess of what we're allowed */ + if (st->constrained_vbr) + st->vbr_reservoir += target - vbr_rate; + /*printf ("%d\n", st->vbr_reservoir);*/ + + /* Compute the offset we need to apply in order to reach the target */ + if (st->constrained_vbr) + { + st->vbr_drift += (opus_int32)MULT16_32_Q15(alpha,(delta*(1<vbr_offset-st->vbr_drift); + st->vbr_offset = -st->vbr_drift; + } + /*printf ("%d\n", st->vbr_drift);*/ + + if (st->constrained_vbr && st->vbr_reservoir < 0) + { + /* We're under the min value -- increase rate */ + int adjust = (-st->vbr_reservoir)/(8<vbr_reservoir = 0; + /*printf ("+%d\n", adjust);*/ + } + nbCompressedBytes = IMIN(nbCompressedBytes,nbAvailableBytes+nbFilledBytes); + /*printf("%d\n", nbCompressedBytes*50*8);*/ + /* This moves the raw bits to take into account the new compressed size */ + ec_enc_shrink(enc, nbCompressedBytes); + } + + /* Bit allocation */ + ALLOC(fine_quant, nbEBands, int); + ALLOC(pulses, nbEBands, int); + ALLOC(fine_priority, nbEBands, int); + + /* bits = packet size - where we are - safety*/ + bits = (((opus_int32)nbCompressedBytes*8)<=2&&bits>=((LM+2)<end-1; +#ifndef DISABLE_FLOAT_API + if (st->analysis.valid) + { + int min_bandwidth; + if (equiv_rate < (opus_int32)32000*C) + min_bandwidth = 13; + else if (equiv_rate < (opus_int32)48000*C) + min_bandwidth = 16; + else if (equiv_rate < (opus_int32)60000*C) + min_bandwidth = 18; + else if (equiv_rate < (opus_int32)80000*C) + min_bandwidth = 19; + else + min_bandwidth = 20; + signalBandwidth = IMAX(st->analysis.bandwidth, min_bandwidth); + } +#endif + if (st->lfe) + signalBandwidth = 1; + codedBands = compute_allocation(mode, st->start, st->end, offsets, cap, + alloc_trim, &st->intensity, &dual_stereo, bits, &balance, pulses, + fine_quant, fine_priority, C, LM, enc, 1, st->lastCodedBands, signalBandwidth); + if (st->lastCodedBands) + st->lastCodedBands = IMIN(st->lastCodedBands+1,IMAX(st->lastCodedBands-1,codedBands)); + else + st->lastCodedBands = codedBands; + + quant_fine_energy(mode, st->start, st->end, oldBandE, error, fine_quant, enc, C); + + /* Residual quantisation */ + ALLOC(collapse_masks, C*nbEBands, unsigned char); + quant_all_bands(1, mode, st->start, st->end, X, C==2 ? X+N : NULL, collapse_masks, + bandE, pulses, shortBlocks, st->spread_decision, dual_stereo, st->intensity, tf_res, + nbCompressedBytes*(8<rng); + + if (anti_collapse_rsv > 0) + { + anti_collapse_on = st->consec_transient<2; +#ifdef FUZZING + anti_collapse_on = rand()&0x1; +#endif + ec_enc_bits(enc, anti_collapse_on, 1); + } + quant_energy_finalise(mode, st->start, st->end, oldBandE, error, fine_quant, fine_priority, nbCompressedBytes*8-ec_tell(enc), enc, C); + + if (silence) + { + for (i=0;istart, st->end, oldBandE, oldLogE, oldLogE2, pulses, st->rng); + } + + if (silence) + { + for (i=0;istart, effEnd, C, M); + } + + c=0; do { + OPUS_MOVE(st->syn_mem[c], st->syn_mem[c]+N, 2*MAX_PERIOD-N+overlap/2); + } while (++csyn_mem[c]+2*MAX_PERIOD-N; + } while (++cprefilter_period=IMAX(st->prefilter_period, COMBFILTER_MINPERIOD); + st->prefilter_period_old=IMAX(st->prefilter_period_old, COMBFILTER_MINPERIOD); + comb_filter(out_mem[c], out_mem[c], st->prefilter_period_old, st->prefilter_period, mode->shortMdctSize, + st->prefilter_gain_old, st->prefilter_gain, st->prefilter_tapset_old, st->prefilter_tapset, + mode->window, st->overlap); + if (LM!=0) + comb_filter(out_mem[c]+mode->shortMdctSize, out_mem[c]+mode->shortMdctSize, st->prefilter_period, pitch_index, N-mode->shortMdctSize, + st->prefilter_gain, gain1, st->prefilter_tapset, prefilter_tapset, + mode->window, overlap); + } while (++cupsample, mode->preemph, st->preemph_memD, freq); + st->prefilter_period_old = st->prefilter_period; + st->prefilter_gain_old = st->prefilter_gain; + st->prefilter_tapset_old = st->prefilter_tapset; + } +#endif + + st->prefilter_period = pitch_index; + st->prefilter_gain = gain1; + st->prefilter_tapset = prefilter_tapset; +#ifdef RESYNTH + if (LM!=0) + { + st->prefilter_period_old = st->prefilter_period; + st->prefilter_gain_old = st->prefilter_gain; + st->prefilter_tapset_old = st->prefilter_tapset; + } +#endif + + if (CC==2&&C==1) { + for (i=0;istart;i++) + { + oldBandE[c*nbEBands+i]=0; + oldLogE[c*nbEBands+i]=oldLogE2[c*nbEBands+i]=-QCONST16(28.f,DB_SHIFT); + } + for (i=st->end;iconsec_transient++; + else + st->consec_transient=0; + st->rng = enc->rng; + + /* If there's any room left (can only happen for very high rates), + it's already filled with zeros */ + ec_enc_done(enc); + +#ifdef CUSTOM_MODES + if (st->signalling) + nbCompressedBytes++; +#endif + + RESTORE_STACK; + if (ec_get_error(enc)) + return OPUS_INTERNAL_ERROR; + else + return nbCompressedBytes; +} + + +#ifdef CUSTOM_MODES + +#ifdef FIXED_POINT +int opus_custom_encode(CELTEncoder * OPUS_RESTRICT st, const opus_int16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes) +{ + return celt_encode_with_ec(st, pcm, frame_size, compressed, nbCompressedBytes, NULL); +} + +#ifndef DISABLE_FLOAT_API +int opus_custom_encode_float(CELTEncoder * OPUS_RESTRICT st, const float * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes) +{ + int j, ret, C, N; + VARDECL(opus_int16, in); + ALLOC_STACK; + + if (pcm==NULL) + return OPUS_BAD_ARG; + + C = st->channels; + N = frame_size; + ALLOC(in, C*N, opus_int16); + + for (j=0;jchannels; + N=frame_size; + ALLOC(in, C*N, celt_sig); + for (j=0;j10) + goto bad_arg; + st->complexity = value; + } + break; + case CELT_SET_START_BAND_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<0 || value>=st->mode->nbEBands) + goto bad_arg; + st->start = value; + } + break; + case CELT_SET_END_BAND_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>st->mode->nbEBands) + goto bad_arg; + st->end = value; + } + break; + case CELT_SET_PREDICTION_REQUEST: + { + int value = va_arg(ap, opus_int32); + if (value<0 || value>2) + goto bad_arg; + st->disable_pf = value<=1; + st->force_intra = value==0; + } + break; + case OPUS_SET_PACKET_LOSS_PERC_REQUEST: + { + int value = va_arg(ap, opus_int32); + if (value<0 || value>100) + goto bad_arg; + st->loss_rate = value; + } + break; + case OPUS_SET_VBR_CONSTRAINT_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->constrained_vbr = value; + } + break; + case OPUS_SET_VBR_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->vbr = value; + } + break; + case OPUS_SET_BITRATE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<=500 && value!=OPUS_BITRATE_MAX) + goto bad_arg; + value = IMIN(value, 260000*st->channels); + st->bitrate = value; + } + break; + case CELT_SET_CHANNELS_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>2) + goto bad_arg; + st->stream_channels = value; + } + break; + case OPUS_SET_LSB_DEPTH_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<8 || value>24) + goto bad_arg; + st->lsb_depth=value; + } + break; + case OPUS_GET_LSB_DEPTH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + *value=st->lsb_depth; + } + break; + case OPUS_SET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->variable_duration = value; + } + break; + case OPUS_RESET_STATE: + { + int i; + opus_val16 *oldBandE, *oldLogE, *oldLogE2; + oldBandE = (opus_val16*)(st->in_mem+st->channels*(st->overlap+COMBFILTER_MAXPERIOD)); + oldLogE = oldBandE + st->channels*st->mode->nbEBands; + oldLogE2 = oldLogE + st->channels*st->mode->nbEBands; + OPUS_CLEAR((char*)&st->ENCODER_RESET_START, + opus_custom_encoder_get_size(st->mode, st->channels)- + ((char*)&st->ENCODER_RESET_START - (char*)st)); + for (i=0;ichannels*st->mode->nbEBands;i++) + oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT); + st->vbr_offset = 0; + st->delayedIntra = 1; + st->spread_decision = SPREAD_NORMAL; + st->tonal_average = 256; + st->hf_average = 0; + st->tapset_decision = 0; + } + break; +#ifdef CUSTOM_MODES + case CELT_SET_INPUT_CLIPPING_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->clip = value; + } + break; +#endif + case CELT_SET_SIGNALLING_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->signalling = value; + } + break; + case CELT_SET_ANALYSIS_REQUEST: + { + AnalysisInfo *info = va_arg(ap, AnalysisInfo *); + if (info) + OPUS_COPY(&st->analysis, info, 1); + } + break; + case CELT_GET_MODE_REQUEST: + { + const CELTMode ** value = va_arg(ap, const CELTMode**); + if (value==0) + goto bad_arg; + *value=st->mode; + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + opus_uint32 * value = va_arg(ap, opus_uint32 *); + if (value==0) + goto bad_arg; + *value=st->rng; + } + break; + case OPUS_SET_LFE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->lfe = value; + } + break; + case OPUS_SET_ENERGY_MASK_REQUEST: + { + opus_val16 *value = va_arg(ap, opus_val16*); + st->energy_mask = value; + } + break; + default: + goto bad_request; + } + va_end(ap); + return OPUS_OK; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +bad_request: + va_end(ap); + return OPUS_UNIMPLEMENTED; +} diff --git a/code/opus-1.1/celt/celt_lpc.c b/code/opus-1.1/celt/celt_lpc.c new file mode 100644 index 00000000..fa29d626 --- /dev/null +++ b/code/opus-1.1/celt/celt_lpc.c @@ -0,0 +1,309 @@ +/* Copyright (c) 2009-2010 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "celt_lpc.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "pitch.h" + +void _celt_lpc( + opus_val16 *_lpc, /* out: [0...p-1] LPC coefficients */ +const opus_val32 *ac, /* in: [0...p] autocorrelation values */ +int p +) +{ + int i, j; + opus_val32 r; + opus_val32 error = ac[0]; +#ifdef FIXED_POINT + opus_val32 lpc[LPC_ORDER]; +#else + float *lpc = _lpc; +#endif + + for (i = 0; i < p; i++) + lpc[i] = 0; + if (ac[0] != 0) + { + for (i = 0; i < p; i++) { + /* Sum up this iteration's reflection coefficient */ + opus_val32 rr = 0; + for (j = 0; j < i; j++) + rr += MULT32_32_Q31(lpc[j],ac[i - j]); + rr += SHR32(ac[i + 1],3); + r = -frac_div32(SHL32(rr,3), error); + /* Update LPC coefficients and total error */ + lpc[i] = SHR32(r,3); + for (j = 0; j < (i+1)>>1; j++) + { + opus_val32 tmp1, tmp2; + tmp1 = lpc[j]; + tmp2 = lpc[i-1-j]; + lpc[j] = tmp1 + MULT32_32_Q31(r,tmp2); + lpc[i-1-j] = tmp2 + MULT32_32_Q31(r,tmp1); + } + + error = error - MULT32_32_Q31(MULT32_32_Q31(r,r),error); + /* Bail out once we get 30 dB gain */ +#ifdef FIXED_POINT + if (error=1;j--) + { + mem[j]=mem[j-1]; + } + mem[0] = ROUND16(sum,SIG_SHIFT); + _y[i] = sum; + } +#else + int i,j; + VARDECL(opus_val16, rden); + VARDECL(opus_val16, y); + SAVE_STACK; + + celt_assert((ord&3)==0); + ALLOC(rden, ord, opus_val16); + ALLOC(y, N+ord, opus_val16); + for(i=0;i0); + celt_assert(overlap>=0); + if (overlap == 0) + { + xptr = x; + } else { + for (i=0;i0) + { + for(i=0;i= 536870912) + { + int shift2=1; + if (ac[0] >= 1073741824) + shift2++; + for (i=0;i<=lag;i++) + ac[i] = SHR32(ac[i], shift2); + shift += shift2; + } +#endif + + RESTORE_STACK; + return shift; +} diff --git a/code/opus-1.0.2/celt/celt_lpc.h b/code/opus-1.1/celt/celt_lpc.h similarity index 93% rename from code/opus-1.0.2/celt/celt_lpc.h rename to code/opus-1.1/celt/celt_lpc.h index 2baa77ed..dc2a0a3d 100644 --- a/code/opus-1.0.2/celt/celt_lpc.h +++ b/code/opus-1.1/celt/celt_lpc.h @@ -48,6 +48,7 @@ void celt_iir(const opus_val32 *x, int ord, opus_val16 *mem); -void _celt_autocorr(const opus_val16 *x, opus_val32 *ac, const opus_val16 *window, int overlap, int lag, int n); +int _celt_autocorr(const opus_val16 *x, opus_val32 *ac, + const opus_val16 *window, int overlap, int lag, int n, int arch); #endif /* PLC_H */ diff --git a/code/opus-1.0.2/celt/pitch.h b/code/opus-1.1/celt/cpu_support.h similarity index 68% rename from code/opus-1.0.2/celt/pitch.h rename to code/opus-1.1/celt/cpu_support.h index 2757071a..d68dbe62 100644 --- a/code/opus-1.0.2/celt/pitch.h +++ b/code/opus-1.1/celt/cpu_support.h @@ -1,11 +1,5 @@ -/* Copyright (c) 2007-2008 CSIRO - Copyright (c) 2007-2009 Xiph.Org Foundation - Written by Jean-Marc Valin */ -/** - @file pitch.h - @brief Pitch analysis - */ - +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ /* Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions @@ -31,18 +25,30 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ -#ifndef PITCH_H -#define PITCH_H +#ifndef CPU_SUPPORT_H +#define CPU_SUPPORT_H -#include "modes.h" +#include "opus_types.h" +#include "opus_defines.h" -void pitch_downsample(celt_sig * OPUS_RESTRICT x[], opus_val16 * OPUS_RESTRICT x_lp, - int len, int C); +#if defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_ASM) +#include "arm/armcpu.h" -void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTRICT y, - int len, int max_pitch, int *pitch); +/* We currently support 4 ARM variants: + * arch[0] -> ARMv4 + * arch[1] -> ARMv5E + * arch[2] -> ARMv6 + * arch[3] -> NEON + */ +#define OPUS_ARCHMASK 3 -opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod, - int N, int *T0, int prev_period, opus_val16 prev_gain); +#else +#define OPUS_ARCHMASK 0 + +static OPUS_INLINE int opus_select_arch(void) +{ + return 0; +} +#endif #endif diff --git a/code/opus-1.1/celt/cwrs.c b/code/opus-1.1/celt/cwrs.c new file mode 100644 index 00000000..ad980cc7 --- /dev/null +++ b/code/opus-1.1/celt/cwrs.c @@ -0,0 +1,697 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2007-2009 Timothy B. Terriberry + Written by Timothy B. Terriberry and Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "os_support.h" +#include "cwrs.h" +#include "mathops.h" +#include "arch.h" + +#ifdef CUSTOM_MODES + +/*Guaranteed to return a conservatively large estimate of the binary logarithm + with frac bits of fractional precision. + Tested for all possible 32-bit inputs with frac=4, where the maximum + overestimation is 0.06254243 bits.*/ +int log2_frac(opus_uint32 val, int frac) +{ + int l; + l=EC_ILOG(val); + if(val&(val-1)){ + /*This is (val>>l-16), but guaranteed to round up, even if adding a bias + before the shift would cause overflow (e.g., for 0xFFFFxxxx). + Doesn't work for val=0, but that case fails the test above.*/ + if(l>16)val=((val-1)>>(l-16))+1; + else val<<=16-l; + l=(l-1)<>16); + l+=b<>b; + val=(val*val+0x7FFF)>>15; + } + while(frac-->0); + /*If val is not exactly 0x8000, then we have to round up the remainder.*/ + return l+(val>0x8000); + } + /*Exact powers of two require no rounding.*/ + else return (l-1)<0 ? sum(k=1...K,2**k*choose(N,k)*choose(K-1,k-1)) : 1, + where choose() is the binomial function. + A table of values for N<10 and K<10 looks like: + V[10][10] = { + {1, 0, 0, 0, 0, 0, 0, 0, 0, 0}, + {1, 2, 2, 2, 2, 2, 2, 2, 2, 2}, + {1, 4, 8, 12, 16, 20, 24, 28, 32, 36}, + {1, 6, 18, 38, 66, 102, 146, 198, 258, 326}, + {1, 8, 32, 88, 192, 360, 608, 952, 1408, 1992}, + {1, 10, 50, 170, 450, 1002, 1970, 3530, 5890, 9290}, + {1, 12, 72, 292, 912, 2364, 5336, 10836, 20256, 35436}, + {1, 14, 98, 462, 1666, 4942, 12642, 28814, 59906, 115598}, + {1, 16, 128, 688, 2816, 9424, 27008, 68464, 157184, 332688}, + {1, 18, 162, 978, 4482, 16722, 53154, 148626, 374274, 864146} + }; + + U(N,K) = the number of such combinations wherein N-1 objects are taken at + most K-1 at a time. + This is given by + U(N,K) = sum(k=0...K-1,V(N-1,k)) + = K>0 ? (V(N-1,K-1) + V(N,K-1))/2 : 0. + The latter expression also makes clear that U(N,K) is half the number of such + combinations wherein the first object is taken at least once. + Although it may not be clear from either of these definitions, U(N,K) is the + natural function to work with when enumerating the pulse vector codebooks, + not V(N,K). + U(N,K) is not well-defined for N=0, but with the extension + U(0,K) = K>0 ? 0 : 1, + the function becomes symmetric: U(N,K) = U(K,N), with a similar table: + U[10][10] = { + {1, 0, 0, 0, 0, 0, 0, 0, 0, 0}, + {0, 1, 1, 1, 1, 1, 1, 1, 1, 1}, + {0, 1, 3, 5, 7, 9, 11, 13, 15, 17}, + {0, 1, 5, 13, 25, 41, 61, 85, 113, 145}, + {0, 1, 7, 25, 63, 129, 231, 377, 575, 833}, + {0, 1, 9, 41, 129, 321, 681, 1289, 2241, 3649}, + {0, 1, 11, 61, 231, 681, 1683, 3653, 7183, 13073}, + {0, 1, 13, 85, 377, 1289, 3653, 8989, 19825, 40081}, + {0, 1, 15, 113, 575, 2241, 7183, 19825, 48639, 108545}, + {0, 1, 17, 145, 833, 3649, 13073, 40081, 108545, 265729} + }; + + With this extension, V(N,K) may be written in terms of U(N,K): + V(N,K) = U(N,K) + U(N,K+1) + for all N>=0, K>=0. + Thus U(N,K+1) represents the number of combinations where the first element + is positive or zero, and U(N,K) represents the number of combinations where + it is negative. + With a large enough table of U(N,K) values, we could write O(N) encoding + and O(min(N*log(K),N+K)) decoding routines, but such a table would be + prohibitively large for small embedded devices (K may be as large as 32767 + for small N, and N may be as large as 200). + + Both functions obey the same recurrence relation: + V(N,K) = V(N-1,K) + V(N,K-1) + V(N-1,K-1), + U(N,K) = U(N-1,K) + U(N,K-1) + U(N-1,K-1), + for all N>0, K>0, with different initial conditions at N=0 or K=0. + This allows us to construct a row of one of the tables above given the + previous row or the next row. + Thus we can derive O(NK) encoding and decoding routines with O(K) memory + using only addition and subtraction. + + When encoding, we build up from the U(2,K) row and work our way forwards. + When decoding, we need to start at the U(N,K) row and work our way backwards, + which requires a means of computing U(N,K). + U(N,K) may be computed from two previous values with the same N: + U(N,K) = ((2*N-1)*U(N,K-1) - U(N,K-2))/(K-1) + U(N,K-2) + for all N>1, and since U(N,K) is symmetric, a similar relation holds for two + previous values with the same K: + U(N,K>1) = ((2*K-1)*U(N-1,K) - U(N-2,K))/(N-1) + U(N-2,K) + for all K>1. + This allows us to construct an arbitrary row of the U(N,K) table by starting + with the first two values, which are constants. + This saves roughly 2/3 the work in our O(NK) decoding routine, but costs O(K) + multiplications. + Similar relations can be derived for V(N,K), but are not used here. + + For N>0 and K>0, U(N,K) and V(N,K) take on the form of an (N-1)-degree + polynomial for fixed N. + The first few are + U(1,K) = 1, + U(2,K) = 2*K-1, + U(3,K) = (2*K-2)*K+1, + U(4,K) = (((4*K-6)*K+8)*K-3)/3, + U(5,K) = ((((2*K-4)*K+10)*K-8)*K+3)/3, + and + V(1,K) = 2, + V(2,K) = 4*K, + V(3,K) = 4*K*K+2, + V(4,K) = 8*(K*K+2)*K/3, + V(5,K) = ((4*K*K+20)*K*K+6)/3, + for all K>0. + This allows us to derive O(N) encoding and O(N*log(K)) decoding routines for + small N (and indeed decoding is also O(N) for N<3). + + @ARTICLE{Fis86, + author="Thomas R. Fischer", + title="A Pyramid Vector Quantizer", + journal="IEEE Transactions on Information Theory", + volume="IT-32", + number=4, + pages="568--583", + month=Jul, + year=1986 + }*/ + +#if !defined(SMALL_FOOTPRINT) + +/*U(N,K) = U(K,N) := N>0?K>0?U(N-1,K)+U(N,K-1)+U(N-1,K-1):0:K>0?1:0*/ +# define CELT_PVQ_U(_n,_k) (CELT_PVQ_U_ROW[IMIN(_n,_k)][IMAX(_n,_k)]) +/*V(N,K) := U(N,K)+U(N,K+1) = the number of PVQ codewords for a band of size N + with K pulses allocated to it.*/ +# define CELT_PVQ_V(_n,_k) (CELT_PVQ_U(_n,_k)+CELT_PVQ_U(_n,(_k)+1)) + +/*For each V(N,K) supported, we will access element U(min(N,K+1),max(N,K+1)). + Thus, the number of entries in row I is the larger of the maximum number of + pulses we will ever allocate for a given N=I (K=128, or however many fit in + 32 bits, whichever is smaller), plus one, and the maximum N for which + K=I-1 pulses fit in 32 bits. + The largest band size in an Opus Custom mode is 208. + Otherwise, we can limit things to the set of N which can be achieved by + splitting a band from a standard Opus mode: 176, 144, 96, 88, 72, 64, 48, + 44, 36, 32, 24, 22, 18, 16, 8, 4, 2).*/ +#if defined(CUSTOM_MODES) +static const opus_uint32 CELT_PVQ_U_DATA[1488]={ +#else +static const opus_uint32 CELT_PVQ_U_DATA[1272]={ +#endif + /*N=0, K=0...176:*/ + 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +#if defined(CUSTOM_MODES) + /*...208:*/ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, +#endif + /*N=1, K=1...176:*/ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, +#if defined(CUSTOM_MODES) + /*...208:*/ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, +#endif + /*N=2, K=2...176:*/ + 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, + 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65, 67, 69, 71, 73, 75, 77, 79, + 81, 83, 85, 87, 89, 91, 93, 95, 97, 99, 101, 103, 105, 107, 109, 111, 113, + 115, 117, 119, 121, 123, 125, 127, 129, 131, 133, 135, 137, 139, 141, 143, + 145, 147, 149, 151, 153, 155, 157, 159, 161, 163, 165, 167, 169, 171, 173, + 175, 177, 179, 181, 183, 185, 187, 189, 191, 193, 195, 197, 199, 201, 203, + 205, 207, 209, 211, 213, 215, 217, 219, 221, 223, 225, 227, 229, 231, 233, + 235, 237, 239, 241, 243, 245, 247, 249, 251, 253, 255, 257, 259, 261, 263, + 265, 267, 269, 271, 273, 275, 277, 279, 281, 283, 285, 287, 289, 291, 293, + 295, 297, 299, 301, 303, 305, 307, 309, 311, 313, 315, 317, 319, 321, 323, + 325, 327, 329, 331, 333, 335, 337, 339, 341, 343, 345, 347, 349, 351, +#if defined(CUSTOM_MODES) + /*...208:*/ + 353, 355, 357, 359, 361, 363, 365, 367, 369, 371, 373, 375, 377, 379, 381, + 383, 385, 387, 389, 391, 393, 395, 397, 399, 401, 403, 405, 407, 409, 411, + 413, 415, +#endif + /*N=3, K=3...176:*/ + 13, 25, 41, 61, 85, 113, 145, 181, 221, 265, 313, 365, 421, 481, 545, 613, + 685, 761, 841, 925, 1013, 1105, 1201, 1301, 1405, 1513, 1625, 1741, 1861, + 1985, 2113, 2245, 2381, 2521, 2665, 2813, 2965, 3121, 3281, 3445, 3613, 3785, + 3961, 4141, 4325, 4513, 4705, 4901, 5101, 5305, 5513, 5725, 5941, 6161, 6385, + 6613, 6845, 7081, 7321, 7565, 7813, 8065, 8321, 8581, 8845, 9113, 9385, 9661, + 9941, 10225, 10513, 10805, 11101, 11401, 11705, 12013, 12325, 12641, 12961, + 13285, 13613, 13945, 14281, 14621, 14965, 15313, 15665, 16021, 16381, 16745, + 17113, 17485, 17861, 18241, 18625, 19013, 19405, 19801, 20201, 20605, 21013, + 21425, 21841, 22261, 22685, 23113, 23545, 23981, 24421, 24865, 25313, 25765, + 26221, 26681, 27145, 27613, 28085, 28561, 29041, 29525, 30013, 30505, 31001, + 31501, 32005, 32513, 33025, 33541, 34061, 34585, 35113, 35645, 36181, 36721, + 37265, 37813, 38365, 38921, 39481, 40045, 40613, 41185, 41761, 42341, 42925, + 43513, 44105, 44701, 45301, 45905, 46513, 47125, 47741, 48361, 48985, 49613, + 50245, 50881, 51521, 52165, 52813, 53465, 54121, 54781, 55445, 56113, 56785, + 57461, 58141, 58825, 59513, 60205, 60901, 61601, +#if defined(CUSTOM_MODES) + /*...208:*/ + 62305, 63013, 63725, 64441, 65161, 65885, 66613, 67345, 68081, 68821, 69565, + 70313, 71065, 71821, 72581, 73345, 74113, 74885, 75661, 76441, 77225, 78013, + 78805, 79601, 80401, 81205, 82013, 82825, 83641, 84461, 85285, 86113, +#endif + /*N=4, K=4...176:*/ + 63, 129, 231, 377, 575, 833, 1159, 1561, 2047, 2625, 3303, 4089, 4991, 6017, + 7175, 8473, 9919, 11521, 13287, 15225, 17343, 19649, 22151, 24857, 27775, + 30913, 34279, 37881, 41727, 45825, 50183, 54809, 59711, 64897, 70375, 76153, + 82239, 88641, 95367, 102425, 109823, 117569, 125671, 134137, 142975, 152193, + 161799, 171801, 182207, 193025, 204263, 215929, 228031, 240577, 253575, + 267033, 280959, 295361, 310247, 325625, 341503, 357889, 374791, 392217, + 410175, 428673, 447719, 467321, 487487, 508225, 529543, 551449, 573951, + 597057, 620775, 645113, 670079, 695681, 721927, 748825, 776383, 804609, + 833511, 863097, 893375, 924353, 956039, 988441, 1021567, 1055425, 1090023, + 1125369, 1161471, 1198337, 1235975, 1274393, 1313599, 1353601, 1394407, + 1436025, 1478463, 1521729, 1565831, 1610777, 1656575, 1703233, 1750759, + 1799161, 1848447, 1898625, 1949703, 2001689, 2054591, 2108417, 2163175, + 2218873, 2275519, 2333121, 2391687, 2451225, 2511743, 2573249, 2635751, + 2699257, 2763775, 2829313, 2895879, 2963481, 3032127, 3101825, 3172583, + 3244409, 3317311, 3391297, 3466375, 3542553, 3619839, 3698241, 3777767, + 3858425, 3940223, 4023169, 4107271, 4192537, 4278975, 4366593, 4455399, + 4545401, 4636607, 4729025, 4822663, 4917529, 5013631, 5110977, 5209575, + 5309433, 5410559, 5512961, 5616647, 5721625, 5827903, 5935489, 6044391, + 6154617, 6266175, 6379073, 6493319, 6608921, 6725887, 6844225, 6963943, + 7085049, 7207551, +#if defined(CUSTOM_MODES) + /*...208:*/ + 7331457, 7456775, 7583513, 7711679, 7841281, 7972327, 8104825, 8238783, + 8374209, 8511111, 8649497, 8789375, 8930753, 9073639, 9218041, 9363967, + 9511425, 9660423, 9810969, 9963071, 10116737, 10271975, 10428793, 10587199, + 10747201, 10908807, 11072025, 11236863, 11403329, 11571431, 11741177, + 11912575, +#endif + /*N=5, K=5...176:*/ + 321, 681, 1289, 2241, 3649, 5641, 8361, 11969, 16641, 22569, 29961, 39041, + 50049, 63241, 78889, 97281, 118721, 143529, 172041, 204609, 241601, 283401, + 330409, 383041, 441729, 506921, 579081, 658689, 746241, 842249, 947241, + 1061761, 1186369, 1321641, 1468169, 1626561, 1797441, 1981449, 2179241, + 2391489, 2618881, 2862121, 3121929, 3399041, 3694209, 4008201, 4341801, + 4695809, 5071041, 5468329, 5888521, 6332481, 6801089, 7295241, 7815849, + 8363841, 8940161, 9545769, 10181641, 10848769, 11548161, 12280841, 13047849, + 13850241, 14689089, 15565481, 16480521, 17435329, 18431041, 19468809, + 20549801, 21675201, 22846209, 24064041, 25329929, 26645121, 28010881, + 29428489, 30899241, 32424449, 34005441, 35643561, 37340169, 39096641, + 40914369, 42794761, 44739241, 46749249, 48826241, 50971689, 53187081, + 55473921, 57833729, 60268041, 62778409, 65366401, 68033601, 70781609, + 73612041, 76526529, 79526721, 82614281, 85790889, 89058241, 92418049, + 95872041, 99421961, 103069569, 106816641, 110664969, 114616361, 118672641, + 122835649, 127107241, 131489289, 135983681, 140592321, 145317129, 150160041, + 155123009, 160208001, 165417001, 170752009, 176215041, 181808129, 187533321, + 193392681, 199388289, 205522241, 211796649, 218213641, 224775361, 231483969, + 238341641, 245350569, 252512961, 259831041, 267307049, 274943241, 282741889, + 290705281, 298835721, 307135529, 315607041, 324252609, 333074601, 342075401, + 351257409, 360623041, 370174729, 379914921, 389846081, 399970689, 410291241, + 420810249, 431530241, 442453761, 453583369, 464921641, 476471169, 488234561, + 500214441, 512413449, 524834241, 537479489, 550351881, 563454121, 576788929, + 590359041, 604167209, 618216201, 632508801, +#if defined(CUSTOM_MODES) + /*...208:*/ + 647047809, 661836041, 676876329, 692171521, 707724481, 723538089, 739615241, + 755958849, 772571841, 789457161, 806617769, 824056641, 841776769, 859781161, + 878072841, 896654849, 915530241, 934702089, 954173481, 973947521, 994027329, + 1014416041, 1035116809, 1056132801, 1077467201, 1099123209, 1121104041, + 1143412929, 1166053121, 1189027881, 1212340489, 1235994241, +#endif + /*N=6, K=6...96:*/ + 1683, 3653, 7183, 13073, 22363, 36365, 56695, 85305, 124515, 177045, 246047, + 335137, 448427, 590557, 766727, 982729, 1244979, 1560549, 1937199, 2383409, + 2908411, 3522221, 4235671, 5060441, 6009091, 7095093, 8332863, 9737793, + 11326283, 13115773, 15124775, 17372905, 19880915, 22670725, 25765455, + 29189457, 32968347, 37129037, 41699767, 46710137, 52191139, 58175189, + 64696159, 71789409, 79491819, 87841821, 96879431, 106646281, 117185651, + 128542501, 140763503, 153897073, 167993403, 183104493, 199284183, 216588185, + 235074115, 254801525, 275831935, 298228865, 322057867, 347386557, 374284647, + 402823977, 433078547, 465124549, 499040399, 534906769, 572806619, 612825229, + 655050231, 699571641, 746481891, 795875861, 847850911, 902506913, 959946283, + 1020274013, 1083597703, 1150027593, 1219676595, 1292660325, 1369097135, + 1449108145, 1532817275, 1620351277, 1711839767, 1807415257, 1907213187, + 2011371957, 2120032959, +#if defined(CUSTOM_MODES) + /*...109:*/ + 2233340609U, 2351442379U, 2474488829U, 2602633639U, 2736033641U, 2874848851U, + 3019242501U, 3169381071U, 3325434321U, 3487575323U, 3655980493U, 3830829623U, + 4012305913U, +#endif + /*N=7, K=7...54*/ + 8989, 19825, 40081, 75517, 134245, 227305, 369305, 579125, 880685, 1303777, + 1884961, 2668525, 3707509, 5064793, 6814249, 9041957, 11847485, 15345233, + 19665841, 24957661, 31388293, 39146185, 48442297, 59511829, 72616013, + 88043969, 106114625, 127178701, 151620757, 179861305, 212358985, 249612805, + 292164445, 340600625, 395555537, 457713341, 527810725, 606639529, 695049433, + 793950709, 904317037, 1027188385, 1163673953, 1314955181, 1482288821, + 1667010073, 1870535785, 2094367717, +#if defined(CUSTOM_MODES) + /*...60:*/ + 2340095869U, 2609401873U, 2904062449U, 3225952925U, 3577050821U, 3959439497U, +#endif + /*N=8, K=8...37*/ + 48639, 108545, 224143, 433905, 795455, 1392065, 2340495, 3800305, 5984767, + 9173505, 13726991, 20103025, 28875327, 40754369, 56610575, 77500017, + 104692735, 139703809, 184327311, 240673265, 311207743, 398796225, 506750351, + 638878193, 799538175, 993696769, 1226990095, 1505789553, 1837271615, + 2229491905U, +#if defined(CUSTOM_MODES) + /*...40:*/ + 2691463695U, 3233240945U, 3866006015U, +#endif + /*N=9, K=9...28:*/ + 265729, 598417, 1256465, 2485825, 4673345, 8405905, 14546705, 24331777, + 39490049, 62390545, 96220561, 145198913, 214828609, 312193553, 446304145, + 628496897, 872893441, 1196924561, 1621925137, 2173806145U, +#if defined(CUSTOM_MODES) + /*...29:*/ + 2883810113U, +#endif + /*N=10, K=10...24:*/ + 1462563, 3317445, 7059735, 14218905, 27298155, 50250765, 89129247, 152951073, + 254831667, 413442773, 654862247, 1014889769, 1541911931, 2300409629U, + 3375210671U, + /*N=11, K=11...19:*/ + 8097453, 18474633, 39753273, 81270333, 158819253, 298199265, 540279585, + 948062325, 1616336765, +#if defined(CUSTOM_MODES) + /*...20:*/ + 2684641785U, +#endif + /*N=12, K=12...18:*/ + 45046719, 103274625, 224298231, 464387817, 921406335, 1759885185, + 3248227095U, + /*N=13, K=13...16:*/ + 251595969, 579168825, 1267854873, 2653649025U, + /*N=14, K=14:*/ + 1409933619 +}; + +#if defined(CUSTOM_MODES) +static const opus_uint32 *const CELT_PVQ_U_ROW[15]={ + CELT_PVQ_U_DATA+ 0,CELT_PVQ_U_DATA+ 208,CELT_PVQ_U_DATA+ 415, + CELT_PVQ_U_DATA+ 621,CELT_PVQ_U_DATA+ 826,CELT_PVQ_U_DATA+1030, + CELT_PVQ_U_DATA+1233,CELT_PVQ_U_DATA+1336,CELT_PVQ_U_DATA+1389, + CELT_PVQ_U_DATA+1421,CELT_PVQ_U_DATA+1441,CELT_PVQ_U_DATA+1455, + CELT_PVQ_U_DATA+1464,CELT_PVQ_U_DATA+1470,CELT_PVQ_U_DATA+1473 +}; +#else +static const opus_uint32 *const CELT_PVQ_U_ROW[15]={ + CELT_PVQ_U_DATA+ 0,CELT_PVQ_U_DATA+ 176,CELT_PVQ_U_DATA+ 351, + CELT_PVQ_U_DATA+ 525,CELT_PVQ_U_DATA+ 698,CELT_PVQ_U_DATA+ 870, + CELT_PVQ_U_DATA+1041,CELT_PVQ_U_DATA+1131,CELT_PVQ_U_DATA+1178, + CELT_PVQ_U_DATA+1207,CELT_PVQ_U_DATA+1226,CELT_PVQ_U_DATA+1240, + CELT_PVQ_U_DATA+1248,CELT_PVQ_U_DATA+1254,CELT_PVQ_U_DATA+1257 +}; +#endif + +#if defined(CUSTOM_MODES) +void get_required_bits(opus_int16 *_bits,int _n,int _maxk,int _frac){ + int k; + /*_maxk==0 => there's nothing to do.*/ + celt_assert(_maxk>0); + _bits[0]=0; + for(k=1;k<=_maxk;k++)_bits[k]=log2_frac(CELT_PVQ_V(_n,k),_frac); +} +#endif + +static opus_uint32 icwrs(int _n,const int *_y){ + opus_uint32 i; + int j; + int k; + celt_assert(_n>=2); + j=_n-1; + i=_y[j]<0; + k=abs(_y[j]); + do{ + j--; + i+=CELT_PVQ_U(_n-j,k); + k+=abs(_y[j]); + if(_y[j]<0)i+=CELT_PVQ_U(_n-j,k+1); + } + while(j>0); + return i; +} + +void encode_pulses(const int *_y,int _n,int _k,ec_enc *_enc){ + celt_assert(_k>0); + ec_enc_uint(_enc,icwrs(_n,_y),CELT_PVQ_V(_n,_k)); +} + +static void cwrsi(int _n,int _k,opus_uint32 _i,int *_y){ + opus_uint32 p; + int s; + int k0; + celt_assert(_k>0); + celt_assert(_n>1); + while(_n>2){ + opus_uint32 q; + /*Lots of pulses case:*/ + if(_k>=_n){ + const opus_uint32 *row; + row=CELT_PVQ_U_ROW[_n]; + /*Are the pulses in this dimension negative?*/ + p=row[_k+1]; + s=-(_i>=p); + _i-=p&s; + /*Count how many pulses were placed in this dimension.*/ + k0=_k; + q=row[_n]; + if(q>_i){ + celt_assert(p>q); + _k=_n; + do p=CELT_PVQ_U_ROW[--_k][_n]; + while(p>_i); + } + else for(p=row[_k];p>_i;p=row[_k])_k--; + _i-=p; + *_y++=(k0-_k+s)^s; + } + /*Lots of dimensions case:*/ + else{ + /*Are there any pulses in this dimension at all?*/ + p=CELT_PVQ_U_ROW[_k][_n]; + q=CELT_PVQ_U_ROW[_k+1][_n]; + if(p<=_i&&_i=q); + _i-=q&s; + /*Count how many pulses were placed in this dimension.*/ + k0=_k; + do p=CELT_PVQ_U_ROW[--_k][_n]; + while(p>_i); + _i-=p; + *_y++=(k0-_k+s)^s; + } + } + _n--; + } + /*_n==2*/ + p=2*_k+1; + s=-(_i>=p); + _i-=p&s; + k0=_k; + _k=(_i+1)>>1; + if(_k)_i-=2*_k-1; + *_y++=(k0-_k+s)^s; + /*_n==1*/ + s=-(int)_i; + *_y=(_k+s)^s; +} + +void decode_pulses(int *_y,int _n,int _k,ec_dec *_dec){ + cwrsi(_n,_k,ec_dec_uint(_dec,CELT_PVQ_V(_n,_k)),_y); +} + +#else /* SMALL_FOOTPRINT */ + +/*Computes the next row/column of any recurrence that obeys the relation + u[i][j]=u[i-1][j]+u[i][j-1]+u[i-1][j-1]. + _ui0 is the base case for the new row/column.*/ +static OPUS_INLINE void unext(opus_uint32 *_ui,unsigned _len,opus_uint32 _ui0){ + opus_uint32 ui1; + unsigned j; + /*This do-while will overrun the array if we don't have storage for at least + 2 values.*/ + j=1; do { + ui1=UADD32(UADD32(_ui[j],_ui[j-1]),_ui0); + _ui[j-1]=_ui0; + _ui0=ui1; + } while (++j<_len); + _ui[j-1]=_ui0; +} + +/*Computes the previous row/column of any recurrence that obeys the relation + u[i-1][j]=u[i][j]-u[i][j-1]-u[i-1][j-1]. + _ui0 is the base case for the new row/column.*/ +static OPUS_INLINE void uprev(opus_uint32 *_ui,unsigned _n,opus_uint32 _ui0){ + opus_uint32 ui1; + unsigned j; + /*This do-while will overrun the array if we don't have storage for at least + 2 values.*/ + j=1; do { + ui1=USUB32(USUB32(_ui[j],_ui[j-1]),_ui0); + _ui[j-1]=_ui0; + _ui0=ui1; + } while (++j<_n); + _ui[j-1]=_ui0; +} + +/*Compute V(_n,_k), as well as U(_n,0..._k+1). + _u: On exit, _u[i] contains U(_n,i) for i in [0..._k+1].*/ +static opus_uint32 ncwrs_urow(unsigned _n,unsigned _k,opus_uint32 *_u){ + opus_uint32 um2; + unsigned len; + unsigned k; + len=_k+2; + /*We require storage at least 3 values (e.g., _k>0).*/ + celt_assert(len>=3); + _u[0]=0; + _u[1]=um2=1; + /*If _n==0, _u[0] should be 1 and the rest should be 0.*/ + /*If _n==1, _u[i] should be 1 for i>1.*/ + celt_assert(_n>=2); + /*If _k==0, the following do-while loop will overflow the buffer.*/ + celt_assert(_k>0); + k=2; + do _u[k]=(k<<1)-1; + while(++k0); + j=0; + do{ + opus_uint32 p; + int s; + int yj; + p=_u[_k+1]; + s=-(_i>=p); + _i-=p&s; + yj=_k; + p=_u[_k]; + while(p>_i)p=_u[--_k]; + _i-=p; + yj-=_k; + _y[j]=(yj+s)^s; + uprev(_u,_k+2,0); + } + while(++j<_n); +} + +/*Returns the index of the given combination of K elements chosen from a set + of size 1 with associated sign bits. + _y: The vector of pulses, whose sum of absolute values is K. + _k: Returns K.*/ +static OPUS_INLINE opus_uint32 icwrs1(const int *_y,int *_k){ + *_k=abs(_y[0]); + return _y[0]<0; +} + +/*Returns the index of the given combination of K elements chosen from a set + of size _n with associated sign bits. + _y: The vector of pulses, whose sum of absolute values must be _k. + _nc: Returns V(_n,_k).*/ +static OPUS_INLINE opus_uint32 icwrs(int _n,int _k,opus_uint32 *_nc,const int *_y, + opus_uint32 *_u){ + opus_uint32 i; + int j; + int k; + /*We can't unroll the first two iterations of the loop unless _n>=2.*/ + celt_assert(_n>=2); + _u[0]=0; + for(k=1;k<=_k+1;k++)_u[k]=(k<<1)-1; + i=icwrs1(_y+_n-1,&k); + j=_n-2; + i+=_u[k]; + k+=abs(_y[j]); + if(_y[j]<0)i+=_u[k+1]; + while(j-->0){ + unext(_u,_k+2,0); + i+=_u[k]; + k+=abs(_y[j]); + if(_y[j]<0)i+=_u[k+1]; + } + *_nc=_u[k]+_u[k+1]; + return i; +} + +#ifdef CUSTOM_MODES +void get_required_bits(opus_int16 *_bits,int _n,int _maxk,int _frac){ + int k; + /*_maxk==0 => there's nothing to do.*/ + celt_assert(_maxk>0); + _bits[0]=0; + if (_n==1) + { + for (k=1;k<=_maxk;k++) + _bits[k] = 1<<_frac; + } + else { + VARDECL(opus_uint32,u); + SAVE_STACK; + ALLOC(u,_maxk+2U,opus_uint32); + ncwrs_urow(_n,_maxk,u); + for(k=1;k<=_maxk;k++) + _bits[k]=log2_frac(u[k]+u[k+1],_frac); + RESTORE_STACK; + } +} +#endif /* CUSTOM_MODES */ + +void encode_pulses(const int *_y,int _n,int _k,ec_enc *_enc){ + opus_uint32 i; + VARDECL(opus_uint32,u); + opus_uint32 nc; + SAVE_STACK; + celt_assert(_k>0); + ALLOC(u,_k+2U,opus_uint32); + i=icwrs(_n,_k,&nc,_y,u); + ec_enc_uint(_enc,i,nc); + RESTORE_STACK; +} + +void decode_pulses(int *_y,int _n,int _k,ec_dec *_dec){ + VARDECL(opus_uint32,u); + SAVE_STACK; + celt_assert(_k>0); + ALLOC(u,_k+2U,opus_uint32); + cwrsi(_n,_k,ec_dec_uint(_dec,ncwrs_urow(_n,_k,u)),_y,u); + RESTORE_STACK; +} + +#endif /* SMALL_FOOTPRINT */ diff --git a/code/opus-1.0.2/celt/cwrs.h b/code/opus-1.1/celt/cwrs.h similarity index 100% rename from code/opus-1.0.2/celt/cwrs.h rename to code/opus-1.1/celt/cwrs.h diff --git a/code/opus-1.0.2/celt/ecintrin.h b/code/opus-1.1/celt/ecintrin.h similarity index 97% rename from code/opus-1.0.2/celt/ecintrin.h rename to code/opus-1.1/celt/ecintrin.h index be57dd40..2263cff6 100644 --- a/code/opus-1.0.2/celt/ecintrin.h +++ b/code/opus-1.1/celt/ecintrin.h @@ -33,7 +33,7 @@ #if !defined(_ecintrin_H) # define _ecintrin_H (1) -/*Some specific platforms may have optimized intrinsic or inline assembly +/*Some specific platforms may have optimized intrinsic or OPUS_INLINE assembly versions of these functions which can substantially improve performance. We define macros for them to allow easy incorporation of these non-ANSI features.*/ diff --git a/code/opus-1.0.2/celt/entcode.c b/code/opus-1.1/celt/entcode.c similarity index 100% rename from code/opus-1.0.2/celt/entcode.c rename to code/opus-1.1/celt/entcode.c diff --git a/code/opus-1.0.2/celt/entcode.h b/code/opus-1.1/celt/entcode.h similarity index 94% rename from code/opus-1.0.2/celt/entcode.h rename to code/opus-1.1/celt/entcode.h index aebecc06..dd13e49e 100644 --- a/code/opus-1.0.2/celt/entcode.h +++ b/code/opus-1.1/celt/entcode.h @@ -26,6 +26,7 @@ */ #include "opus_types.h" +#include "opus_defines.h" #if !defined(_entcode_H) # define _entcode_H (1) @@ -83,15 +84,15 @@ struct ec_ctx{ int error; }; -static inline opus_uint32 ec_range_bytes(ec_ctx *_this){ +static OPUS_INLINE opus_uint32 ec_range_bytes(ec_ctx *_this){ return _this->offs; } -static inline unsigned char *ec_get_buffer(ec_ctx *_this){ +static OPUS_INLINE unsigned char *ec_get_buffer(ec_ctx *_this){ return _this->buf; } -static inline int ec_get_error(ec_ctx *_this){ +static OPUS_INLINE int ec_get_error(ec_ctx *_this){ return _this->error; } @@ -101,7 +102,7 @@ static inline int ec_get_error(ec_ctx *_this){ Return: The number of bits. This will always be slightly larger than the exact value (e.g., all rounding error is in the positive direction).*/ -static inline int ec_tell(ec_ctx *_this){ +static OPUS_INLINE int ec_tell(ec_ctx *_this){ return _this->nbits_total-EC_ILOG(_this->rng); } diff --git a/code/opus-1.0.2/celt/entdec.c b/code/opus-1.1/celt/entdec.c similarity index 98% rename from code/opus-1.0.2/celt/entdec.c rename to code/opus-1.1/celt/entdec.c index 75e3e45a..3c264685 100644 --- a/code/opus-1.0.2/celt/entdec.c +++ b/code/opus-1.1/celt/entdec.c @@ -85,7 +85,7 @@ number=3, pages="256--294", month=Jul, - URL="http://www.stanford.edu/class/ee398/handouts/papers/Moffat98ArithmCoding.pdf" + URL="http://www.stanford.edu/class/ee398a/handouts/papers/Moffat98ArithmCoding.pdf" }*/ static int ec_read_byte(ec_dec *_this){ diff --git a/code/opus-1.0.2/celt/entdec.h b/code/opus-1.1/celt/entdec.h similarity index 100% rename from code/opus-1.0.2/celt/entdec.h rename to code/opus-1.1/celt/entdec.h diff --git a/code/opus-1.0.2/celt/entenc.c b/code/opus-1.1/celt/entenc.c similarity index 100% rename from code/opus-1.0.2/celt/entenc.c rename to code/opus-1.1/celt/entenc.c diff --git a/code/opus-1.0.2/celt/entenc.h b/code/opus-1.1/celt/entenc.h similarity index 100% rename from code/opus-1.0.2/celt/entenc.h rename to code/opus-1.1/celt/entenc.h diff --git a/code/opus-1.0.2/celt/fixed_debug.h b/code/opus-1.1/celt/fixed_debug.h similarity index 90% rename from code/opus-1.0.2/celt/fixed_debug.h rename to code/opus-1.1/celt/fixed_debug.h index f11d890d..80bc9491 100644 --- a/code/opus-1.0.2/celt/fixed_debug.h +++ b/code/opus-1.1/celt/fixed_debug.h @@ -33,9 +33,9 @@ #define FIXED_DEBUG_H #include +#include "opus_defines.h" #ifdef CELT_C -#include "opus_defines.h" OPUS_EXPORT opus_int64 celt_mips=0; #else extern opus_int64 celt_mips; @@ -59,7 +59,7 @@ extern opus_int64 celt_mips; #define SHR(a,b) SHR32(a,b) #define PSHR(a,b) PSHR32(a,b) -static inline short NEG16(int x) +static OPUS_INLINE short NEG16(int x) { int res; if (!VERIFY_SHORT(x)) @@ -80,7 +80,7 @@ static inline short NEG16(int x) celt_mips++; return res; } -static inline int NEG32(opus_int64 x) +static OPUS_INLINE int NEG32(opus_int64 x) { opus_int64 res; if (!VERIFY_INT(x)) @@ -103,7 +103,7 @@ static inline int NEG32(opus_int64 x) } #define EXTRACT16(x) EXTRACT16_(x, __FILE__, __LINE__) -static inline short EXTRACT16_(int x, char *file, int line) +static OPUS_INLINE short EXTRACT16_(int x, char *file, int line) { int res; if (!VERIFY_SHORT(x)) @@ -119,7 +119,7 @@ static inline short EXTRACT16_(int x, char *file, int line) } #define EXTEND32(x) EXTEND32_(x, __FILE__, __LINE__) -static inline int EXTEND32_(int x, char *file, int line) +static OPUS_INLINE int EXTEND32_(int x, char *file, int line) { int res; if (!VERIFY_SHORT(x)) @@ -135,7 +135,7 @@ static inline int EXTEND32_(int x, char *file, int line) } #define SHR16(a, shift) SHR16_(a, shift, __FILE__, __LINE__) -static inline short SHR16_(int a, int shift, char *file, int line) +static OPUS_INLINE short SHR16_(int a, int shift, char *file, int line) { int res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(shift)) @@ -157,7 +157,7 @@ static inline short SHR16_(int a, int shift, char *file, int line) return res; } #define SHL16(a, shift) SHL16_(a, shift, __FILE__, __LINE__) -static inline short SHL16_(int a, int shift, char *file, int line) +static OPUS_INLINE short SHL16_(int a, int shift, char *file, int line) { int res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(shift)) @@ -179,7 +179,7 @@ static inline short SHL16_(int a, int shift, char *file, int line) return res; } -static inline int SHR32(opus_int64 a, int shift) +static OPUS_INLINE int SHR32(opus_int64 a, int shift) { opus_int64 res; if (!VERIFY_INT(a) || !VERIFY_SHORT(shift)) @@ -201,7 +201,7 @@ static inline int SHR32(opus_int64 a, int shift) return res; } #define SHL32(a, shift) SHL32_(a, shift, __FILE__, __LINE__) -static inline int SHL32_(opus_int64 a, int shift, char *file, int line) +static OPUS_INLINE int SHL32_(opus_int64 a, int shift, char *file, int line) { opus_int64 res; if (!VERIFY_INT(a) || !VERIFY_SHORT(shift)) @@ -234,7 +234,7 @@ static inline int SHL32_(opus_int64 a, int shift, char *file, int line) //#define SHL(a,shift) ((a) << (shift)) #define ADD16(a, b) ADD16_(a, b, __FILE__, __LINE__) -static inline short ADD16_(int a, int b, char *file, int line) +static OPUS_INLINE short ADD16_(int a, int b, char *file, int line) { int res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -257,7 +257,7 @@ static inline short ADD16_(int a, int b, char *file, int line) } #define SUB16(a, b) SUB16_(a, b, __FILE__, __LINE__) -static inline short SUB16_(int a, int b, char *file, int line) +static OPUS_INLINE short SUB16_(int a, int b, char *file, int line) { int res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -280,7 +280,7 @@ static inline short SUB16_(int a, int b, char *file, int line) } #define ADD32(a, b) ADD32_(a, b, __FILE__, __LINE__) -static inline int ADD32_(opus_int64 a, opus_int64 b, char *file, int line) +static OPUS_INLINE int ADD32_(opus_int64 a, opus_int64 b, char *file, int line) { opus_int64 res; if (!VERIFY_INT(a) || !VERIFY_INT(b)) @@ -303,7 +303,7 @@ static inline int ADD32_(opus_int64 a, opus_int64 b, char *file, int line) } #define SUB32(a, b) SUB32_(a, b, __FILE__, __LINE__) -static inline int SUB32_(opus_int64 a, opus_int64 b, char *file, int line) +static OPUS_INLINE int SUB32_(opus_int64 a, opus_int64 b, char *file, int line) { opus_int64 res; if (!VERIFY_INT(a) || !VERIFY_INT(b)) @@ -327,7 +327,7 @@ static inline int SUB32_(opus_int64 a, opus_int64 b, char *file, int line) #undef UADD32 #define UADD32(a, b) UADD32_(a, b, __FILE__, __LINE__) -static inline unsigned int UADD32_(opus_uint64 a, opus_uint64 b, char *file, int line) +static OPUS_INLINE unsigned int UADD32_(opus_uint64 a, opus_uint64 b, char *file, int line) { opus_uint64 res; if (!VERIFY_UINT(a) || !VERIFY_UINT(b)) @@ -351,7 +351,7 @@ static inline unsigned int UADD32_(opus_uint64 a, opus_uint64 b, char *file, int #undef USUB32 #define USUB32(a, b) USUB32_(a, b, __FILE__, __LINE__) -static inline unsigned int USUB32_(opus_uint64 a, opus_uint64 b, char *file, int line) +static OPUS_INLINE unsigned int USUB32_(opus_uint64 a, opus_uint64 b, char *file, int line) { opus_uint64 res; if (!VERIFY_UINT(a) || !VERIFY_UINT(b)) @@ -381,7 +381,7 @@ static inline unsigned int USUB32_(opus_uint64 a, opus_uint64 b, char *file, int } /* result fits in 16 bits */ -static inline short MULT16_16_16(int a, int b) +static OPUS_INLINE short MULT16_16_16(int a, int b) { int res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -404,7 +404,7 @@ static inline short MULT16_16_16(int a, int b) } #define MULT16_16(a, b) MULT16_16_(a, b, __FILE__, __LINE__) -static inline int MULT16_16_(int a, int b, char *file, int line) +static OPUS_INLINE int MULT16_16_(int a, int b, char *file, int line) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -429,7 +429,7 @@ static inline int MULT16_16_(int a, int b, char *file, int line) #define MAC16_16(c,a,b) (celt_mips-=2,ADD32((c),MULT16_16((a),(b)))) #define MULT16_32_QX(a, b, Q) MULT16_32_QX_(a, b, Q, __FILE__, __LINE__) -static inline int MULT16_32_QX_(int a, opus_int64 b, int Q, char *file, int line) +static OPUS_INLINE int MULT16_32_QX_(int a, opus_int64 b, int Q, char *file, int line) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_INT(b)) @@ -462,7 +462,7 @@ static inline int MULT16_32_QX_(int a, opus_int64 b, int Q, char *file, int line } #define MULT16_32_PX(a, b, Q) MULT16_32_PX_(a, b, Q, __FILE__, __LINE__) -static inline int MULT16_32_PX_(int a, opus_int64 b, int Q, char *file, int line) +static OPUS_INLINE int MULT16_32_PX_(int a, opus_int64 b, int Q, char *file, int line) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_INT(b)) @@ -497,7 +497,7 @@ static inline int MULT16_32_PX_(int a, opus_int64 b, int Q, char *file, int line #define MULT16_32_Q15(a,b) MULT16_32_QX(a,b,15) #define MAC16_32_Q15(c,a,b) (celt_mips-=2,ADD32((c),MULT16_32_Q15((a),(b)))) -static inline int SATURATE(int a, int b) +static OPUS_INLINE int SATURATE(int a, int b) { if (a>b) a=b; @@ -507,7 +507,17 @@ static inline int SATURATE(int a, int b) return a; } -static inline int MULT16_16_Q11_32(int a, int b) +static OPUS_INLINE opus_int16 SATURATE16(opus_int32 a) +{ + celt_mips+=3; + if (a>32767) + return 32767; + else if (a<-32768) + return -32768; + else return a; +} + +static OPUS_INLINE int MULT16_16_Q11_32(int a, int b) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -529,7 +539,7 @@ static inline int MULT16_16_Q11_32(int a, int b) celt_mips+=3; return res; } -static inline short MULT16_16_Q13(int a, int b) +static OPUS_INLINE short MULT16_16_Q13(int a, int b) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -551,7 +561,7 @@ static inline short MULT16_16_Q13(int a, int b) celt_mips+=3; return res; } -static inline short MULT16_16_Q14(int a, int b) +static OPUS_INLINE short MULT16_16_Q14(int a, int b) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -575,7 +585,7 @@ static inline short MULT16_16_Q14(int a, int b) } #define MULT16_16_Q15(a, b) MULT16_16_Q15_(a, b, __FILE__, __LINE__) -static inline short MULT16_16_Q15_(int a, int b, char *file, int line) +static OPUS_INLINE short MULT16_16_Q15_(int a, int b, char *file, int line) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -598,7 +608,7 @@ static inline short MULT16_16_Q15_(int a, int b, char *file, int line) return res; } -static inline short MULT16_16_P13(int a, int b) +static OPUS_INLINE short MULT16_16_P13(int a, int b) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -628,7 +638,7 @@ static inline short MULT16_16_P13(int a, int b) celt_mips+=4; return res; } -static inline short MULT16_16_P14(int a, int b) +static OPUS_INLINE short MULT16_16_P14(int a, int b) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -658,7 +668,7 @@ static inline short MULT16_16_P14(int a, int b) celt_mips+=4; return res; } -static inline short MULT16_16_P15(int a, int b) +static OPUS_INLINE short MULT16_16_P15(int a, int b) { opus_int64 res; if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) @@ -691,7 +701,7 @@ static inline short MULT16_16_P15(int a, int b) #define DIV32_16(a, b) DIV32_16_(a, b, __FILE__, __LINE__) -static inline int DIV32_16_(opus_int64 a, opus_int64 b, char *file, int line) +static OPUS_INLINE int DIV32_16_(opus_int64 a, opus_int64 b, char *file, int line) { opus_int64 res; if (b==0) @@ -726,7 +736,7 @@ static inline int DIV32_16_(opus_int64 a, opus_int64 b, char *file, int line) } #define DIV32(a, b) DIV32_(a, b, __FILE__, __LINE__) -static inline int DIV32_(opus_int64 a, opus_int64 b, char *file, int line) +static OPUS_INLINE int DIV32_(opus_int64 a, opus_int64 b, char *file, int line) { opus_int64 res; if (b==0) diff --git a/code/opus-1.0.2/celt/fixed_generic.h b/code/opus-1.1/celt/fixed_generic.h similarity index 95% rename from code/opus-1.0.2/celt/fixed_generic.h rename to code/opus-1.1/celt/fixed_generic.h index 71e28d62..ecf018a2 100644 --- a/code/opus-1.0.2/celt/fixed_generic.h +++ b/code/opus-1.1/celt/fixed_generic.h @@ -40,7 +40,7 @@ #define MULT16_32_Q16(a,b) ADD32(MULT16_16((a),SHR((b),16)), SHR(MULT16_16SU((a),((b)&0x0000ffff)),16)) /** 16x32 multiplication, followed by a 16-bit shift right (round-to-nearest). Results fits in 32 bits */ -#define MULT16_32_P16(a,b) ADD32(MULT16_16((a),SHR((b),16)), PSHR(MULT16_16((a),((b)&0x0000ffff)),16)) +#define MULT16_32_P16(a,b) ADD32(MULT16_16((a),SHR((b),16)), PSHR(MULT16_16SU((a),((b)&0x0000ffff)),16)) /** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */ #define MULT16_32_Q15(a,b) ADD32(SHL(MULT16_16((a),SHR((b),16)),1), SHR(MULT16_16SU((a),((b)&0x0000ffff)),15)) @@ -84,6 +84,8 @@ #define PSHR(a,shift) (SHR((a)+((EXTEND32(1)<<((shift))>>1)),shift)) #define SATURATE(x,a) (((x)>(a) ? (a) : (x)<-(a) ? -(a) : (x))) +#define SATURATE16(x) (EXTRACT16((x)>32767 ? 32767 : (x)<-32768 ? -32768 : (x))) + /** Shift by a and round-to-neareast 32-bit value. Result is a 16-bit value */ #define ROUND16(x,a) (EXTRACT16(PSHR32((x),(a)))) /** Divide by two */ @@ -108,10 +110,13 @@ /** 16x16 multiply-add where the result fits in 32 bits */ #define MAC16_16(c,a,b) (ADD32((c),MULT16_16((a),(b)))) -/** 16x32 multiply-add, followed by a 15-bit shift right. Results fits in 32 bits */ +/** 16x32 multiply, followed by a 15-bit shift right and 32-bit add. + b must fit in 31 bits. + Result fits in 32 bits. */ #define MAC16_32_Q15(c,a,b) ADD32(c,ADD32(MULT16_16((a),SHR((b),15)), SHR(MULT16_16((a),((b)&0x00007fff)),15))) #define MULT16_16_Q11_32(a,b) (SHR(MULT16_16((a),(b)),11)) +#define MULT16_16_Q11(a,b) (SHR(MULT16_16((a),(b)),11)) #define MULT16_16_Q13(a,b) (SHR(MULT16_16((a),(b)),13)) #define MULT16_16_Q14(a,b) (SHR(MULT16_16((a),(b)),14)) #define MULT16_16_Q15(a,b) (SHR(MULT16_16((a),(b)),15)) diff --git a/code/opus-1.0.2/celt/float_cast.h b/code/opus-1.1/celt/float_cast.h similarity index 97% rename from code/opus-1.0.2/celt/float_cast.h rename to code/opus-1.1/celt/float_cast.h index 5ded2915..ede65748 100644 --- a/code/opus-1.0.2/celt/float_cast.h +++ b/code/opus-1.1/celt/float_cast.h @@ -101,7 +101,7 @@ #include /* Win32 doesn't seem to have these functions. - ** Therefore implement inline versions of these functions here. + ** Therefore implement OPUS_INLINE versions of these functions here. */ __inline long int @@ -128,7 +128,7 @@ #endif #ifndef DISABLE_FLOAT_API -static inline opus_int16 FLOAT2INT16(float x) +static OPUS_INLINE opus_int16 FLOAT2INT16(float x) { x = x*CELT_SIG_SCALE; x = MAX32(x, -32768); diff --git a/code/opus-1.0.2/celt/kiss_fft.c b/code/opus-1.1/celt/kiss_fft.c similarity index 99% rename from code/opus-1.0.2/celt/kiss_fft.c rename to code/opus-1.1/celt/kiss_fft.c index dcd69686..ad706c73 100644 --- a/code/opus-1.0.2/celt/kiss_fft.c +++ b/code/opus-1.1/celt/kiss_fft.c @@ -40,7 +40,6 @@ #include "os_support.h" #include "mathops.h" #include "stack_alloc.h" -#include "os_support.h" /* The guts header contains all the multiplication and addition macros that are defined for complex numbers. It also delares the kf_ internal functions. @@ -142,8 +141,6 @@ static void kf_bfly4( C_ADDTO(*Fout, scratch[1]); C_ADD( scratch[3] , scratch[0] , scratch[2] ); C_SUB( scratch[4] , scratch[0] , scratch[2] ); - Fout[m2].r = PSHR32(Fout[m2].r, 2); - Fout[m2].i = PSHR32(Fout[m2].i, 2); C_SUB( Fout[m2], *Fout, scratch[3] ); tw1 += fstride; tw2 += fstride*2; diff --git a/code/opus-1.0.2/celt/kiss_fft.h b/code/opus-1.1/celt/kiss_fft.h similarity index 100% rename from code/opus-1.0.2/celt/kiss_fft.h rename to code/opus-1.1/celt/kiss_fft.h diff --git a/code/opus-1.0.2/celt/laplace.c b/code/opus-1.1/celt/laplace.c similarity index 100% rename from code/opus-1.0.2/celt/laplace.c rename to code/opus-1.1/celt/laplace.c diff --git a/code/opus-1.0.2/celt/laplace.h b/code/opus-1.1/celt/laplace.h similarity index 100% rename from code/opus-1.0.2/celt/laplace.h rename to code/opus-1.1/celt/laplace.h diff --git a/code/opus-1.0.2/celt/mathops.c b/code/opus-1.1/celt/mathops.c similarity index 98% rename from code/opus-1.0.2/celt/mathops.c rename to code/opus-1.1/celt/mathops.c index ce472c9f..3f8c5dcc 100644 --- a/code/opus-1.0.2/celt/mathops.c +++ b/code/opus-1.1/celt/mathops.c @@ -123,6 +123,8 @@ opus_val32 celt_sqrt(opus_val32 x) static const opus_val16 C[5] = {23175, 11561, -3011, 1699, -664}; if (x==0) return 0; + else if (x>=1073741824) + return 32767; k = (celt_ilog2(x)>>1)-7; x = VSHR32(x, 2*k); n = x-32768; @@ -137,7 +139,7 @@ opus_val32 celt_sqrt(opus_val32 x) #define L3 8277 #define L4 -626 -static inline opus_val16 _celt_cos_pi_2(opus_val16 x) +static OPUS_INLINE opus_val16 _celt_cos_pi_2(opus_val16 x) { opus_val16 x2; diff --git a/code/opus-1.0.2/celt/mathops.h b/code/opus-1.1/celt/mathops.h similarity index 82% rename from code/opus-1.0.2/celt/mathops.h rename to code/opus-1.1/celt/mathops.h index 4e977956..a0525a96 100644 --- a/code/opus-1.0.2/celt/mathops.h +++ b/code/opus-1.1/celt/mathops.h @@ -43,6 +43,41 @@ unsigned isqrt32(opus_uint32 _val); +#ifndef OVERRIDE_CELT_MAXABS16 +static OPUS_INLINE opus_val32 celt_maxabs16(const opus_val16 *x, int len) +{ + int i; + opus_val16 maxval = 0; + opus_val16 minval = 0; + for (i=0;i0, "celt_ilog2() only defined for strictly positive numbers"); return EC_ILOG(x)-1; } #endif -#ifndef OVERRIDE_CELT_MAXABS16 -static inline opus_val16 celt_maxabs16(opus_val16 *x, int len) -{ - int i; - opus_val16 maxval = 0; - for (i=0;in; N >>= shift; N2 = N>>1; N4 = N>>2; ALLOC(f, N2, kiss_fft_scalar); + ALLOC(f2, N2, kiss_fft_scalar); /* sin(x) ~= x here */ #ifdef FIXED_POINT sine = TRIG_UPSCALE*(QCONST16(0.7853981f, 15)+N2)/N; @@ -131,7 +133,7 @@ void clt_mdct_forward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar kiss_fft_scalar * OPUS_RESTRICT yp = f; const opus_val16 * OPUS_RESTRICT wp1 = window+(overlap>>1); const opus_val16 * OPUS_RESTRICT wp2 = window+(overlap>>1)-1; - for(i=0;i<(overlap>>2);i++) + for(i=0;i<((overlap+3)>>2);i++) { /* Real part arranged as -d-cR, Imag part arranged as -b+aR*/ *yp++ = MULT16_32_Q15(*wp2, xp1[N2]) + MULT16_32_Q15(*wp1,*xp2); @@ -143,7 +145,7 @@ void clt_mdct_forward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar } wp1 = window; wp2 = window+overlap-1; - for(;i>2);i++) + for(;i>2);i++) { /* Real part arranged as a-bR, Imag part arranged as -c-dR */ *yp++ = *xp2; @@ -180,12 +182,12 @@ void clt_mdct_forward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar } /* N/4 complex FFT, down-scales by 4/N */ - opus_fft(l->kfft[shift], (kiss_fft_cpx *)f, (kiss_fft_cpx *)in); + opus_fft(l->kfft[shift], (kiss_fft_cpx *)f, (kiss_fft_cpx *)f2); /* Post-rotate */ { /* Temp pointers to make it really clear to the compiler what we're doing */ - const kiss_fft_scalar * OPUS_RESTRICT fp = in; + const kiss_fft_scalar * OPUS_RESTRICT fp = f2; kiss_fft_scalar * OPUS_RESTRICT yp1 = out; kiss_fft_scalar * OPUS_RESTRICT yp2 = out+stride*(N2-1); const kiss_twiddle_scalar *t = &l->trig[0]; @@ -212,14 +214,12 @@ void clt_mdct_backward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scala int i; int N, N2, N4; kiss_twiddle_scalar sine; - VARDECL(kiss_fft_scalar, f); VARDECL(kiss_fft_scalar, f2); SAVE_STACK; N = l->n; N >>= shift; N2 = N>>1; N4 = N>>2; - ALLOC(f, N2, kiss_fft_scalar); ALLOC(f2, N2, kiss_fft_scalar); /* sin(x) ~= x here */ #ifdef FIXED_POINT @@ -249,81 +249,60 @@ void clt_mdct_backward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scala } /* Inverse N/4 complex FFT. This one should *not* downscale even in fixed-point */ - opus_ifft(l->kfft[shift], (kiss_fft_cpx *)f2, (kiss_fft_cpx *)f); + opus_ifft(l->kfft[shift], (kiss_fft_cpx *)f2, (kiss_fft_cpx *)(out+(overlap>>1))); - /* Post-rotate */ + /* Post-rotate and de-shuffle from both ends of the buffer at once to make + it in-place. */ { - kiss_fft_scalar * OPUS_RESTRICT fp = f; + kiss_fft_scalar * OPUS_RESTRICT yp0 = out+(overlap>>1); + kiss_fft_scalar * OPUS_RESTRICT yp1 = out+(overlap>>1)+N2-2; const kiss_twiddle_scalar *t = &l->trig[0]; - - for(i=0;i>1 to handle odd N4. When N4 is odd, the + middle pair will be computed twice. */ + for(i=0;i<(N4+1)>>1;i++) { kiss_fft_scalar re, im, yr, yi; - re = fp[0]; - im = fp[1]; + kiss_twiddle_scalar t0, t1; + re = yp0[0]; + im = yp0[1]; + t0 = t[i<>1; + /* Mirror on both sides for TDAC */ { - kiss_fft_scalar * OPUS_RESTRICT fp1 = f2+N4-1; - kiss_fft_scalar * OPUS_RESTRICT xp1 = out+N2-1; - kiss_fft_scalar * OPUS_RESTRICT yp1 = out+N4-overlap/2; + kiss_fft_scalar * OPUS_RESTRICT xp1 = out+overlap-1; + kiss_fft_scalar * OPUS_RESTRICT yp1 = out; const opus_val16 * OPUS_RESTRICT wp1 = window; const opus_val16 * OPUS_RESTRICT wp2 = window+overlap-1; - for(i = 0; i< N4-overlap/2; i++) + + for(i = 0; i < overlap/2; i++) { - *xp1 = *fp1; - xp1--; - fp1--; - } - for(; i < N4; i++) - { - kiss_fft_scalar x1; - x1 = *fp1--; - *yp1++ +=-MULT16_32_Q15(*wp1, x1); - *xp1-- += MULT16_32_Q15(*wp2, x1); - wp1++; - wp2--; - } - } - { - kiss_fft_scalar * OPUS_RESTRICT fp2 = f2+N4; - kiss_fft_scalar * OPUS_RESTRICT xp2 = out+N2; - kiss_fft_scalar * OPUS_RESTRICT yp2 = out+N-1-(N4-overlap/2); - const opus_val16 * OPUS_RESTRICT wp1 = window; - const opus_val16 * OPUS_RESTRICT wp2 = window+overlap-1; - for(i = 0; i< N4-overlap/2; i++) - { - *xp2 = *fp2; - xp2++; - fp2++; - } - for(; i < N4; i++) - { - kiss_fft_scalar x2; - x2 = *fp2++; - *yp2-- = MULT16_32_Q15(*wp1, x2); - *xp2++ = MULT16_32_Q15(*wp2, x2); + kiss_fft_scalar x1, x2; + x1 = *xp1; + x2 = *yp1; + *yp1++ = MULT16_32_Q15(*wp2, x2) - MULT16_32_Q15(*wp1, x1); + *xp1-- = MULT16_32_Q15(*wp1, x2) + MULT16_32_Q15(*wp2, x1); wp1++; wp2--; } diff --git a/code/opus-1.0.2/celt/mdct.h b/code/opus-1.1/celt/mdct.h similarity index 100% rename from code/opus-1.0.2/celt/mdct.h rename to code/opus-1.1/celt/mdct.h diff --git a/code/opus-1.0.2/celt/mfrngcod.h b/code/opus-1.1/celt/mfrngcod.h similarity index 100% rename from code/opus-1.0.2/celt/mfrngcod.h rename to code/opus-1.1/celt/mfrngcod.h diff --git a/code/opus-1.0.2/celt/modes.c b/code/opus-1.1/celt/modes.c similarity index 98% rename from code/opus-1.0.2/celt/modes.c rename to code/opus-1.1/celt/modes.c index ed204d7d..42e68e1c 100644 --- a/code/opus-1.0.2/celt/modes.c +++ b/code/opus-1.1/celt/modes.c @@ -345,6 +345,14 @@ CELTMode *opus_custom_mode_create(opus_int32 Fs, int frame_size, int *error) mode->eBands = compute_ebands(Fs, mode->shortMdctSize, res, &mode->nbEBands); if (mode->eBands==NULL) goto failure; +#if !defined(SMALL_FOOTPRINT) + /* Make sure we don't allocate a band larger than our PVQ table. + 208 should be enough, but let's be paranoid. */ + if ((mode->eBands[mode->nbEBands] - mode->eBands[mode->nbEBands-1])< + 208) { + goto failure; + } +#endif mode->effEBands = mode->nbEBands; while (mode->eBands[mode->effEBands] > mode->shortMdctSize) diff --git a/code/opus-1.0.2/celt/modes.h b/code/opus-1.1/celt/modes.h similarity index 100% rename from code/opus-1.0.2/celt/modes.h rename to code/opus-1.1/celt/modes.h diff --git a/code/opus-1.0.2/celt/opus_custom_demo.c b/code/opus-1.1/celt/opus_custom_demo.c similarity index 100% rename from code/opus-1.0.2/celt/opus_custom_demo.c rename to code/opus-1.1/celt/opus_custom_demo.c diff --git a/code/opus-1.0.2/celt/os_support.h b/code/opus-1.1/celt/os_support.h similarity index 93% rename from code/opus-1.0.2/celt/os_support.h rename to code/opus-1.1/celt/os_support.h index 2484f0b2..5e47e3cf 100644 --- a/code/opus-1.0.2/celt/os_support.h +++ b/code/opus-1.1/celt/os_support.h @@ -35,13 +35,16 @@ # include "custom_support.h" #endif +#include "opus_types.h" +#include "opus_defines.h" + #include #include #include /** Opus wrapper for malloc(). To do your own dynamic allocation, all you need to do is replace this function and opus_free */ #ifndef OVERRIDE_OPUS_ALLOC -static inline void *opus_alloc (size_t size) +static OPUS_INLINE void *opus_alloc (size_t size) { return malloc(size); } @@ -49,7 +52,7 @@ static inline void *opus_alloc (size_t size) /** Same as celt_alloc(), except that the area is only needed inside a CELT call (might cause problem with wideband though) */ #ifndef OVERRIDE_OPUS_ALLOC_SCRATCH -static inline void *opus_alloc_scratch (size_t size) +static OPUS_INLINE void *opus_alloc_scratch (size_t size) { /* Scratch space doesn't need to be cleared */ return opus_alloc(size); @@ -58,7 +61,7 @@ static inline void *opus_alloc_scratch (size_t size) /** Opus wrapper for free(). To do your own dynamic allocation, all you need to do is replace this function and opus_alloc */ #ifndef OVERRIDE_OPUS_FREE -static inline void opus_free (void *ptr) +static OPUS_INLINE void opus_free (void *ptr) { free(ptr); } diff --git a/code/opus-1.0.2/celt/pitch.c b/code/opus-1.1/celt/pitch.c similarity index 69% rename from code/opus-1.0.2/celt/pitch.c rename to code/opus-1.1/celt/pitch.c index ca0f523e..d2b30544 100644 --- a/code/opus-1.0.2/celt/pitch.c +++ b/code/opus-1.1/celt/pitch.c @@ -102,13 +102,57 @@ static void find_best_pitch(opus_val32 *xcorr, opus_val16 *y, int len, } } +static void celt_fir5(const opus_val16 *x, + const opus_val16 *num, + opus_val16 *y, + int N, + opus_val16 *mem) +{ + int i; + opus_val16 num0, num1, num2, num3, num4; + opus_val32 mem0, mem1, mem2, mem3, mem4; + num0=num[0]; + num1=num[1]; + num2=num[2]; + num3=num[3]; + num4=num[4]; + mem0=mem[0]; + mem1=mem[1]; + mem2=mem[2]; + mem3=mem[3]; + mem4=mem[4]; + for (i=0;i>1); + 4, len>>1, arch); /* Noise floor -40 dB */ #ifdef FIXED_POINT @@ -161,16 +205,96 @@ void pitch_downsample(celt_sig * OPUS_RESTRICT x[], opus_val16 * OPUS_RESTRICT x tmp = MULT16_16_Q15(QCONST16(.9f,15), tmp); lpc[i] = MULT16_16_Q15(lpc[i], tmp); } - celt_fir(x_lp, lpc, x_lp, len>>1, 4, mem); - - mem[0]=0; - lpc[0]=QCONST16(.8f,12); - celt_fir(x_lp, lpc, x_lp, len>>1, 1, mem); - + /* Add a zero */ + lpc2[0] = lpc[0] + QCONST16(.8f,SIG_SHIFT); + lpc2[1] = lpc[1] + MULT16_16_Q15(c1,lpc[0]); + lpc2[2] = lpc[2] + MULT16_16_Q15(c1,lpc[1]); + lpc2[3] = lpc[3] + MULT16_16_Q15(c1,lpc[2]); + lpc2[4] = MULT16_16_Q15(c1,lpc[3]); + celt_fir5(x_lp, lpc2, x_lp, len>>1, mem); } +#if 0 /* This is a simple version of the pitch correlation that should work + well on DSPs like Blackfin and TI C5x/C6x */ + +#ifdef FIXED_POINT +opus_val32 +#else +void +#endif +celt_pitch_xcorr(opus_val16 *x, opus_val16 *y, opus_val32 *xcorr, int len, int max_pitch) +{ + int i, j; +#ifdef FIXED_POINT + opus_val32 maxcorr=1; +#endif + for (i=0;i0); + celt_assert((((unsigned char *)_x-(unsigned char *)NULL)&3)==0); +#ifdef FIXED_POINT + opus_val32 maxcorr=1; +#endif + for (i=0;i>2); ymax = celt_maxabs16(y_lp4, lag>>2); - shift = celt_ilog2(MAX16(1, MAX16(xmax, ymax)))-11; + shift = celt_ilog2(MAX32(1, MAX32(xmax, ymax)))-11; if (shift>0) { for (j=0;j>2;j++) @@ -220,16 +344,11 @@ void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTR /* Coarse search with 4x decimation */ - for (i=0;i>2;i++) - { - opus_val32 sum = 0; - for (j=0;j>2;j++) - sum = MAC16_16(sum, x_lp4[j],y_lp4[i+j]); - xcorr[i] = MAX32(-1, sum); #ifdef FIXED_POINT - maxcorr = MAX32(maxcorr, sum); + maxcorr = #endif - } + celt_pitch_xcorr(x_lp4, y_lp4, xcorr, len>>2, max_pitch>>2, arch); + find_best_pitch(xcorr, y_lp4, len>>2, max_pitch>>2, best_pitch #ifdef FIXED_POINT , 0, maxcorr @@ -287,11 +406,13 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod, int k, i, T, T0; opus_val16 g, g0; opus_val16 pg; - opus_val32 xy,xx,yy; + opus_val32 xy,xx,yy,xy2; opus_val32 xcorr[3]; opus_val32 best_xy, best_yy; int offset; int minperiod0; + VARDECL(opus_val32, yy_lookup); + SAVE_STACK; minperiod0 = minperiod; maxperiod /= 2; @@ -304,13 +425,16 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod, *T0_=maxperiod-1; T = T0 = *T0_; - xx=xy=yy=0; - for (i=0;i QCONST16(.3f,15) + MULT16_16_Q15(QCONST16(.4f,15),g0)-cont) + thresh = MAX16(QCONST16(.3f,15), MULT16_16_Q15(QCONST16(.7f,15),g0)-cont); + /* Bias against very high pitch (very short period) to avoid false-positives + due to short-term correlation */ + if (T1<3*minperiod) + thresh = MAX16(QCONST16(.4f,15), MULT16_16_Q15(QCONST16(.85f,15),g0)-cont); + else if (T1<2*minperiod) + thresh = MAX16(QCONST16(.5f,15), MULT16_16_Q15(QCONST16(.9f,15),g0)-cont); + if (g1 > thresh) { best_xy = xy; best_yy = yy; @@ -406,5 +532,6 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod, if (*T0_=3); + y_3=0; /* gcc doesn't realize that y_3 can't be used uninitialized */ + y_0=*y++; + y_1=*y++; + y_2=*y++; + for (j=0;j=2) + qi = IMIN(qi, 0); if (budget-tell >= 15) { int pi; @@ -253,13 +255,13 @@ static int quant_coarse_energy_impl(const CELTMode *m, int start, int end, prev[c] = prev[c] + SHL32(q,7) - MULT16_16(beta,PSHR32(q,8)); } while (++c < C); } - return badness; + return lfe ? 0 : badness; } void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, const opus_val16 *eBands, opus_val16 *oldEBands, opus_uint32 budget, opus_val16 *error, ec_enc *enc, int C, int LM, int nbAvailableBytes, - int force_intra, opus_val32 *delayedIntra, int two_pass, int loss_rate) + int force_intra, opus_val32 *delayedIntra, int two_pass, int loss_rate, int lfe) { int intra; opus_val16 max_decay; @@ -280,9 +282,6 @@ void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, if (tell+3 > budget) two_pass = intra = 0; - /* Encode the global flags using a simple probability model - (first symbols in the stream) */ - max_decay = QCONST16(16.f,DB_SHIFT); if (end-start>10) { @@ -292,6 +291,8 @@ void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, max_decay = MIN32(max_decay, .125f*nbAvailableBytes); #endif } + if (lfe) + max_decay=3; enc_start_state = *enc; ALLOC(oldEBands_intra, C*m->nbEBands, opus_val16); @@ -301,7 +302,7 @@ void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, if (two_pass || intra) { badness1 = quant_coarse_energy_impl(m, start, end, eBands, oldEBands_intra, budget, - tell, e_prob_model[LM][1], error_intra, enc, C, LM, 1, max_decay); + tell, e_prob_model[LM][1], error_intra, enc, C, LM, 1, max_decay, lfe); } if (!intra) @@ -311,6 +312,7 @@ void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, opus_int32 tell_intra; opus_uint32 nstart_bytes; opus_uint32 nintra_bytes; + opus_uint32 save_bytes; int badness2; VARDECL(unsigned char, intra_bits); @@ -321,14 +323,17 @@ void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, nstart_bytes = ec_range_bytes(&enc_start_state); nintra_bytes = ec_range_bytes(&enc_intra_state); intra_buf = ec_get_buffer(&enc_intra_state) + nstart_bytes; - ALLOC(intra_bits, nintra_bytes-nstart_bytes, unsigned char); + save_bytes = nintra_bytes-nstart_bytes; + if (save_bytes == 0) + save_bytes = ALLOC_NONE; + ALLOC(intra_bits, save_bytes, unsigned char); /* Copy bits from intra bit-stream */ OPUS_COPY(intra_bits, intra_buf, nintra_bytes - nstart_bytes); *enc = enc_start_state; badness2 = quant_coarse_energy_impl(m, start, end, eBands, oldEBands, budget, - tell, e_prob_model[LM][intra], error, enc, C, LM, 0, max_decay); + tell, e_prob_model[LM][intra], error, enc, C, LM, 0, max_decay, lfe); if (two_pass && (badness1 < badness2 || (badness1 == badness2 && ((opus_int32)ec_tell_frac(enc))+intra_bias > tell_intra))) { @@ -535,25 +540,6 @@ void unquant_energy_finalise(const CELTMode *m, int start, int end, opus_val16 * } } -void log2Amp(const CELTMode *m, int start, int end, - celt_ener *eBands, const opus_val16 *oldEBands, int C) -{ - int c, i; - c=0; - do { - for (i=0;inbEBands] = 0; - for (;inbEBands], - SHL16((opus_val16)eMeans[i],6)); - eBands[i+c*m->nbEBands] = PSHR32(celt_exp2(lg),4); - } - for (;inbEBands;i++) - eBands[i+c*m->nbEBands] = 0; - } while (++c < C); -} - void amp2Log2(const CELTMode *m, int effEnd, int end, celt_ener *bandE, opus_val16 *bandLogE, int C) { diff --git a/code/opus-1.0.2/celt/quant_bands.h b/code/opus-1.1/celt/quant_bands.h similarity index 94% rename from code/opus-1.0.2/celt/quant_bands.h rename to code/opus-1.1/celt/quant_bands.h index bec2855c..0490bca4 100644 --- a/code/opus-1.0.2/celt/quant_bands.h +++ b/code/opus-1.1/celt/quant_bands.h @@ -35,6 +35,12 @@ #include "entdec.h" #include "mathops.h" +#ifdef FIXED_POINT +extern const signed char eMeans[25]; +#else +extern const opus_val16 eMeans[25]; +#endif + void amp2Log2(const CELTMode *m, int effEnd, int end, celt_ener *bandE, opus_val16 *bandLogE, int C); @@ -45,7 +51,7 @@ void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, const opus_val16 *eBands, opus_val16 *oldEBands, opus_uint32 budget, opus_val16 *error, ec_enc *enc, int C, int LM, int nbAvailableBytes, int force_intra, opus_val32 *delayedIntra, - int two_pass, int loss_rate); + int two_pass, int loss_rate, int lfe); void quant_fine_energy(const CELTMode *m, int start, int end, opus_val16 *oldEBands, opus_val16 *error, int *fine_quant, ec_enc *enc, int C); diff --git a/code/opus-1.0.2/celt/rate.c b/code/opus-1.1/celt/rate.c similarity index 98% rename from code/opus-1.0.2/celt/rate.c rename to code/opus-1.1/celt/rate.c index 4e96787f..e13d839d 100644 --- a/code/opus-1.0.2/celt/rate.c +++ b/code/opus-1.1/celt/rate.c @@ -245,10 +245,10 @@ void compute_pulse_cache(CELTMode *m, int LM) #define ALLOC_STEPS 6 -static inline int interp_bits2pulses(const CELTMode *m, int start, int end, int skip_start, +static OPUS_INLINE int interp_bits2pulses(const CELTMode *m, int start, int end, int skip_start, const int *bits1, const int *bits2, const int *thresh, const int *cap, opus_int32 total, opus_int32 *_balance, int skip_rsv, int *intensity, int intensity_rsv, int *dual_stereo, int dual_stereo_rsv, int *bits, - int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev) + int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev, int signalBandwidth) { opus_int32 psum; int lo, hi; @@ -353,7 +353,7 @@ static inline int interp_bits2pulses(const CELTMode *m, int start, int end, int #ifdef FUZZING if ((rand()&0x1) == 0) #else - if (codedBands<=start+2 || band_bits > ((j>4) + if (codedBands<=start+2 || (band_bits > ((j>4 && j<=signalBandwidth)) #endif { ec_enc_bit_logp(ec, 1, 1); @@ -524,7 +524,7 @@ static inline int interp_bits2pulses(const CELTMode *m, int start, int end, int } int compute_allocation(const CELTMode *m, int start, int end, const int *offsets, const int *cap, int alloc_trim, int *intensity, int *dual_stereo, - opus_int32 total, opus_int32 *balance, int *pulses, int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev) + opus_int32 total, opus_int32 *balance, int *pulses, int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev, int signalBandwidth) { int lo, hi, len, j; int codedBands; @@ -631,7 +631,7 @@ int compute_allocation(const CELTMode *m, int start, int end, const int *offsets } codedBands = interp_bits2pulses(m, start, end, skip_start, bits1, bits2, thresh, cap, total, balance, skip_rsv, intensity, intensity_rsv, dual_stereo, dual_stereo_rsv, - pulses, ebits, fine_priority, C, LM, ec, encode, prev); + pulses, ebits, fine_priority, C, LM, ec, encode, prev, signalBandwidth); RESTORE_STACK; return codedBands; } diff --git a/code/opus-1.0.2/celt/rate.h b/code/opus-1.1/celt/rate.h similarity index 92% rename from code/opus-1.0.2/celt/rate.h rename to code/opus-1.1/celt/rate.h index e0d50223..f1e06611 100644 --- a/code/opus-1.0.2/celt/rate.h +++ b/code/opus-1.1/celt/rate.h @@ -45,12 +45,12 @@ void compute_pulse_cache(CELTMode *m, int LM); -static inline int get_pulses(int i) +static OPUS_INLINE int get_pulses(int i) { return i<8 ? i : (8 + (i&7)) << ((i>>3)-1); } -static inline int bits2pulses(const CELTMode *m, int band, int LM, int bits) +static OPUS_INLINE int bits2pulses(const CELTMode *m, int band, int LM, int bits) { int i; int lo, hi; @@ -77,7 +77,7 @@ static inline int bits2pulses(const CELTMode *m, int band, int LM, int bits) return hi; } -static inline int pulses2bits(const CELTMode *m, int band, int LM, int pulses) +static OPUS_INLINE int pulses2bits(const CELTMode *m, int band, int LM, int pulses) { const unsigned char *cache; @@ -96,6 +96,6 @@ static inline int pulses2bits(const CELTMode *m, int band, int LM, int pulses) @return Total number of bits allocated */ int compute_allocation(const CELTMode *m, int start, int end, const int *offsets, const int *cap, int alloc_trim, int *intensity, int *dual_stero, - opus_int32 total, opus_int32 *balance, int *pulses, int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev); + opus_int32 total, opus_int32 *balance, int *pulses, int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev, int signalBandwidth); #endif diff --git a/code/opus-1.0.2/celt/stack_alloc.h b/code/opus-1.1/celt/stack_alloc.h similarity index 83% rename from code/opus-1.0.2/celt/stack_alloc.h rename to code/opus-1.1/celt/stack_alloc.h index a6f06d22..316a6ce1 100644 --- a/code/opus-1.0.2/celt/stack_alloc.h +++ b/code/opus-1.1/celt/stack_alloc.h @@ -32,6 +32,9 @@ #ifndef STACK_ALLOC_H #define STACK_ALLOC_H +#include "opus_types.h" +#include "opus_defines.h" + #if (!defined (VAR_ARRAYS) && !defined (USE_ALLOCA) && !defined (NONTHREADSAFE_PSEUDOSTACK)) #error "Opus requires one of VAR_ARRAYS, USE_ALLOCA, or NONTHREADSAFE_PSEUDOSTACK be defined to select the temporary allocation mode." #endif @@ -92,6 +95,8 @@ #define SAVE_STACK #define RESTORE_STACK #define ALLOC_STACK +/* C99 does not allow VLAs of size zero */ +#define ALLOC_NONE 1 #elif defined(USE_ALLOCA) @@ -106,6 +111,7 @@ #define SAVE_STACK #define RESTORE_STACK #define ALLOC_STACK +#define ALLOC_NONE 0 #else @@ -143,7 +149,30 @@ extern char *global_stack_top; #define VARDECL(type, var) type *var #define ALLOC(var, size, type) var = PUSH(global_stack, size, type) #define SAVE_STACK char *_saved_stack = global_stack; +#define ALLOC_NONE 0 #endif /* VAR_ARRAYS */ + +#ifdef ENABLE_VALGRIND + +#include +#define OPUS_CHECK_ARRAY(ptr, len) VALGRIND_CHECK_MEM_IS_DEFINED(ptr, len*sizeof(*ptr)) +#define OPUS_CHECK_VALUE(value) VALGRIND_CHECK_VALUE_IS_DEFINED(value) +#define OPUS_CHECK_ARRAY_COND(ptr, len) VALGRIND_CHECK_MEM_IS_DEFINED(ptr, len*sizeof(*ptr)) +#define OPUS_CHECK_VALUE_COND(value) VALGRIND_CHECK_VALUE_IS_DEFINED(value) +#define OPUS_PRINT_INT(value) do {fprintf(stderr, #value " = %d at %s:%d\n", value, __FILE__, __LINE__);}while(0) +#define OPUS_FPRINTF fprintf + +#else + +static OPUS_INLINE int _opus_false(void) {return 0;} +#define OPUS_CHECK_ARRAY(ptr, len) _opus_false() +#define OPUS_CHECK_VALUE(value) _opus_false() +#define OPUS_PRINT_INT(value) do{}while(0) +#define OPUS_FPRINTF (void) + +#endif + + #endif /* STACK_ALLOC_H */ diff --git a/code/opus-1.0.2/celt/static_modes_fixed.h b/code/opus-1.1/celt/static_modes_fixed.h similarity index 100% rename from code/opus-1.0.2/celt/static_modes_fixed.h rename to code/opus-1.1/celt/static_modes_fixed.h diff --git a/code/opus-1.0.2/celt/static_modes_float.h b/code/opus-1.1/celt/static_modes_float.h similarity index 100% rename from code/opus-1.0.2/celt/static_modes_float.h rename to code/opus-1.1/celt/static_modes_float.h diff --git a/code/opus-1.0.2/celt/vq.c b/code/opus-1.1/celt/vq.c similarity index 100% rename from code/opus-1.0.2/celt/vq.c rename to code/opus-1.1/celt/vq.c diff --git a/code/opus-1.0.2/celt/vq.h b/code/opus-1.1/celt/vq.h similarity index 89% rename from code/opus-1.0.2/celt/vq.h rename to code/opus-1.1/celt/vq.h index 1ceeeeb2..ffdc69cd 100644 --- a/code/opus-1.0.2/celt/vq.h +++ b/code/opus-1.1/celt/vq.h @@ -40,11 +40,9 @@ /** Algebraic pulse-vector quantiser. The signal x is replaced by the sum of * the pitch and a combination of pulses such that its norm is still equal * to 1. This is the function that will typically require the most CPU. - * @param x Residual signal to quantise/encode (returns quantised version) - * @param W Perceptual weight to use when optimising (currently unused) + * @param X Residual signal to quantise/encode (returns quantised version) * @param N Number of samples to encode * @param K Number of pulses to use - * @param p Pitch vector (it is assumed that p+x is a unit vector) * @param enc Entropy encoder state * @ret A mask indicating which blocks in the band received pulses */ @@ -56,10 +54,9 @@ unsigned alg_quant(celt_norm *X, int N, int K, int spread, int B, ); /** Algebraic pulse decoder - * @param x Decoded normalised spectrum (returned) + * @param X Decoded normalised spectrum (returned) * @param N Number of samples to decode * @param K Number of pulses to use - * @param p Pitch vector (automatically added to x) * @param dec Entropy decoder state * @ret A mask indicating which blocks in the band received pulses */ diff --git a/code/opus-1.1/celt/x86/pitch_sse.h b/code/opus-1.1/celt/x86/pitch_sse.h new file mode 100644 index 00000000..695122a5 --- /dev/null +++ b/code/opus-1.1/celt/x86/pitch_sse.h @@ -0,0 +1,156 @@ +/* Copyright (c) 2013 Jean-Marc Valin and John Ridges */ +/** + @file pitch_sse.h + @brief Pitch analysis + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef PITCH_SSE_H +#define PITCH_SSE_H + +#include +#include "arch.h" + +#define OVERRIDE_XCORR_KERNEL +static OPUS_INLINE void xcorr_kernel(const opus_val16 *x, const opus_val16 *y, opus_val32 sum[4], int len) +{ + int j; + __m128 xsum1, xsum2; + xsum1 = _mm_loadu_ps(sum); + xsum2 = _mm_setzero_ps(); + + for (j = 0; j < len-3; j += 4) + { + __m128 x0 = _mm_loadu_ps(x+j); + __m128 yj = _mm_loadu_ps(y+j); + __m128 y3 = _mm_loadu_ps(y+j+3); + + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0x00),yj)); + xsum2 = _mm_add_ps(xsum2,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0x55), + _mm_shuffle_ps(yj,y3,0x49))); + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0xaa), + _mm_shuffle_ps(yj,y3,0x9e))); + xsum2 = _mm_add_ps(xsum2,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0xff),y3)); + } + if (j < len) + { + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_load1_ps(x+j),_mm_loadu_ps(y+j))); + if (++j < len) + { + xsum2 = _mm_add_ps(xsum2,_mm_mul_ps(_mm_load1_ps(x+j),_mm_loadu_ps(y+j))); + if (++j < len) + { + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_load1_ps(x+j),_mm_loadu_ps(y+j))); + } + } + } + _mm_storeu_ps(sum,_mm_add_ps(xsum1,xsum2)); +} + +#define OVERRIDE_DUAL_INNER_PROD +static OPUS_INLINE void dual_inner_prod(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02, + int N, opus_val32 *xy1, opus_val32 *xy2) +{ + int i; + __m128 xsum1, xsum2; + xsum1 = _mm_setzero_ps(); + xsum2 = _mm_setzero_ps(); + for (i=0;iopus_int32: size of data * @param [out] out_toc char*: TOC pointer * @param [out] frames char*[48] encapsulated frames - * @param [out] size short[48] sizes of the encapsulated frames + * @param [out] size opus_int16[48] sizes of the encapsulated frames * @param [out] payload_offset int*: returns the position of the payload within the packet (in bytes) * @returns number of frames */ @@ -529,7 +529,7 @@ OPUS_EXPORT int opus_packet_parse( opus_int32 len, unsigned char *out_toc, const unsigned char *frames[48], - short size[48], + opus_int16 size[48], int *payload_offset ) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); @@ -566,6 +566,7 @@ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_channels(const unsign * @param [in] packet char*: Opus packet * @param [in] len opus_int32: Length of packet * @returns Number of frames + * @retval OPUS_BAD_ARG Insufficient data was passed to the function * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type */ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_frames(const unsigned char packet[], opus_int32 len) OPUS_ARG_NONNULL(1); @@ -577,6 +578,7 @@ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_frames(const unsigned * This must be a multiple of 400, or * inaccurate results will be returned. * @returns Number of samples + * @retval OPUS_BAD_ARG Insufficient data was passed to the function * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type */ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_samples(const unsigned char packet[], opus_int32 len, opus_int32 Fs) OPUS_ARG_NONNULL(1); @@ -586,9 +588,24 @@ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_samples(const unsigne * @param [in] packet char*: Opus packet * @param [in] len opus_int32: Length of packet * @returns Number of samples + * @retval OPUS_BAD_ARG Insufficient data was passed to the function * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type */ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_decoder_get_nb_samples(const OpusDecoder *dec, const unsigned char packet[], opus_int32 len) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); + +/** Applies soft-clipping to bring a float signal within the [-1,1] range. If + * the signal is already in that range, nothing is done. If there are values + * outside of [-1,1], then the signal is clipped as smoothly as possible to + * both fit in the range and avoid creating excessive distortion in the + * process. + * @param [in,out] pcm float*: Input PCM and modified PCM + * @param [in] frame_size int Number of samples per channel to process + * @param [in] channels int: Number of channels + * @param [in,out] softclip_mem float*: State memory for the soft clipping process (one float per channel, initialized to zero) + */ +OPUS_EXPORT void opus_pcm_soft_clip(float *pcm, int frame_size, int channels, float *softclip_mem); + + /**@}*/ /** @defgroup opus_repacketizer Repacketizer @@ -894,6 +911,64 @@ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_repacketizer_get_nb_frames(OpusRepa */ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_repacketizer_out(OpusRepacketizer *rp, unsigned char *data, opus_int32 maxlen) OPUS_ARG_NONNULL(1); +/** Pads a given Opus packet to a larger size (possibly changing the TOC sequence). + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to pad. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @param new_len opus_int32: The desired size of the packet after padding. + * This must be at least as large as len. + * @returns an error code + * @retval #OPUS_OK \a on success. + * @retval #OPUS_BAD_ARG \a len was less than 1 or new_len was less than len. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT int opus_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len); + +/** Remove all padding from a given Opus packet and rewrite the TOC sequence to + * minimize space usage. + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to strip. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @returns The new size of the output packet on success, or an error code + * on failure. + * @retval #OPUS_BAD_ARG \a len was less than 1. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_packet_unpad(unsigned char *data, opus_int32 len); + +/** Pads a given Opus multi-stream packet to a larger size (possibly changing the TOC sequence). + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to pad. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @param new_len opus_int32: The desired size of the packet after padding. + * This must be at least 1. + * @param nb_streams opus_int32: The number of streams (not channels) in the packet. + * This must be at least as large as len. + * @returns an error code + * @retval #OPUS_OK \a on success. + * @retval #OPUS_BAD_ARG \a len was less than 1. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT int opus_multistream_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len, int nb_streams); + +/** Remove all padding from a given Opus multi-stream packet and rewrite the TOC sequence to + * minimize space usage. + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to strip. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @param nb_streams opus_int32: The number of streams (not channels) in the packet. + * This must be at least 1. + * @returns The new size of the output packet on success, or an error code + * on failure. + * @retval #OPUS_BAD_ARG \a len was less than 1 or new_len was less than len. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_multistream_packet_unpad(unsigned char *data, opus_int32 len, int nb_streams); + /**@}*/ #ifdef __cplusplus diff --git a/code/opus-1.0.2/include/opus_custom.h b/code/opus-1.1/include/opus_custom.h similarity index 97% rename from code/opus-1.0.2/include/opus_custom.h rename to code/opus-1.1/include/opus_custom.h index e7861d6f..41f36bf2 100644 --- a/code/opus-1.0.2/include/opus_custom.h +++ b/code/opus-1.1/include/opus_custom.h @@ -42,15 +42,15 @@ extern "C" { #endif #ifdef CUSTOM_MODES -#define OPUS_CUSTOM_EXPORT OPUS_EXPORT -#define OPUS_CUSTOM_EXPORT_STATIC OPUS_EXPORT +# define OPUS_CUSTOM_EXPORT OPUS_EXPORT +# define OPUS_CUSTOM_EXPORT_STATIC OPUS_EXPORT #else -#define OPUS_CUSTOM_EXPORT -#ifdef CELT_C -#define OPUS_CUSTOM_EXPORT_STATIC static inline -#else -#define OPUS_CUSTOM_EXPORT_STATIC -#endif +# define OPUS_CUSTOM_EXPORT +# ifdef OPUS_BUILD +# define OPUS_CUSTOM_EXPORT_STATIC static OPUS_INLINE +# else +# define OPUS_CUSTOM_EXPORT_STATIC +# endif #endif /** @defgroup opus_custom Opus Custom @@ -126,6 +126,9 @@ OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT OpusCustomMode *opus_custom_mode_crea */ OPUS_CUSTOM_EXPORT void opus_custom_mode_destroy(OpusCustomMode *mode); + +#if !defined(OPUS_BUILD) || defined(CELT_ENCODER_C) + /* Encoder */ /** Gets the size of an OpusCustomEncoder structure. * @param [in] mode OpusCustomMode *: Mode configuration @@ -137,6 +140,28 @@ OPUS_CUSTOM_EXPORT_STATIC OPUS_WARN_UNUSED_RESULT int opus_custom_encoder_get_si int channels ) OPUS_ARG_NONNULL(1); +# ifdef CUSTOM_MODES +/** Initializes a previously allocated encoder state + * The memory pointed to by st must be the size returned by opus_custom_encoder_get_size. + * This is intended for applications which use their own allocator instead of malloc. + * @see opus_custom_encoder_create(),opus_custom_encoder_get_size() + * To reset a previously initialized state use the OPUS_RESET_STATE CTL. + * @param [in] st OpusCustomEncoder*: Encoder state + * @param [in] mode OpusCustomMode *: Contains all the information about the characteristics of + * the stream (must be the same characteristics as used for the + * decoder) + * @param [in] channels int: Number of channels + * @return OPUS_OK Success or @ref opus_errorcodes + */ +OPUS_CUSTOM_EXPORT int opus_custom_encoder_init( + OpusCustomEncoder *st, + const OpusCustomMode *mode, + int channels +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); +# endif +#endif + + /** Creates a new encoder state. Each stream needs its own encoder * state (can't be shared across simultaneous streams). * @param [in] mode OpusCustomMode*: Contains all the information about the characteristics of @@ -152,23 +177,6 @@ OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT OpusCustomEncoder *opus_custom_encode int *error ) OPUS_ARG_NONNULL(1); -/** Initializes a previously allocated encoder state - * The memory pointed to by st must be the size returned by opus_custom_encoder_get_size. - * This is intended for applications which use their own allocator instead of malloc. - * @see opus_custom_encoder_create(),opus_custom_encoder_get_size() - * To reset a previously initialized state use the OPUS_RESET_STATE CTL. - * @param [in] st OpusCustomEncoder*: Encoder state - * @param [in] mode OpusCustomMode *: Contains all the information about the characteristics of - * the stream (must be the same characteristics as used for the - * decoder) - * @param [in] channels int: Number of channels - * @return OPUS_OK Success or @ref opus_errorcodes - */ -OPUS_CUSTOM_EXPORT_STATIC int opus_custom_encoder_init( - OpusCustomEncoder *st, - const OpusCustomMode *mode, - int channels -) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); /** Destroys a an encoder state. * @param[in] st OpusCustomEncoder*: State to be freed. @@ -229,6 +237,8 @@ OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT int opus_custom_encode( */ OPUS_CUSTOM_EXPORT int opus_custom_encoder_ctl(OpusCustomEncoder * OPUS_RESTRICT st, int request, ...) OPUS_ARG_NONNULL(1); + +#if !defined(OPUS_BUILD) || defined(CELT_DECODER_C) /* Decoder */ /** Gets the size of an OpusCustomDecoder structure. @@ -241,20 +251,6 @@ OPUS_CUSTOM_EXPORT_STATIC OPUS_WARN_UNUSED_RESULT int opus_custom_decoder_get_si int channels ) OPUS_ARG_NONNULL(1); -/** Creates a new decoder state. Each stream needs its own decoder state (can't - * be shared across simultaneous streams). - * @param [in] mode OpusCustomMode: Contains all the information about the characteristics of the - * stream (must be the same characteristics as used for the encoder) - * @param [in] channels int: Number of channels - * @param [out] error int*: Returns an error code - * @return Newly created decoder state. - */ -OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT OpusCustomDecoder *opus_custom_decoder_create( - const OpusCustomMode *mode, - int channels, - int *error -) OPUS_ARG_NONNULL(1); - /** Initializes a previously allocated decoder state * The memory pointed to by st must be the size returned by opus_custom_decoder_get_size. * This is intended for applications which use their own allocator instead of malloc. @@ -273,6 +269,23 @@ OPUS_CUSTOM_EXPORT_STATIC int opus_custom_decoder_init( int channels ) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); +#endif + + +/** Creates a new decoder state. Each stream needs its own decoder state (can't + * be shared across simultaneous streams). + * @param [in] mode OpusCustomMode: Contains all the information about the characteristics of the + * stream (must be the same characteristics as used for the encoder) + * @param [in] channels int: Number of channels + * @param [out] error int*: Returns an error code + * @return Newly created decoder state. + */ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT OpusCustomDecoder *opus_custom_decoder_create( + const OpusCustomMode *mode, + int channels, + int *error +) OPUS_ARG_NONNULL(1); + /** Destroys a an decoder state. * @param[in] st OpusCustomDecoder*: State to be freed. */ diff --git a/code/opus-1.0.2/include/opus_defines.h b/code/opus-1.1/include/opus_defines.h similarity index 88% rename from code/opus-1.0.2/include/opus_defines.h rename to code/opus-1.1/include/opus_defines.h index cdde061a..265089f6 100644 --- a/code/opus-1.0.2/include/opus_defines.h +++ b/code/opus-1.1/include/opus_defines.h @@ -64,14 +64,14 @@ extern "C" { /**Export control for opus functions */ #ifndef OPUS_EXPORT -# if defined(__GNUC__) && defined(OPUS_BUILD) -# define OPUS_EXPORT __attribute__ ((visibility ("default"))) -# elif defined(WIN32) && !defined(__MINGW32__) +# if defined(WIN32) # ifdef OPUS_BUILD # define OPUS_EXPORT __declspec(dllexport) # else # define OPUS_EXPORT # endif +# elif defined(__GNUC__) && defined(OPUS_BUILD) +# define OPUS_EXPORT __attribute__ ((visibility ("default"))) # else # define OPUS_EXPORT # endif @@ -98,6 +98,18 @@ extern "C" { # define OPUS_RESTRICT restrict #endif +#if (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L) ) +# if OPUS_GNUC_PREREQ(2,7) +# define OPUS_INLINE __inline__ +# elif (defined(_MSC_VER)) +# define OPUS_INLINE __inline +# else +# define OPUS_INLINE +# endif +#else +# define OPUS_INLINE inline +#endif + /**Warning attributes for opus functions * NONNULL is not used in OPUS_BUILD to avoid the compiler optimizing out * some paranoid null checks. */ @@ -148,8 +160,11 @@ extern "C" { #define OPUS_GET_GAIN_REQUEST 4045 /* Should have been 4035 */ #define OPUS_SET_LSB_DEPTH_REQUEST 4036 #define OPUS_GET_LSB_DEPTH_REQUEST 4037 - #define OPUS_GET_LAST_PACKET_DURATION_REQUEST 4039 +#define OPUS_SET_EXPERT_FRAME_DURATION_REQUEST 4040 +#define OPUS_GET_EXPERT_FRAME_DURATION_REQUEST 4041 +#define OPUS_SET_PREDICTION_DISABLED_REQUEST 4042 +#define OPUS_GET_PREDICTION_DISABLED_REQUEST 4043 /* Don't use 4045, it's already taken by OPUS_GET_GAIN_REQUEST */ @@ -157,6 +172,7 @@ extern "C" { #define __opus_check_int(x) (((void)((x) == (opus_int32)0)), (opus_int32)(x)) #define __opus_check_int_ptr(ptr) ((ptr) + ((ptr) - (opus_int32*)(ptr))) #define __opus_check_uint_ptr(ptr) ((ptr) + ((ptr) - (opus_uint32*)(ptr))) +#define __opus_check_val16_ptr(ptr) ((ptr) + ((ptr) - (opus_val16*)(ptr))) /** @endcond */ /** @defgroup opus_ctlvalues Pre-defined values for CTL interface @@ -185,6 +201,14 @@ extern "C" { #define OPUS_BANDWIDTH_SUPERWIDEBAND 1104 /**<12 kHz bandpass @hideinitializer*/ #define OPUS_BANDWIDTH_FULLBAND 1105 /**<20 kHz bandpass @hideinitializer*/ +#define OPUS_FRAMESIZE_ARG 5000 /**< Select frame size from the argument (default) */ +#define OPUS_FRAMESIZE_2_5_MS 5001 /**< Use 2.5 ms frames */ +#define OPUS_FRAMESIZE_5_MS 5002 /**< Use 5 ms frames */ +#define OPUS_FRAMESIZE_10_MS 5003 /**< Use 10 ms frames */ +#define OPUS_FRAMESIZE_20_MS 5004 /**< Use 20 ms frames */ +#define OPUS_FRAMESIZE_40_MS 5005 /**< Use 40 ms frames */ +#define OPUS_FRAMESIZE_60_MS 5006 /**< Use 60 ms frames */ + /**@}*/ @@ -525,6 +549,53 @@ extern "C" { * @param[out] x opus_int32 *: Number of samples (at current sampling rate). * @hideinitializer */ #define OPUS_GET_LAST_PACKET_DURATION(x) OPUS_GET_LAST_PACKET_DURATION_REQUEST, __opus_check_int_ptr(x) + +/** Configures the encoder's use of variable duration frames. + * When variable duration is enabled, the encoder is free to use a shorter frame + * size than the one requested in the opus_encode*() call. + * It is then the user's responsibility + * to verify how much audio was encoded by checking the ToC byte of the encoded + * packet. The part of the audio that was not encoded needs to be resent to the + * encoder for the next call. Do not use this option unless you really + * know what you are doing. + * @see OPUS_GET_EXPERT_VARIABLE_DURATION + * @param[in] x opus_int32: Allowed values: + *
+ *
OPUS_FRAMESIZE_ARG
Select frame size from the argument (default).
+ *
OPUS_FRAMESIZE_2_5_MS
Use 2.5 ms frames.
+ *
OPUS_FRAMESIZE_5_MS
Use 2.5 ms frames.
+ *
OPUS_FRAMESIZE_10_MS
Use 10 ms frames.
+ *
OPUS_FRAMESIZE_20_MS
Use 20 ms frames.
+ *
OPUS_FRAMESIZE_40_MS
Use 40 ms frames.
+ *
OPUS_FRAMESIZE_60_MS
Use 60 ms frames.
+ *
OPUS_FRAMESIZE_VARIABLE
Optimize the frame size dynamically.
+ *
+ * @hideinitializer */ +#define OPUS_SET_EXPERT_FRAME_DURATION(x) OPUS_SET_EXPERT_FRAME_DURATION_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured use of variable duration frames. + * @see OPUS_SET_EXPERT_VARIABLE_DURATION + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
OPUS_FRAMESIZE_ARG
Select frame size from the argument (default).
+ *
OPUS_FRAMESIZE_2_5_MS
Use 2.5 ms frames.
+ *
OPUS_FRAMESIZE_5_MS
Use 2.5 ms frames.
+ *
OPUS_FRAMESIZE_10_MS
Use 10 ms frames.
+ *
OPUS_FRAMESIZE_20_MS
Use 20 ms frames.
+ *
OPUS_FRAMESIZE_40_MS
Use 40 ms frames.
+ *
OPUS_FRAMESIZE_60_MS
Use 60 ms frames.
+ *
OPUS_FRAMESIZE_VARIABLE
Optimize the frame size dynamically.
+ *
+ * @hideinitializer */ +#define OPUS_GET_EXPERT_FRAME_DURATION(x) OPUS_GET_EXPERT_FRAME_DURATION_REQUEST, __opus_check_int_ptr(x) + +/** If set to 1, disables almost all use of prediction, making frames almost + completely independent. This reduces quality. (default : 0) + * @hideinitializer */ +#define OPUS_SET_PREDICTION_DISABLED(x) OPUS_SET_PREDICTION_DISABLED_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured prediction status. + * @hideinitializer */ +#define OPUS_GET_PREDICTION_DISABLED(x) OPUS_GET_PREDICTION_DISABLED_REQUEST, __opus_check_int_ptr(x) + /**@}*/ /** @defgroup opus_genericctls Generic CTLs diff --git a/code/opus-1.0.2/include/opus_multistream.h b/code/opus-1.1/include/opus_multistream.h similarity index 97% rename from code/opus-1.0.2/include/opus_multistream.h rename to code/opus-1.1/include/opus_multistream.h index 658067f7..ae599793 100644 --- a/code/opus-1.0.2/include/opus_multistream.h +++ b/code/opus-1.1/include/opus_multistream.h @@ -205,6 +205,12 @@ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_multistream_encoder_get_size int coupled_streams ); +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_multistream_surround_encoder_get_size( + int channels, + int mapping_family +); + + /** Allocates and initializes a multistream encoder state. * Call opus_multistream_encoder_destroy() to release * this object when finished. @@ -258,6 +264,17 @@ OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusMSEncoder *opus_multistream_encoder_crea int *error ) OPUS_ARG_NONNULL(5); +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusMSEncoder *opus_multistream_surround_encoder_create( + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application, + int *error +) OPUS_ARG_NONNULL(5); + /** Initialize a previously allocated multistream encoder state. * The memory pointed to by \a st must be at least the size returned by * opus_multistream_encoder_get_size(). @@ -316,6 +333,17 @@ OPUS_EXPORT int opus_multistream_encoder_init( int application ) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(6); +OPUS_EXPORT int opus_multistream_surround_encoder_init( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(6); + /** Encodes a multistream Opus frame. * @param st OpusMSEncoder*: Multistream encoder state. * @param[in] pcm const opus_int16*: The input signal as interleaved diff --git a/code/opus-1.0.2/include/opus_types.h b/code/opus-1.1/include/opus_types.h similarity index 100% rename from code/opus-1.0.2/include/opus_types.h rename to code/opus-1.1/include/opus_types.h diff --git a/code/opus-1.0.2/silk/A2NLSF.c b/code/opus-1.1/silk/A2NLSF.c similarity index 97% rename from code/opus-1.0.2/silk/A2NLSF.c rename to code/opus-1.1/silk/A2NLSF.c index 49d5d9d9..74b1b95d 100644 --- a/code/opus-1.0.2/silk/A2NLSF.c +++ b/code/opus-1.1/silk/A2NLSF.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -44,7 +44,7 @@ POSSIBILITY OF SUCH DAMAGE. /* Helper function for A2NLSF(..) */ /* Transforms polynomials from cos(n*f) to cos(f)^n */ -static inline void silk_A2NLSF_trans_poly( +static OPUS_INLINE void silk_A2NLSF_trans_poly( opus_int32 *p, /* I/O Polynomial */ const opus_int dd /* I Polynomial order (= filter order / 2 ) */ ) @@ -60,7 +60,7 @@ static inline void silk_A2NLSF_trans_poly( } /* Helper function for A2NLSF(..) */ /* Polynomial evaluation */ -static inline opus_int32 silk_A2NLSF_eval_poly( /* return the polynomial evaluation, in Q16 */ +static OPUS_INLINE opus_int32 silk_A2NLSF_eval_poly( /* return the polynomial evaluation, in Q16 */ opus_int32 *p, /* I Polynomial, Q16 */ const opus_int32 x, /* I Evaluation point, Q12 */ const opus_int dd /* I Order */ @@ -77,7 +77,7 @@ static inline opus_int32 silk_A2NLSF_eval_poly( /* return the polynomial evaluat return y32; } -static inline void silk_A2NLSF_init( +static OPUS_INLINE void silk_A2NLSF_init( const opus_int32 *a_Q16, opus_int32 *P, opus_int32 *Q, diff --git a/code/opus-1.0.2/silk/API.h b/code/opus-1.1/silk/API.h similarity index 98% rename from code/opus-1.0.2/silk/API.h rename to code/opus-1.1/silk/API.h index 4b8ca12a..f0601bcf 100644 --- a/code/opus-1.0.2/silk/API.h +++ b/code/opus-1.1/silk/API.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -64,6 +64,7 @@ opus_int silk_Get_Encoder_Size( /* O Returns error co /*************************/ opus_int silk_InitEncoder( /* O Returns error code */ void *encState, /* I/O State */ + int arch, /* I Run-time architecture */ silk_EncControlStruct *encStatus /* O Encoder Status */ ); diff --git a/code/opus-1.0.2/silk/CNG.c b/code/opus-1.1/silk/CNG.c similarity index 97% rename from code/opus-1.0.2/silk/CNG.c rename to code/opus-1.1/silk/CNG.c index d0a619c1..8481d95d 100644 --- a/code/opus-1.0.2/silk/CNG.c +++ b/code/opus-1.1/silk/CNG.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,9 +30,10 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "stack_alloc.h" /* Generates excitation for CNG LPC synthesis */ -static inline void silk_CNG_exc( +static OPUS_INLINE void silk_CNG_exc( opus_int32 residual_Q10[], /* O CNG residual signal Q10 */ opus_int32 exc_buf_Q14[], /* I Random samples buffer Q10 */ opus_int32 Gain_Q16, /* I Gain to apply */ @@ -86,8 +87,8 @@ void silk_CNG( opus_int i, subfr; opus_int32 sum_Q6, max_Gain_Q16; opus_int16 A_Q12[ MAX_LPC_ORDER ]; - opus_int32 CNG_sig_Q10[ MAX_FRAME_LENGTH + MAX_LPC_ORDER ]; silk_CNG_struct *psCNG = &psDec->sCNG; + SAVE_STACK; if( psDec->fs_kHz != psCNG->fs_kHz ) { /* Reset state */ @@ -123,6 +124,9 @@ void silk_CNG( /* Add CNG when packet is lost or during DTX */ if( psDec->lossCnt ) { + VARDECL( opus_int32, CNG_sig_Q10 ); + + ALLOC( CNG_sig_Q10, length + MAX_LPC_ORDER, opus_int32 ); /* Generate CNG excitation */ silk_CNG_exc( CNG_sig_Q10 + MAX_LPC_ORDER, psCNG->CNG_exc_buf_Q14, psCNG->CNG_smth_Gain_Q16, length, &psCNG->rand_seed ); @@ -164,4 +168,5 @@ void silk_CNG( } else { silk_memset( psCNG->CNG_synth_state, 0, psDec->LPC_order * sizeof( opus_int32 ) ); } + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/HP_variable_cutoff.c b/code/opus-1.1/silk/HP_variable_cutoff.c similarity index 99% rename from code/opus-1.0.2/silk/HP_variable_cutoff.c rename to code/opus-1.1/silk/HP_variable_cutoff.c index 199dbb34..bbe10f04 100644 --- a/code/opus-1.0.2/silk/HP_variable_cutoff.c +++ b/code/opus-1.1/silk/HP_variable_cutoff.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/Inlines.h b/code/opus-1.1/silk/Inlines.h similarity index 93% rename from code/opus-1.0.2/silk/Inlines.h rename to code/opus-1.1/silk/Inlines.h index 87ac2e20..ec986cdf 100644 --- a/code/opus-1.0.2/silk/Inlines.h +++ b/code/opus-1.1/silk/Inlines.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -26,7 +26,7 @@ POSSIBILITY OF SUCH DAMAGE. ***********************************************************************/ /*! \file silk_Inlines.h - * \brief silk_Inlines.h defines inline signal processing functions. + * \brief silk_Inlines.h defines OPUS_INLINE signal processing functions. */ #ifndef SILK_FIX_INLINES_H @@ -38,7 +38,7 @@ extern "C" #endif /* count leading zeros of opus_int64 */ -static inline opus_int32 silk_CLZ64( opus_int64 in ) +static OPUS_INLINE opus_int32 silk_CLZ64( opus_int64 in ) { opus_int32 in_upper; @@ -53,7 +53,7 @@ static inline opus_int32 silk_CLZ64( opus_int64 in ) } /* get number of leading zeros and fractional part (the bits right after the leading one */ -static inline void silk_CLZ_FRAC( +static OPUS_INLINE void silk_CLZ_FRAC( opus_int32 in, /* I input */ opus_int32 *lz, /* O number of leading zeros */ opus_int32 *frac_Q7 /* O the 7 bits right after the leading one */ @@ -68,7 +68,7 @@ static inline void silk_CLZ_FRAC( /* Approximation of square root */ /* Accuracy: < +/- 10% for output values > 15 */ /* < +/- 2.5% for output values > 120 */ -static inline opus_int32 silk_SQRT_APPROX( opus_int32 x ) +static OPUS_INLINE opus_int32 silk_SQRT_APPROX( opus_int32 x ) { opus_int32 y, lz, frac_Q7; @@ -94,7 +94,7 @@ static inline opus_int32 silk_SQRT_APPROX( opus_int32 x ) } /* Divide two int32 values and return result as int32 in a given Q-domain */ -static inline opus_int32 silk_DIV32_varQ( /* O returns a good approximation of "(a32 << Qres) / b32" */ +static OPUS_INLINE opus_int32 silk_DIV32_varQ( /* O returns a good approximation of "(a32 << Qres) / b32" */ const opus_int32 a32, /* I numerator (Q0) */ const opus_int32 b32, /* I denominator (Q0) */ const opus_int Qres /* I Q-domain of result (>= 0) */ @@ -140,7 +140,7 @@ static inline opus_int32 silk_DIV32_varQ( /* O returns a good approximation } /* Invert int32 value and return result as int32 in a given Q-domain */ -static inline opus_int32 silk_INVERSE32_varQ( /* O returns a good approximation of "(1 << Qres) / b32" */ +static OPUS_INLINE opus_int32 silk_INVERSE32_varQ( /* O returns a good approximation of "(1 << Qres) / b32" */ const opus_int32 b32, /* I denominator (Q0) */ const opus_int Qres /* I Q-domain of result (> 0) */ ) diff --git a/code/opus-1.0.2/silk/LPC_analysis_filter.c b/code/opus-1.1/silk/LPC_analysis_filter.c similarity index 89% rename from code/opus-1.0.2/silk/LPC_analysis_filter.c rename to code/opus-1.1/silk/LPC_analysis_filter.c index 421dba0b..9d1f16cb 100644 --- a/code/opus-1.0.2/silk/LPC_analysis_filter.c +++ b/code/opus-1.1/silk/LPC_analysis_filter.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "SigProc_FIX.h" +#include "celt_lpc.h" /*******************************************/ /* LPC analysis filter */ @@ -46,14 +47,33 @@ void silk_LPC_analysis_filter( const opus_int32 d /* I Filter order */ ) { - opus_int ix, j; + opus_int j; +#ifdef FIXED_POINT + opus_int16 mem[SILK_MAX_ORDER_LPC]; + opus_int16 num[SILK_MAX_ORDER_LPC]; +#else + int ix; opus_int32 out32_Q12, out32; const opus_int16 *in_ptr; +#endif silk_assert( d >= 6 ); silk_assert( (d & 1) == 0 ); silk_assert( d <= len ); +#ifdef FIXED_POINT + silk_assert( d <= SILK_MAX_ORDER_LPC ); + for ( j = 0; j < d; j++ ) { + num[ j ] = -B[ j ]; + } + for (j=0;j> 16) * (opus_int32)((opus_int16)b32) + (((a32 & 0x0000FFFF) * (opus_int32)((opus_int16)b32)) >> 16); return ret; } #undef silk_SMLAWB -static inline opus_int32 silk_SMLAWB(opus_int32 a32, opus_int32 b32, opus_int32 c32){ +static OPUS_INLINE opus_int32 silk_SMLAWB(opus_int32 a32, opus_int32 b32, opus_int32 c32){ opus_int32 ret; ops_count += 5; ret = ((a32) + ((((b32) >> 16) * (opus_int32)((opus_int16)(c32))) + ((((b32) & 0x0000FFFF) * (opus_int32)((opus_int16)(c32))) >> 16))); @@ -97,14 +97,14 @@ static inline opus_int32 silk_SMLAWB(opus_int32 a32, opus_int32 b32, opus_int32 } #undef silk_SMULWT -static inline opus_int32 silk_SMULWT(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int32 silk_SMULWT(opus_int32 a32, opus_int32 b32){ opus_int32 ret; ops_count += 4; ret = (a32 >> 16) * (b32 >> 16) + (((a32 & 0x0000FFFF) * (b32 >> 16)) >> 16); return ret; } #undef silk_SMLAWT -static inline opus_int32 silk_SMLAWT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ +static OPUS_INLINE opus_int32 silk_SMLAWT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ opus_int32 ret; ops_count += 4; ret = a32 + ((b32 >> 16) * (c32 >> 16)) + (((b32 & 0x0000FFFF) * ((c32 >> 16)) >> 16)); @@ -112,14 +112,14 @@ static inline opus_int32 silk_SMLAWT(opus_int32 a32, opus_int32 b32, opus_int32 } #undef silk_SMULBB -static inline opus_int32 silk_SMULBB(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int32 silk_SMULBB(opus_int32 a32, opus_int32 b32){ opus_int32 ret; ops_count += 1; ret = (opus_int32)((opus_int16)a32) * (opus_int32)((opus_int16)b32); return ret; } #undef silk_SMLABB -static inline opus_int32 silk_SMLABB(opus_int32 a32, opus_int32 b32, opus_int32 c32){ +static OPUS_INLINE opus_int32 silk_SMLABB(opus_int32 a32, opus_int32 b32, opus_int32 c32){ opus_int32 ret; ops_count += 1; ret = a32 + (opus_int32)((opus_int16)b32) * (opus_int32)((opus_int16)c32); @@ -127,7 +127,7 @@ static inline opus_int32 silk_SMLABB(opus_int32 a32, opus_int32 b32, opus_int32 } #undef silk_SMULBT -static inline opus_int32 silk_SMULBT(opus_int32 a32, opus_int32 b32 ){ +static OPUS_INLINE opus_int32 silk_SMULBT(opus_int32 a32, opus_int32 b32 ){ opus_int32 ret; ops_count += 4; ret = ((opus_int32)((opus_int16)a32)) * (b32 >> 16); @@ -135,7 +135,7 @@ static inline opus_int32 silk_SMULBT(opus_int32 a32, opus_int32 b32 ){ } #undef silk_SMLABT -static inline opus_int32 silk_SMLABT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ +static OPUS_INLINE opus_int32 silk_SMLABT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ opus_int32 ret; ops_count += 1; ret = a32 + ((opus_int32)((opus_int16)b32)) * (c32 >> 16); @@ -143,7 +143,7 @@ static inline opus_int32 silk_SMLABT(opus_int32 a32, opus_int32 b32, opus_int32 } #undef silk_SMULTT -static inline opus_int32 silk_SMULTT(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int32 silk_SMULTT(opus_int32 a32, opus_int32 b32){ opus_int32 ret; ops_count += 1; ret = (a32 >> 16) * (b32 >> 16); @@ -151,7 +151,7 @@ static inline opus_int32 silk_SMULTT(opus_int32 a32, opus_int32 b32){ } #undef silk_SMLATT -static inline opus_int32 silk_SMLATT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ +static OPUS_INLINE opus_int32 silk_SMLATT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ opus_int32 ret; ops_count += 1; ret = a32 + (b32 >> 16) * (c32 >> 16); @@ -179,7 +179,7 @@ static inline opus_int32 silk_SMLATT(opus_int32 a32, opus_int32 b32, opus_int32 #define silk_SMLAWT_ovflw silk_SMLAWT #undef silk_SMULL -static inline opus_int64 silk_SMULL(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int64 silk_SMULL(opus_int32 a32, opus_int32 b32){ opus_int64 ret; ops_count += 8; ret = ((opus_int64)(a32) * /*(opus_int64)*/(b32)); @@ -187,14 +187,14 @@ static inline opus_int64 silk_SMULL(opus_int32 a32, opus_int32 b32){ } #undef silk_SMLAL -static inline opus_int64 silk_SMLAL(opus_int64 a64, opus_int32 b32, opus_int32 c32){ +static OPUS_INLINE opus_int64 silk_SMLAL(opus_int64 a64, opus_int32 b32, opus_int32 c32){ opus_int64 ret; ops_count += 8; ret = a64 + ((opus_int64)(b32) * /*(opus_int64)*/(c32)); return ret; } #undef silk_SMLALBB -static inline opus_int64 silk_SMLALBB(opus_int64 a64, opus_int16 b16, opus_int16 c16){ +static OPUS_INLINE opus_int64 silk_SMLALBB(opus_int64 a64, opus_int16 b16, opus_int16 c16){ opus_int64 ret; ops_count += 4; ret = a64 + ((opus_int64)(b16) * /*(opus_int64)*/(c16)); @@ -202,7 +202,7 @@ static inline opus_int64 silk_SMLALBB(opus_int64 a64, opus_int16 b16, opus_int16 } #undef SigProcFIX_CLZ16 -static inline opus_int32 SigProcFIX_CLZ16(opus_int16 in16) +static OPUS_INLINE opus_int32 SigProcFIX_CLZ16(opus_int16 in16) { opus_int32 out32 = 0; ops_count += 10; @@ -240,7 +240,7 @@ static inline opus_int32 SigProcFIX_CLZ16(opus_int16 in16) } #undef SigProcFIX_CLZ32 -static inline opus_int32 SigProcFIX_CLZ32(opus_int32 in32) +static OPUS_INLINE opus_int32 SigProcFIX_CLZ32(opus_int32 in32) { /* test highest 16 bits and convert to opus_int16 */ ops_count += 2; @@ -252,19 +252,19 @@ static inline opus_int32 SigProcFIX_CLZ32(opus_int32 in32) } #undef silk_DIV32 -static inline opus_int32 silk_DIV32(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int32 silk_DIV32(opus_int32 a32, opus_int32 b32){ ops_count += 64; return a32 / b32; } #undef silk_DIV32_16 -static inline opus_int32 silk_DIV32_16(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int32 silk_DIV32_16(opus_int32 a32, opus_int32 b32){ ops_count += 32; return a32 / b32; } #undef silk_SAT8 -static inline opus_int8 silk_SAT8(opus_int64 a){ +static OPUS_INLINE opus_int8 silk_SAT8(opus_int64 a){ opus_int8 tmp; ops_count += 1; tmp = (opus_int8)((a) > silk_int8_MAX ? silk_int8_MAX : \ @@ -273,7 +273,7 @@ static inline opus_int8 silk_SAT8(opus_int64 a){ } #undef silk_SAT16 -static inline opus_int16 silk_SAT16(opus_int64 a){ +static OPUS_INLINE opus_int16 silk_SAT16(opus_int64 a){ opus_int16 tmp; ops_count += 1; tmp = (opus_int16)((a) > silk_int16_MAX ? silk_int16_MAX : \ @@ -281,7 +281,7 @@ static inline opus_int16 silk_SAT16(opus_int64 a){ return(tmp); } #undef silk_SAT32 -static inline opus_int32 silk_SAT32(opus_int64 a){ +static OPUS_INLINE opus_int32 silk_SAT32(opus_int64 a){ opus_int32 tmp; ops_count += 1; tmp = (opus_int32)((a) > silk_int32_MAX ? silk_int32_MAX : \ @@ -289,7 +289,7 @@ static inline opus_int32 silk_SAT32(opus_int64 a){ return(tmp); } #undef silk_POS_SAT32 -static inline opus_int32 silk_POS_SAT32(opus_int64 a){ +static OPUS_INLINE opus_int32 silk_POS_SAT32(opus_int64 a){ opus_int32 tmp; ops_count += 1; tmp = (opus_int32)((a) > silk_int32_MAX ? silk_int32_MAX : (a)); @@ -297,14 +297,14 @@ static inline opus_int32 silk_POS_SAT32(opus_int64 a){ } #undef silk_ADD_POS_SAT8 -static inline opus_int8 silk_ADD_POS_SAT8(opus_int64 a, opus_int64 b){ +static OPUS_INLINE opus_int8 silk_ADD_POS_SAT8(opus_int64 a, opus_int64 b){ opus_int8 tmp; ops_count += 1; tmp = (opus_int8)((((a)+(b)) & 0x80) ? silk_int8_MAX : ((a)+(b))); return(tmp); } #undef silk_ADD_POS_SAT16 -static inline opus_int16 silk_ADD_POS_SAT16(opus_int64 a, opus_int64 b){ +static OPUS_INLINE opus_int16 silk_ADD_POS_SAT16(opus_int64 a, opus_int64 b){ opus_int16 tmp; ops_count += 1; tmp = (opus_int16)((((a)+(b)) & 0x8000) ? silk_int16_MAX : ((a)+(b))); @@ -312,7 +312,7 @@ static inline opus_int16 silk_ADD_POS_SAT16(opus_int64 a, opus_int64 b){ } #undef silk_ADD_POS_SAT32 -static inline opus_int32 silk_ADD_POS_SAT32(opus_int64 a, opus_int64 b){ +static OPUS_INLINE opus_int32 silk_ADD_POS_SAT32(opus_int64 a, opus_int64 b){ opus_int32 tmp; ops_count += 1; tmp = (opus_int32)((((a)+(b)) & 0x80000000) ? silk_int32_MAX : ((a)+(b))); @@ -320,7 +320,7 @@ static inline opus_int32 silk_ADD_POS_SAT32(opus_int64 a, opus_int64 b){ } #undef silk_ADD_POS_SAT64 -static inline opus_int64 silk_ADD_POS_SAT64(opus_int64 a, opus_int64 b){ +static OPUS_INLINE opus_int64 silk_ADD_POS_SAT64(opus_int64 a, opus_int64 b){ opus_int64 tmp; ops_count += 1; tmp = ((((a)+(b)) & 0x8000000000000000LL) ? silk_int64_MAX : ((a)+(b))); @@ -328,40 +328,40 @@ static inline opus_int64 silk_ADD_POS_SAT64(opus_int64 a, opus_int64 b){ } #undef silk_LSHIFT8 -static inline opus_int8 silk_LSHIFT8(opus_int8 a, opus_int32 shift){ +static OPUS_INLINE opus_int8 silk_LSHIFT8(opus_int8 a, opus_int32 shift){ opus_int8 ret; ops_count += 1; ret = a << shift; return ret; } #undef silk_LSHIFT16 -static inline opus_int16 silk_LSHIFT16(opus_int16 a, opus_int32 shift){ +static OPUS_INLINE opus_int16 silk_LSHIFT16(opus_int16 a, opus_int32 shift){ opus_int16 ret; ops_count += 1; ret = a << shift; return ret; } #undef silk_LSHIFT32 -static inline opus_int32 silk_LSHIFT32(opus_int32 a, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_LSHIFT32(opus_int32 a, opus_int32 shift){ opus_int32 ret; ops_count += 1; ret = a << shift; return ret; } #undef silk_LSHIFT64 -static inline opus_int64 silk_LSHIFT64(opus_int64 a, opus_int shift){ +static OPUS_INLINE opus_int64 silk_LSHIFT64(opus_int64 a, opus_int shift){ ops_count += 1; return a << shift; } #undef silk_LSHIFT_ovflw -static inline opus_int32 silk_LSHIFT_ovflw(opus_int32 a, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_LSHIFT_ovflw(opus_int32 a, opus_int32 shift){ ops_count += 1; return a << shift; } #undef silk_LSHIFT_uint -static inline opus_uint32 silk_LSHIFT_uint(opus_uint32 a, opus_int32 shift){ +static OPUS_INLINE opus_uint32 silk_LSHIFT_uint(opus_uint32 a, opus_int32 shift){ opus_uint32 ret; ops_count += 1; ret = a << shift; @@ -369,83 +369,83 @@ static inline opus_uint32 silk_LSHIFT_uint(opus_uint32 a, opus_int32 shift){ } #undef silk_RSHIFT8 -static inline opus_int8 silk_RSHIFT8(opus_int8 a, opus_int32 shift){ +static OPUS_INLINE opus_int8 silk_RSHIFT8(opus_int8 a, opus_int32 shift){ ops_count += 1; return a >> shift; } #undef silk_RSHIFT16 -static inline opus_int16 silk_RSHIFT16(opus_int16 a, opus_int32 shift){ +static OPUS_INLINE opus_int16 silk_RSHIFT16(opus_int16 a, opus_int32 shift){ ops_count += 1; return a >> shift; } #undef silk_RSHIFT32 -static inline opus_int32 silk_RSHIFT32(opus_int32 a, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_RSHIFT32(opus_int32 a, opus_int32 shift){ ops_count += 1; return a >> shift; } #undef silk_RSHIFT64 -static inline opus_int64 silk_RSHIFT64(opus_int64 a, opus_int64 shift){ +static OPUS_INLINE opus_int64 silk_RSHIFT64(opus_int64 a, opus_int64 shift){ ops_count += 1; return a >> shift; } #undef silk_RSHIFT_uint -static inline opus_uint32 silk_RSHIFT_uint(opus_uint32 a, opus_int32 shift){ +static OPUS_INLINE opus_uint32 silk_RSHIFT_uint(opus_uint32 a, opus_int32 shift){ ops_count += 1; return a >> shift; } #undef silk_ADD_LSHIFT -static inline opus_int32 silk_ADD_LSHIFT(opus_int32 a, opus_int32 b, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_ADD_LSHIFT(opus_int32 a, opus_int32 b, opus_int32 shift){ opus_int32 ret; ops_count += 1; ret = a + (b << shift); return ret; /* shift >= 0*/ } #undef silk_ADD_LSHIFT32 -static inline opus_int32 silk_ADD_LSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_ADD_LSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ opus_int32 ret; ops_count += 1; ret = a + (b << shift); return ret; /* shift >= 0*/ } #undef silk_ADD_LSHIFT_uint -static inline opus_uint32 silk_ADD_LSHIFT_uint(opus_uint32 a, opus_uint32 b, opus_int32 shift){ +static OPUS_INLINE opus_uint32 silk_ADD_LSHIFT_uint(opus_uint32 a, opus_uint32 b, opus_int32 shift){ opus_uint32 ret; ops_count += 1; ret = a + (b << shift); return ret; /* shift >= 0*/ } #undef silk_ADD_RSHIFT -static inline opus_int32 silk_ADD_RSHIFT(opus_int32 a, opus_int32 b, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_ADD_RSHIFT(opus_int32 a, opus_int32 b, opus_int32 shift){ opus_int32 ret; ops_count += 1; ret = a + (b >> shift); return ret; /* shift > 0*/ } #undef silk_ADD_RSHIFT32 -static inline opus_int32 silk_ADD_RSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_ADD_RSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ opus_int32 ret; ops_count += 1; ret = a + (b >> shift); return ret; /* shift > 0*/ } #undef silk_ADD_RSHIFT_uint -static inline opus_uint32 silk_ADD_RSHIFT_uint(opus_uint32 a, opus_uint32 b, opus_int32 shift){ +static OPUS_INLINE opus_uint32 silk_ADD_RSHIFT_uint(opus_uint32 a, opus_uint32 b, opus_int32 shift){ opus_uint32 ret; ops_count += 1; ret = a + (b >> shift); return ret; /* shift > 0*/ } #undef silk_SUB_LSHIFT32 -static inline opus_int32 silk_SUB_LSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_SUB_LSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ opus_int32 ret; ops_count += 1; ret = a - (b << shift); return ret; /* shift >= 0*/ } #undef silk_SUB_RSHIFT32 -static inline opus_int32 silk_SUB_RSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_SUB_RSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ opus_int32 ret; ops_count += 1; ret = a - (b >> shift); @@ -453,7 +453,7 @@ static inline opus_int32 silk_SUB_RSHIFT32(opus_int32 a, opus_int32 b, opus_int3 } #undef silk_RSHIFT_ROUND -static inline opus_int32 silk_RSHIFT_ROUND(opus_int32 a, opus_int32 shift){ +static OPUS_INLINE opus_int32 silk_RSHIFT_ROUND(opus_int32 a, opus_int32 shift){ opus_int32 ret; ops_count += 3; ret = shift == 1 ? (a >> 1) + (a & 1) : ((a >> (shift - 1)) + 1) >> 1; @@ -461,7 +461,7 @@ static inline opus_int32 silk_RSHIFT_ROUND(opus_int32 a, opus_int32 shift){ } #undef silk_RSHIFT_ROUND64 -static inline opus_int64 silk_RSHIFT_ROUND64(opus_int64 a, opus_int32 shift){ +static OPUS_INLINE opus_int64 silk_RSHIFT_ROUND64(opus_int64 a, opus_int32 shift){ opus_int64 ret; ops_count += 6; ret = shift == 1 ? (a >> 1) + (a & 1) : ((a >> (shift - 1)) + 1) >> 1; @@ -469,13 +469,13 @@ static inline opus_int64 silk_RSHIFT_ROUND64(opus_int64 a, opus_int32 shift){ } #undef silk_abs_int64 -static inline opus_int64 silk_abs_int64(opus_int64 a){ +static OPUS_INLINE opus_int64 silk_abs_int64(opus_int64 a){ ops_count += 1; return (((a) > 0) ? (a) : -(a)); /* Be careful, silk_abs returns wrong when input equals to silk_intXX_MIN*/ } #undef silk_abs_int32 -static inline opus_int32 silk_abs_int32(opus_int32 a){ +static OPUS_INLINE opus_int32 silk_abs_int32(opus_int32 a){ ops_count += 1; return silk_abs(a); } @@ -498,7 +498,7 @@ static silk_sign(a){ } #undef silk_ADD16 -static inline opus_int16 silk_ADD16(opus_int16 a, opus_int16 b){ +static OPUS_INLINE opus_int16 silk_ADD16(opus_int16 a, opus_int16 b){ opus_int16 ret; ops_count += 1; ret = a + b; @@ -506,7 +506,7 @@ static inline opus_int16 silk_ADD16(opus_int16 a, opus_int16 b){ } #undef silk_ADD32 -static inline opus_int32 silk_ADD32(opus_int32 a, opus_int32 b){ +static OPUS_INLINE opus_int32 silk_ADD32(opus_int32 a, opus_int32 b){ opus_int32 ret; ops_count += 1; ret = a + b; @@ -514,7 +514,7 @@ static inline opus_int32 silk_ADD32(opus_int32 a, opus_int32 b){ } #undef silk_ADD64 -static inline opus_int64 silk_ADD64(opus_int64 a, opus_int64 b){ +static OPUS_INLINE opus_int64 silk_ADD64(opus_int64 a, opus_int64 b){ opus_int64 ret; ops_count += 2; ret = a + b; @@ -522,7 +522,7 @@ static inline opus_int64 silk_ADD64(opus_int64 a, opus_int64 b){ } #undef silk_SUB16 -static inline opus_int16 silk_SUB16(opus_int16 a, opus_int16 b){ +static OPUS_INLINE opus_int16 silk_SUB16(opus_int16 a, opus_int16 b){ opus_int16 ret; ops_count += 1; ret = a - b; @@ -530,7 +530,7 @@ static inline opus_int16 silk_SUB16(opus_int16 a, opus_int16 b){ } #undef silk_SUB32 -static inline opus_int32 silk_SUB32(opus_int32 a, opus_int32 b){ +static OPUS_INLINE opus_int32 silk_SUB32(opus_int32 a, opus_int32 b){ opus_int32 ret; ops_count += 1; ret = a - b; @@ -538,7 +538,7 @@ static inline opus_int32 silk_SUB32(opus_int32 a, opus_int32 b){ } #undef silk_SUB64 -static inline opus_int64 silk_SUB64(opus_int64 a, opus_int64 b){ +static OPUS_INLINE opus_int64 silk_SUB64(opus_int64 a, opus_int64 b){ opus_int64 ret; ops_count += 2; ret = a - b; @@ -546,7 +546,7 @@ static inline opus_int64 silk_SUB64(opus_int64 a, opus_int64 b){ } #undef silk_ADD_SAT16 -static inline opus_int16 silk_ADD_SAT16( opus_int16 a16, opus_int16 b16 ) { +static OPUS_INLINE opus_int16 silk_ADD_SAT16( opus_int16 a16, opus_int16 b16 ) { opus_int16 res; /* Nb will be counted in AKP_add32 and silk_SAT16*/ res = (opus_int16)silk_SAT16( silk_ADD32( (opus_int32)(a16), (b16) ) ); @@ -554,7 +554,7 @@ static inline opus_int16 silk_ADD_SAT16( opus_int16 a16, opus_int16 b16 ) { } #undef silk_ADD_SAT32 -static inline opus_int32 silk_ADD_SAT32(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int32 silk_ADD_SAT32(opus_int32 a32, opus_int32 b32){ opus_int32 res; ops_count += 1; res = ((((a32) + (b32)) & 0x80000000) == 0 ? \ @@ -564,7 +564,7 @@ static inline opus_int32 silk_ADD_SAT32(opus_int32 a32, opus_int32 b32){ } #undef silk_ADD_SAT64 -static inline opus_int64 silk_ADD_SAT64( opus_int64 a64, opus_int64 b64 ) { +static OPUS_INLINE opus_int64 silk_ADD_SAT64( opus_int64 a64, opus_int64 b64 ) { opus_int64 res; ops_count += 1; res = ((((a64) + (b64)) & 0x8000000000000000LL) == 0 ? \ @@ -574,7 +574,7 @@ static inline opus_int64 silk_ADD_SAT64( opus_int64 a64, opus_int64 b64 ) { } #undef silk_SUB_SAT16 -static inline opus_int16 silk_SUB_SAT16( opus_int16 a16, opus_int16 b16 ) { +static OPUS_INLINE opus_int16 silk_SUB_SAT16( opus_int16 a16, opus_int16 b16 ) { opus_int16 res; silk_assert(0); /* Nb will be counted in sub-macros*/ @@ -583,7 +583,7 @@ static inline opus_int16 silk_SUB_SAT16( opus_int16 a16, opus_int16 b16 ) { } #undef silk_SUB_SAT32 -static inline opus_int32 silk_SUB_SAT32( opus_int32 a32, opus_int32 b32 ) { +static OPUS_INLINE opus_int32 silk_SUB_SAT32( opus_int32 a32, opus_int32 b32 ) { opus_int32 res; ops_count += 1; res = ((((a32)-(b32)) & 0x80000000) == 0 ? \ @@ -593,7 +593,7 @@ static inline opus_int32 silk_SUB_SAT32( opus_int32 a32, opus_int32 b32 ) { } #undef silk_SUB_SAT64 -static inline opus_int64 silk_SUB_SAT64( opus_int64 a64, opus_int64 b64 ) { +static OPUS_INLINE opus_int64 silk_SUB_SAT64( opus_int64 a64, opus_int64 b64 ) { opus_int64 res; ops_count += 1; res = ((((a64)-(b64)) & 0x8000000000000000LL) == 0 ? \ @@ -604,7 +604,7 @@ static inline opus_int64 silk_SUB_SAT64( opus_int64 a64, opus_int64 b64 ) { } #undef silk_SMULWW -static inline opus_int32 silk_SMULWW(opus_int32 a32, opus_int32 b32){ +static OPUS_INLINE opus_int32 silk_SMULWW(opus_int32 a32, opus_int32 b32){ opus_int32 ret; /* Nb will be counted in sub-macros*/ ret = silk_MLA(silk_SMULWB((a32), (b32)), (a32), silk_RSHIFT_ROUND((b32), 16)); @@ -612,7 +612,7 @@ static inline opus_int32 silk_SMULWW(opus_int32 a32, opus_int32 b32){ } #undef silk_SMLAWW -static inline opus_int32 silk_SMLAWW(opus_int32 a32, opus_int32 b32, opus_int32 c32){ +static OPUS_INLINE opus_int32 silk_SMLAWW(opus_int32 a32, opus_int32 b32, opus_int32 c32){ opus_int32 ret; /* Nb will be counted in sub-macros*/ ret = silk_MLA(silk_SMLAWB((a32), (b32), (c32)), (b32), silk_RSHIFT_ROUND((c32), 16)); @@ -620,26 +620,26 @@ static inline opus_int32 silk_SMLAWW(opus_int32 a32, opus_int32 b32, opus_int32 } #undef silk_min_int -static inline opus_int silk_min_int(opus_int a, opus_int b) +static OPUS_INLINE opus_int silk_min_int(opus_int a, opus_int b) { ops_count += 1; return (((a) < (b)) ? (a) : (b)); } #undef silk_min_16 -static inline opus_int16 silk_min_16(opus_int16 a, opus_int16 b) +static OPUS_INLINE opus_int16 silk_min_16(opus_int16 a, opus_int16 b) { ops_count += 1; return (((a) < (b)) ? (a) : (b)); } #undef silk_min_32 -static inline opus_int32 silk_min_32(opus_int32 a, opus_int32 b) +static OPUS_INLINE opus_int32 silk_min_32(opus_int32 a, opus_int32 b) { ops_count += 1; return (((a) < (b)) ? (a) : (b)); } #undef silk_min_64 -static inline opus_int64 silk_min_64(opus_int64 a, opus_int64 b) +static OPUS_INLINE opus_int64 silk_min_64(opus_int64 a, opus_int64 b) { ops_count += 1; return (((a) < (b)) ? (a) : (b)); @@ -647,26 +647,26 @@ static inline opus_int64 silk_min_64(opus_int64 a, opus_int64 b) /* silk_min() versions with typecast in the function call */ #undef silk_max_int -static inline opus_int silk_max_int(opus_int a, opus_int b) +static OPUS_INLINE opus_int silk_max_int(opus_int a, opus_int b) { ops_count += 1; return (((a) > (b)) ? (a) : (b)); } #undef silk_max_16 -static inline opus_int16 silk_max_16(opus_int16 a, opus_int16 b) +static OPUS_INLINE opus_int16 silk_max_16(opus_int16 a, opus_int16 b) { ops_count += 1; return (((a) > (b)) ? (a) : (b)); } #undef silk_max_32 -static inline opus_int32 silk_max_32(opus_int32 a, opus_int32 b) +static OPUS_INLINE opus_int32 silk_max_32(opus_int32 a, opus_int32 b) { ops_count += 1; return (((a) > (b)) ? (a) : (b)); } #undef silk_max_64 -static inline opus_int64 silk_max_64(opus_int64 a, opus_int64 b) +static OPUS_INLINE opus_int64 silk_max_64(opus_int64 a, opus_int64 b) { ops_count += 1; return (((a) > (b)) ? (a) : (b)); @@ -674,7 +674,7 @@ static inline opus_int64 silk_max_64(opus_int64 a, opus_int64 b) #undef silk_LIMIT_int -static inline opus_int silk_LIMIT_int(opus_int a, opus_int limit1, opus_int limit2) +static OPUS_INLINE opus_int silk_LIMIT_int(opus_int a, opus_int limit1, opus_int limit2) { opus_int ret; ops_count += 6; @@ -686,7 +686,7 @@ static inline opus_int silk_LIMIT_int(opus_int a, opus_int limit1, opus_int limi } #undef silk_LIMIT_16 -static inline opus_int16 silk_LIMIT_16(opus_int16 a, opus_int16 limit1, opus_int16 limit2) +static OPUS_INLINE opus_int16 silk_LIMIT_16(opus_int16 a, opus_int16 limit1, opus_int16 limit2) { opus_int16 ret; ops_count += 6; @@ -699,7 +699,7 @@ return(ret); #undef silk_LIMIT_32 -static inline opus_int silk_LIMIT_32(opus_int32 a, opus_int32 limit1, opus_int32 limit2) +static OPUS_INLINE opus_int silk_LIMIT_32(opus_int32 a, opus_int32 limit1, opus_int32 limit2) { opus_int32 ret; ops_count += 6; diff --git a/code/opus-1.0.2/silk/MacroDebug.h b/code/opus-1.1/silk/MacroDebug.h similarity index 84% rename from code/opus-1.0.2/silk/MacroDebug.h rename to code/opus-1.1/silk/MacroDebug.h index ecd90bc4..35aedc5c 100644 --- a/code/opus-1.0.2/silk/MacroDebug.h +++ b/code/opus-1.1/silk/MacroDebug.h @@ -9,11 +9,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -36,7 +36,7 @@ POSSIBILITY OF SUCH DAMAGE. #undef silk_ADD16 #define silk_ADD16(a,b) silk_ADD16_((a), (b), __FILE__, __LINE__) -static inline opus_int16 silk_ADD16_(opus_int16 a, opus_int16 b, char *file, int line){ +static OPUS_INLINE opus_int16 silk_ADD16_(opus_int16 a, opus_int16 b, char *file, int line){ opus_int16 ret; ret = a + b; @@ -52,7 +52,7 @@ static inline opus_int16 silk_ADD16_(opus_int16 a, opus_int16 b, char *file, int #undef silk_ADD32 #define silk_ADD32(a,b) silk_ADD32_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_ADD32_(opus_int32 a, opus_int32 b, char *file, int line){ +static OPUS_INLINE opus_int32 silk_ADD32_(opus_int32 a, opus_int32 b, char *file, int line){ opus_int32 ret; ret = a + b; @@ -68,7 +68,7 @@ static inline opus_int32 silk_ADD32_(opus_int32 a, opus_int32 b, char *file, int #undef silk_ADD64 #define silk_ADD64(a,b) silk_ADD64_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_ADD64_(opus_int64 a, opus_int64 b, char *file, int line){ +static OPUS_INLINE opus_int64 silk_ADD64_(opus_int64 a, opus_int64 b, char *file, int line){ opus_int64 ret; ret = a + b; @@ -84,7 +84,7 @@ static inline opus_int64 silk_ADD64_(opus_int64 a, opus_int64 b, char *file, int #undef silk_SUB16 #define silk_SUB16(a,b) silk_SUB16_((a), (b), __FILE__, __LINE__) -static inline opus_int16 silk_SUB16_(opus_int16 a, opus_int16 b, char *file, int line){ +static OPUS_INLINE opus_int16 silk_SUB16_(opus_int16 a, opus_int16 b, char *file, int line){ opus_int16 ret; ret = a - b; @@ -100,7 +100,7 @@ static inline opus_int16 silk_SUB16_(opus_int16 a, opus_int16 b, char *file, int #undef silk_SUB32 #define silk_SUB32(a,b) silk_SUB32_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_SUB32_(opus_int32 a, opus_int32 b, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SUB32_(opus_int32 a, opus_int32 b, char *file, int line){ opus_int32 ret; ret = a - b; @@ -116,7 +116,7 @@ static inline opus_int32 silk_SUB32_(opus_int32 a, opus_int32 b, char *file, int #undef silk_SUB64 #define silk_SUB64(a,b) silk_SUB64_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_SUB64_(opus_int64 a, opus_int64 b, char *file, int line){ +static OPUS_INLINE opus_int64 silk_SUB64_(opus_int64 a, opus_int64 b, char *file, int line){ opus_int64 ret; ret = a - b; @@ -132,7 +132,7 @@ static inline opus_int64 silk_SUB64_(opus_int64 a, opus_int64 b, char *file, int #undef silk_ADD_SAT16 #define silk_ADD_SAT16(a,b) silk_ADD_SAT16_((a), (b), __FILE__, __LINE__) -static inline opus_int16 silk_ADD_SAT16_( opus_int16 a16, opus_int16 b16, char *file, int line) { +static OPUS_INLINE opus_int16 silk_ADD_SAT16_( opus_int16 a16, opus_int16 b16, char *file, int line) { opus_int16 res; res = (opus_int16)silk_SAT16( silk_ADD32( (opus_int32)(a16), (b16) ) ); if ( res != silk_SAT16( (opus_int32)a16 + (opus_int32)b16 ) ) @@ -147,7 +147,7 @@ static inline opus_int16 silk_ADD_SAT16_( opus_int16 a16, opus_int16 b16, char * #undef silk_ADD_SAT32 #define silk_ADD_SAT32(a,b) silk_ADD_SAT32_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_ADD_SAT32_(opus_int32 a32, opus_int32 b32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_ADD_SAT32_(opus_int32 a32, opus_int32 b32, char *file, int line){ opus_int32 res; res = ((((opus_uint32)(a32) + (opus_uint32)(b32)) & 0x80000000) == 0 ? \ ((((a32) & (b32)) & 0x80000000) != 0 ? silk_int32_MIN : (a32)+(b32)) : \ @@ -164,7 +164,7 @@ static inline opus_int32 silk_ADD_SAT32_(opus_int32 a32, opus_int32 b32, char *f #undef silk_ADD_SAT64 #define silk_ADD_SAT64(a,b) silk_ADD_SAT64_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_ADD_SAT64_( opus_int64 a64, opus_int64 b64, char *file, int line) { +static OPUS_INLINE opus_int64 silk_ADD_SAT64_( opus_int64 a64, opus_int64 b64, char *file, int line) { opus_int64 res; int fail = 0; res = ((((a64) + (b64)) & 0x8000000000000000LL) == 0 ? \ @@ -193,7 +193,7 @@ static inline opus_int64 silk_ADD_SAT64_( opus_int64 a64, opus_int64 b64, char * #undef silk_SUB_SAT16 #define silk_SUB_SAT16(a,b) silk_SUB_SAT16_((a), (b), __FILE__, __LINE__) -static inline opus_int16 silk_SUB_SAT16_( opus_int16 a16, opus_int16 b16, char *file, int line ) { +static OPUS_INLINE opus_int16 silk_SUB_SAT16_( opus_int16 a16, opus_int16 b16, char *file, int line ) { opus_int16 res; res = (opus_int16)silk_SAT16( silk_SUB32( (opus_int32)(a16), (b16) ) ); if ( res != silk_SAT16( (opus_int32)a16 - (opus_int32)b16 ) ) @@ -208,7 +208,7 @@ static inline opus_int16 silk_SUB_SAT16_( opus_int16 a16, opus_int16 b16, char * #undef silk_SUB_SAT32 #define silk_SUB_SAT32(a,b) silk_SUB_SAT32_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_SUB_SAT32_( opus_int32 a32, opus_int32 b32, char *file, int line ) { +static OPUS_INLINE opus_int32 silk_SUB_SAT32_( opus_int32 a32, opus_int32 b32, char *file, int line ) { opus_int32 res; res = ((((opus_uint32)(a32)-(opus_uint32)(b32)) & 0x80000000) == 0 ? \ (( (a32) & ((b32)^0x80000000) & 0x80000000) ? silk_int32_MIN : (a32)-(b32)) : \ @@ -225,7 +225,7 @@ static inline opus_int32 silk_SUB_SAT32_( opus_int32 a32, opus_int32 b32, char * #undef silk_SUB_SAT64 #define silk_SUB_SAT64(a,b) silk_SUB_SAT64_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_SUB_SAT64_( opus_int64 a64, opus_int64 b64, char *file, int line ) { +static OPUS_INLINE opus_int64 silk_SUB_SAT64_( opus_int64 a64, opus_int64 b64, char *file, int line ) { opus_int64 res; int fail = 0; res = ((((a64)-(b64)) & 0x8000000000000000LL) == 0 ? \ @@ -254,7 +254,7 @@ static inline opus_int64 silk_SUB_SAT64_( opus_int64 a64, opus_int64 b64, char * #undef silk_MUL #define silk_MUL(a,b) silk_MUL_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_MUL_(opus_int32 a32, opus_int32 b32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_MUL_(opus_int32 a32, opus_int32 b32, char *file, int line){ opus_int32 ret; opus_int64 ret64; ret = a32 * b32; @@ -271,7 +271,7 @@ static inline opus_int32 silk_MUL_(opus_int32 a32, opus_int32 b32, char *file, i #undef silk_MUL_uint #define silk_MUL_uint(a,b) silk_MUL_uint_((a), (b), __FILE__, __LINE__) -static inline opus_uint32 silk_MUL_uint_(opus_uint32 a32, opus_uint32 b32, char *file, int line){ +static OPUS_INLINE opus_uint32 silk_MUL_uint_(opus_uint32 a32, opus_uint32 b32, char *file, int line){ opus_uint32 ret; ret = a32 * b32; if ( (opus_uint64)ret != (opus_uint64)a32 * (opus_uint64)b32 ) @@ -286,7 +286,7 @@ static inline opus_uint32 silk_MUL_uint_(opus_uint32 a32, opus_uint32 b32, char #undef silk_MLA #define silk_MLA(a,b,c) silk_MLA_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_MLA_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_MLA_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ opus_int32 ret; ret = a32 + b32 * c32; if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (opus_int64)c32 ) @@ -301,7 +301,7 @@ static inline opus_int32 silk_MLA_(opus_int32 a32, opus_int32 b32, opus_int32 c3 #undef silk_MLA_uint #define silk_MLA_uint(a,b,c) silk_MLA_uint_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_MLA_uint_(opus_uint32 a32, opus_uint32 b32, opus_uint32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_MLA_uint_(opus_uint32 a32, opus_uint32 b32, opus_uint32 c32, char *file, int line){ opus_uint32 ret; ret = a32 + b32 * c32; if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (opus_int64)c32 ) @@ -316,7 +316,7 @@ static inline opus_int32 silk_MLA_uint_(opus_uint32 a32, opus_uint32 b32, opus_u #undef silk_SMULWB #define silk_SMULWB(a,b) silk_SMULWB_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_SMULWB_(opus_int32 a32, opus_int32 b32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMULWB_(opus_int32 a32, opus_int32 b32, char *file, int line){ opus_int32 ret; ret = (a32 >> 16) * (opus_int32)((opus_int16)b32) + (((a32 & 0x0000FFFF) * (opus_int32)((opus_int16)b32)) >> 16); if ( (opus_int64)ret != ((opus_int64)a32 * (opus_int16)b32) >> 16 ) @@ -331,7 +331,7 @@ static inline opus_int32 silk_SMULWB_(opus_int32 a32, opus_int32 b32, char *file #undef silk_SMLAWB #define silk_SMLAWB(a,b,c) silk_SMLAWB_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SMLAWB_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMLAWB_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ opus_int32 ret; ret = silk_ADD32( a32, silk_SMULWB( b32, c32 ) ); if ( silk_ADD32( a32, silk_SMULWB( b32, c32 ) ) != silk_ADD_SAT32( a32, silk_SMULWB( b32, c32 ) ) ) @@ -346,7 +346,7 @@ static inline opus_int32 silk_SMLAWB_(opus_int32 a32, opus_int32 b32, opus_int32 #undef silk_SMULWT #define silk_SMULWT(a,b) silk_SMULWT_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_SMULWT_(opus_int32 a32, opus_int32 b32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMULWT_(opus_int32 a32, opus_int32 b32, char *file, int line){ opus_int32 ret; ret = (a32 >> 16) * (b32 >> 16) + (((a32 & 0x0000FFFF) * (b32 >> 16)) >> 16); if ( (opus_int64)ret != ((opus_int64)a32 * (b32 >> 16)) >> 16 ) @@ -361,7 +361,7 @@ static inline opus_int32 silk_SMULWT_(opus_int32 a32, opus_int32 b32, char *file #undef silk_SMLAWT #define silk_SMLAWT(a,b,c) silk_SMLAWT_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SMLAWT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMLAWT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ opus_int32 ret; ret = a32 + ((b32 >> 16) * (c32 >> 16)) + (((b32 & 0x0000FFFF) * ((c32 >> 16)) >> 16)); if ( (opus_int64)ret != (opus_int64)a32 + (((opus_int64)b32 * (c32 >> 16)) >> 16) ) @@ -376,7 +376,7 @@ static inline opus_int32 silk_SMLAWT_(opus_int32 a32, opus_int32 b32, opus_int32 #undef silk_SMULL #define silk_SMULL(a,b) silk_SMULL_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_SMULL_(opus_int64 a64, opus_int64 b64, char *file, int line){ +static OPUS_INLINE opus_int64 silk_SMULL_(opus_int64 a64, opus_int64 b64, char *file, int line){ opus_int64 ret64; int fail = 0; ret64 = a64 * b64; @@ -398,7 +398,7 @@ static inline opus_int64 silk_SMULL_(opus_int64 a64, opus_int64 b64, char *file, /* no checking needed for silk_SMULBB */ #undef silk_SMLABB #define silk_SMLABB(a,b,c) silk_SMLABB_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SMLABB_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMLABB_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ opus_int32 ret; ret = a32 + (opus_int32)((opus_int16)b32) * (opus_int32)((opus_int16)c32); if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (opus_int16)c32 ) @@ -414,7 +414,7 @@ static inline opus_int32 silk_SMLABB_(opus_int32 a32, opus_int32 b32, opus_int32 /* no checking needed for silk_SMULBT */ #undef silk_SMLABT #define silk_SMLABT(a,b,c) silk_SMLABT_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SMLABT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMLABT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ opus_int32 ret; ret = a32 + ((opus_int32)((opus_int16)b32)) * (c32 >> 16); if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (c32 >> 16) ) @@ -430,7 +430,7 @@ static inline opus_int32 silk_SMLABT_(opus_int32 a32, opus_int32 b32, opus_int32 /* no checking needed for silk_SMULTT */ #undef silk_SMLATT #define silk_SMLATT(a,b,c) silk_SMLATT_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SMLATT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMLATT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ opus_int32 ret; ret = a32 + (b32 >> 16) * (c32 >> 16); if ( (opus_int64)ret != (opus_int64)a32 + (b32 >> 16) * (c32 >> 16) ) @@ -445,7 +445,7 @@ static inline opus_int32 silk_SMLATT_(opus_int32 a32, opus_int32 b32, opus_int32 #undef silk_SMULWW #define silk_SMULWW(a,b) silk_SMULWW_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_SMULWW_(opus_int32 a32, opus_int32 b32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMULWW_(opus_int32 a32, opus_int32 b32, char *file, int line){ opus_int32 ret, tmp1, tmp2; opus_int64 ret64; int fail = 0; @@ -476,7 +476,7 @@ static inline opus_int32 silk_SMULWW_(opus_int32 a32, opus_int32 b32, char *file #undef silk_SMLAWW #define silk_SMLAWW(a,b,c) silk_SMLAWW_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SMLAWW_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SMLAWW_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ opus_int32 ret, tmp; tmp = silk_SMULWW( b32, c32 ); @@ -505,7 +505,7 @@ static inline opus_int32 silk_SMLAWW_(opus_int32 a32, opus_int32 b32, opus_int32 #undef silk_DIV32 #define silk_DIV32(a,b) silk_DIV32_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_DIV32_(opus_int32 a32, opus_int32 b32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_DIV32_(opus_int32 a32, opus_int32 b32, char *file, int line){ if ( b32 == 0 ) { fprintf (stderr, "silk_DIV32(%d, %d) in %s: line %d\n", a32, b32, file, line); @@ -518,7 +518,7 @@ static inline opus_int32 silk_DIV32_(opus_int32 a32, opus_int32 b32, char *file, #undef silk_DIV32_16 #define silk_DIV32_16(a,b) silk_DIV32_16_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_DIV32_16_(opus_int32 a32, opus_int32 b32, char *file, int line){ +static OPUS_INLINE opus_int32 silk_DIV32_16_(opus_int32 a32, opus_int32 b32, char *file, int line){ int fail = 0; fail |= b32 == 0; fail |= b32 > silk_int16_MAX; @@ -544,7 +544,7 @@ static inline opus_int32 silk_DIV32_16_(opus_int32 a32, opus_int32 b32, char *fi #undef silk_LSHIFT8 #define silk_LSHIFT8(a,b) silk_LSHIFT8_((a), (b), __FILE__, __LINE__) -static inline opus_int8 silk_LSHIFT8_(opus_int8 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int8 silk_LSHIFT8_(opus_int8 a, opus_int32 shift, char *file, int line){ opus_int8 ret; int fail = 0; ret = a << shift; @@ -563,7 +563,7 @@ static inline opus_int8 silk_LSHIFT8_(opus_int8 a, opus_int32 shift, char *file, #undef silk_LSHIFT16 #define silk_LSHIFT16(a,b) silk_LSHIFT16_((a), (b), __FILE__, __LINE__) -static inline opus_int16 silk_LSHIFT16_(opus_int16 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int16 silk_LSHIFT16_(opus_int16 a, opus_int32 shift, char *file, int line){ opus_int16 ret; int fail = 0; ret = a << shift; @@ -582,7 +582,7 @@ static inline opus_int16 silk_LSHIFT16_(opus_int16 a, opus_int32 shift, char *fi #undef silk_LSHIFT32 #define silk_LSHIFT32(a,b) silk_LSHIFT32_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_LSHIFT32_(opus_int32 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_LSHIFT32_(opus_int32 a, opus_int32 shift, char *file, int line){ opus_int32 ret; int fail = 0; ret = a << shift; @@ -601,7 +601,7 @@ static inline opus_int32 silk_LSHIFT32_(opus_int32 a, opus_int32 shift, char *fi #undef silk_LSHIFT64 #define silk_LSHIFT64(a,b) silk_LSHIFT64_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_LSHIFT64_(opus_int64 a, opus_int shift, char *file, int line){ +static OPUS_INLINE opus_int64 silk_LSHIFT64_(opus_int64 a, opus_int shift, char *file, int line){ opus_int64 ret; int fail = 0; ret = a << shift; @@ -620,7 +620,7 @@ static inline opus_int64 silk_LSHIFT64_(opus_int64 a, opus_int shift, char *file #undef silk_LSHIFT_ovflw #define silk_LSHIFT_ovflw(a,b) silk_LSHIFT_ovflw_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_LSHIFT_ovflw_(opus_int32 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_LSHIFT_ovflw_(opus_int32 a, opus_int32 shift, char *file, int line){ if ( (shift < 0) || (shift >= 32) ) /* no check for overflow */ { fprintf (stderr, "silk_LSHIFT_ovflw(%d, %d) in %s: line %d\n", a, shift, file, line); @@ -633,7 +633,7 @@ static inline opus_int32 silk_LSHIFT_ovflw_(opus_int32 a, opus_int32 shift, char #undef silk_LSHIFT_uint #define silk_LSHIFT_uint(a,b) silk_LSHIFT_uint_((a), (b), __FILE__, __LINE__) -static inline opus_uint32 silk_LSHIFT_uint_(opus_uint32 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_uint32 silk_LSHIFT_uint_(opus_uint32 a, opus_int32 shift, char *file, int line){ opus_uint32 ret; ret = a << shift; if ( (shift < 0) || ((opus_int64)ret != ((opus_int64)a) << shift)) @@ -648,7 +648,7 @@ static inline opus_uint32 silk_LSHIFT_uint_(opus_uint32 a, opus_int32 shift, cha #undef silk_RSHIFT8 #define silk_RSHITF8(a,b) silk_RSHIFT8_((a), (b), __FILE__, __LINE__) -static inline opus_int8 silk_RSHIFT8_(opus_int8 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int8 silk_RSHIFT8_(opus_int8 a, opus_int32 shift, char *file, int line){ if ( (shift < 0) || (shift>=8) ) { fprintf (stderr, "silk_RSHITF8(%d, %d) in %s: line %d\n", a, shift, file, line); @@ -661,7 +661,7 @@ static inline opus_int8 silk_RSHIFT8_(opus_int8 a, opus_int32 shift, char *file, #undef silk_RSHIFT16 #define silk_RSHITF16(a,b) silk_RSHIFT16_((a), (b), __FILE__, __LINE__) -static inline opus_int16 silk_RSHIFT16_(opus_int16 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int16 silk_RSHIFT16_(opus_int16 a, opus_int32 shift, char *file, int line){ if ( (shift < 0) || (shift>=16) ) { fprintf (stderr, "silk_RSHITF16(%d, %d) in %s: line %d\n", a, shift, file, line); @@ -674,7 +674,7 @@ static inline opus_int16 silk_RSHIFT16_(opus_int16 a, opus_int32 shift, char *fi #undef silk_RSHIFT32 #define silk_RSHIFT32(a,b) silk_RSHIFT32_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_RSHIFT32_(opus_int32 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_RSHIFT32_(opus_int32 a, opus_int32 shift, char *file, int line){ if ( (shift < 0) || (shift>=32) ) { fprintf (stderr, "silk_RSHITF32(%d, %d) in %s: line %d\n", a, shift, file, line); @@ -687,7 +687,7 @@ static inline opus_int32 silk_RSHIFT32_(opus_int32 a, opus_int32 shift, char *fi #undef silk_RSHIFT64 #define silk_RSHIFT64(a,b) silk_RSHIFT64_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_RSHIFT64_(opus_int64 a, opus_int64 shift, char *file, int line){ +static OPUS_INLINE opus_int64 silk_RSHIFT64_(opus_int64 a, opus_int64 shift, char *file, int line){ if ( (shift < 0) || (shift>=64) ) { fprintf (stderr, "silk_RSHITF64(%lld, %lld) in %s: line %d\n", (long long)a, (long long)shift, file, line); @@ -700,7 +700,7 @@ static inline opus_int64 silk_RSHIFT64_(opus_int64 a, opus_int64 shift, char *fi #undef silk_RSHIFT_uint #define silk_RSHIFT_uint(a,b) silk_RSHIFT_uint_((a), (b), __FILE__, __LINE__) -static inline opus_uint32 silk_RSHIFT_uint_(opus_uint32 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_uint32 silk_RSHIFT_uint_(opus_uint32 a, opus_int32 shift, char *file, int line){ if ( (shift < 0) || (shift>32) ) { fprintf (stderr, "silk_RSHIFT_uint(%u, %d) in %s: line %d\n", a, shift, file, line); @@ -713,7 +713,7 @@ static inline opus_uint32 silk_RSHIFT_uint_(opus_uint32 a, opus_int32 shift, cha #undef silk_ADD_LSHIFT #define silk_ADD_LSHIFT(a,b,c) silk_ADD_LSHIFT_((a), (b), (c), __FILE__, __LINE__) -static inline int silk_ADD_LSHIFT_(int a, int b, int shift, char *file, int line){ +static OPUS_INLINE int silk_ADD_LSHIFT_(int a, int b, int shift, char *file, int line){ opus_int16 ret; ret = a + (b << shift); if ( (shift < 0) || (shift>15) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) << shift)) ) @@ -728,7 +728,7 @@ static inline int silk_ADD_LSHIFT_(int a, int b, int shift, char *file, int line #undef silk_ADD_LSHIFT32 #define silk_ADD_LSHIFT32(a,b,c) silk_ADD_LSHIFT32_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_ADD_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_ADD_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ opus_int32 ret; ret = a + (b << shift); if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) << shift)) ) @@ -743,7 +743,7 @@ static inline opus_int32 silk_ADD_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int #undef silk_ADD_LSHIFT_uint #define silk_ADD_LSHIFT_uint(a,b,c) silk_ADD_LSHIFT_uint_((a), (b), (c), __FILE__, __LINE__) -static inline opus_uint32 silk_ADD_LSHIFT_uint_(opus_uint32 a, opus_uint32 b, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_uint32 silk_ADD_LSHIFT_uint_(opus_uint32 a, opus_uint32 b, opus_int32 shift, char *file, int line){ opus_uint32 ret; ret = a + (b << shift); if ( (shift < 0) || (shift>32) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) << shift)) ) @@ -758,7 +758,7 @@ static inline opus_uint32 silk_ADD_LSHIFT_uint_(opus_uint32 a, opus_uint32 b, op #undef silk_ADD_RSHIFT #define silk_ADD_RSHIFT(a,b,c) silk_ADD_RSHIFT_((a), (b), (c), __FILE__, __LINE__) -static inline int silk_ADD_RSHIFT_(int a, int b, int shift, char *file, int line){ +static OPUS_INLINE int silk_ADD_RSHIFT_(int a, int b, int shift, char *file, int line){ opus_int16 ret; ret = a + (b >> shift); if ( (shift < 0) || (shift>15) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) >> shift)) ) @@ -773,7 +773,7 @@ static inline int silk_ADD_RSHIFT_(int a, int b, int shift, char *file, int line #undef silk_ADD_RSHIFT32 #define silk_ADD_RSHIFT32(a,b,c) silk_ADD_RSHIFT32_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_ADD_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_ADD_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ opus_int32 ret; ret = a + (b >> shift); if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) >> shift)) ) @@ -788,7 +788,7 @@ static inline opus_int32 silk_ADD_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int #undef silk_ADD_RSHIFT_uint #define silk_ADD_RSHIFT_uint(a,b,c) silk_ADD_RSHIFT_uint_((a), (b), (c), __FILE__, __LINE__) -static inline opus_uint32 silk_ADD_RSHIFT_uint_(opus_uint32 a, opus_uint32 b, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_uint32 silk_ADD_RSHIFT_uint_(opus_uint32 a, opus_uint32 b, opus_int32 shift, char *file, int line){ opus_uint32 ret; ret = a + (b >> shift); if ( (shift < 0) || (shift>32) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) >> shift)) ) @@ -803,7 +803,7 @@ static inline opus_uint32 silk_ADD_RSHIFT_uint_(opus_uint32 a, opus_uint32 b, op #undef silk_SUB_LSHIFT32 #define silk_SUB_LSHIFT32(a,b,c) silk_SUB_LSHIFT32_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SUB_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SUB_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ opus_int32 ret; ret = a - (b << shift); if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a - (((opus_int64)b) << shift)) ) @@ -818,7 +818,7 @@ static inline opus_int32 silk_SUB_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int #undef silk_SUB_RSHIFT32 #define silk_SUB_RSHIFT32(a,b,c) silk_SUB_RSHIFT32_((a), (b), (c), __FILE__, __LINE__) -static inline opus_int32 silk_SUB_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_SUB_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ opus_int32 ret; ret = a - (b >> shift); if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a - (((opus_int64)b) >> shift)) ) @@ -833,7 +833,7 @@ static inline opus_int32 silk_SUB_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int #undef silk_RSHIFT_ROUND #define silk_RSHIFT_ROUND(a,b) silk_RSHIFT_ROUND_((a), (b), __FILE__, __LINE__) -static inline opus_int32 silk_RSHIFT_ROUND_(opus_int32 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int32 silk_RSHIFT_ROUND_(opus_int32 a, opus_int32 shift, char *file, int line){ opus_int32 ret; ret = shift == 1 ? (a >> 1) + (a & 1) : ((a >> (shift - 1)) + 1) >> 1; /* the marco definition can't handle a shift of zero */ @@ -849,7 +849,7 @@ static inline opus_int32 silk_RSHIFT_ROUND_(opus_int32 a, opus_int32 shift, char #undef silk_RSHIFT_ROUND64 #define silk_RSHIFT_ROUND64(a,b) silk_RSHIFT_ROUND64_((a), (b), __FILE__, __LINE__) -static inline opus_int64 silk_RSHIFT_ROUND64_(opus_int64 a, opus_int32 shift, char *file, int line){ +static OPUS_INLINE opus_int64 silk_RSHIFT_ROUND64_(opus_int64 a, opus_int32 shift, char *file, int line){ opus_int64 ret; /* the marco definition can't handle a shift of zero */ if ( (shift <= 0) || (shift>=64) ) @@ -865,14 +865,14 @@ static inline opus_int64 silk_RSHIFT_ROUND64_(opus_int64 a, opus_int32 shift, ch /* silk_abs is used on floats also, so doesn't work... */ /*#undef silk_abs -static inline opus_int32 silk_abs(opus_int32 a){ +static OPUS_INLINE opus_int32 silk_abs(opus_int32 a){ silk_assert(a != 0x80000000); return (((a) > 0) ? (a) : -(a)); // Be careful, silk_abs returns wrong when input equals to silk_intXX_MIN }*/ #undef silk_abs_int64 #define silk_abs_int64(a) silk_abs_int64_((a), __FILE__, __LINE__) -static inline opus_int64 silk_abs_int64_(opus_int64 a, char *file, int line){ +static OPUS_INLINE opus_int64 silk_abs_int64_(opus_int64 a, char *file, int line){ if ( a == silk_int64_MIN ) { fprintf (stderr, "silk_abs_int64(%lld) in %s: line %d\n", (long long)a, file, line); @@ -885,7 +885,7 @@ static inline opus_int64 silk_abs_int64_(opus_int64 a, char *file, int line){ #undef silk_abs_int32 #define silk_abs_int32(a) silk_abs_int32_((a), __FILE__, __LINE__) -static inline opus_int32 silk_abs_int32_(opus_int32 a, char *file, int line){ +static OPUS_INLINE opus_int32 silk_abs_int32_(opus_int32 a, char *file, int line){ if ( a == silk_int32_MIN ) { fprintf (stderr, "silk_abs_int32(%d) in %s: line %d\n", a, file, line); @@ -898,7 +898,7 @@ static inline opus_int32 silk_abs_int32_(opus_int32 a, char *file, int line){ #undef silk_CHECK_FIT8 #define silk_CHECK_FIT8(a) silk_CHECK_FIT8_((a), __FILE__, __LINE__) -static inline opus_int8 silk_CHECK_FIT8_( opus_int64 a, char *file, int line ){ +static OPUS_INLINE opus_int8 silk_CHECK_FIT8_( opus_int64 a, char *file, int line ){ opus_int8 ret; ret = (opus_int8)a; if ( (opus_int64)ret != a ) @@ -913,7 +913,7 @@ static inline opus_int8 silk_CHECK_FIT8_( opus_int64 a, char *file, int line ){ #undef silk_CHECK_FIT16 #define silk_CHECK_FIT16(a) silk_CHECK_FIT16_((a), __FILE__, __LINE__) -static inline opus_int16 silk_CHECK_FIT16_( opus_int64 a, char *file, int line ){ +static OPUS_INLINE opus_int16 silk_CHECK_FIT16_( opus_int64 a, char *file, int line ){ opus_int16 ret; ret = (opus_int16)a; if ( (opus_int64)ret != a ) @@ -928,7 +928,7 @@ static inline opus_int16 silk_CHECK_FIT16_( opus_int64 a, char *file, int line ) #undef silk_CHECK_FIT32 #define silk_CHECK_FIT32(a) silk_CHECK_FIT32_((a), __FILE__, __LINE__) -static inline opus_int32 silk_CHECK_FIT32_( opus_int64 a, char *file, int line ){ +static OPUS_INLINE opus_int32 silk_CHECK_FIT32_( opus_int64 a, char *file, int line ){ opus_int32 ret; ret = (opus_int32)a; if ( (opus_int64)ret != a ) diff --git a/code/opus-1.0.2/silk/NLSF2A.c b/code/opus-1.1/silk/NLSF2A.c similarity index 99% rename from code/opus-1.0.2/silk/NLSF2A.c rename to code/opus-1.1/silk/NLSF2A.c index 10b66b64..b1c559ea 100644 --- a/code/opus-1.0.2/silk/NLSF2A.c +++ b/code/opus-1.1/silk/NLSF2A.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -41,7 +41,7 @@ POSSIBILITY OF SUCH DAMAGE. #define QA 16 /* helper function for NLSF2A(..) */ -static inline void silk_NLSF2A_find_poly( +static OPUS_INLINE void silk_NLSF2A_find_poly( opus_int32 *out, /* O intermediate polynomial, QA [dd+1] */ const opus_int32 *cLSF, /* I vector of interleaved 2*cos(LSFs), QA [d] */ opus_int dd /* I polynomial order (= 1/2 * filter order) */ diff --git a/code/opus-1.0.2/silk/NLSF_VQ.c b/code/opus-1.1/silk/NLSF_VQ.c similarity index 99% rename from code/opus-1.0.2/silk/NLSF_VQ.c rename to code/opus-1.1/silk/NLSF_VQ.c index 352dda26..69b6e22e 100644 --- a/code/opus-1.0.2/silk/NLSF_VQ.c +++ b/code/opus-1.1/silk/NLSF_VQ.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/NLSF_VQ_weights_laroia.c b/code/opus-1.1/silk/NLSF_VQ_weights_laroia.c similarity index 99% rename from code/opus-1.0.2/silk/NLSF_VQ_weights_laroia.c rename to code/opus-1.1/silk/NLSF_VQ_weights_laroia.c index 05bb17af..04894c59 100644 --- a/code/opus-1.0.2/silk/NLSF_VQ_weights_laroia.c +++ b/code/opus-1.1/silk/NLSF_VQ_weights_laroia.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/NLSF_decode.c b/code/opus-1.1/silk/NLSF_decode.c similarity index 97% rename from code/opus-1.0.2/silk/NLSF_decode.c rename to code/opus-1.1/silk/NLSF_decode.c index e007c49a..9f715060 100644 --- a/code/opus-1.0.2/silk/NLSF_decode.c +++ b/code/opus-1.1/silk/NLSF_decode.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -32,7 +32,7 @@ POSSIBILITY OF SUCH DAMAGE. #include "main.h" /* Predictive dequantizer for NLSF residuals */ -static inline void silk_NLSF_residual_dequant( /* O Returns RD value in Q30 */ +static OPUS_INLINE void silk_NLSF_residual_dequant( /* O Returns RD value in Q30 */ opus_int16 x_Q10[], /* O Output [ order ] */ const opus_int8 indices[], /* I Quantization indices [ order ] */ const opus_uint8 pred_coef_Q8[], /* I Backward predictor coefs [ order ] */ diff --git a/code/opus-1.0.2/silk/NLSF_del_dec_quant.c b/code/opus-1.1/silk/NLSF_del_dec_quant.c similarity index 99% rename from code/opus-1.0.2/silk/NLSF_del_dec_quant.c rename to code/opus-1.1/silk/NLSF_del_dec_quant.c index 78870de5..504dbbd0 100644 --- a/code/opus-1.0.2/silk/NLSF_del_dec_quant.c +++ b/code/opus-1.1/silk/NLSF_del_dec_quant.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -121,7 +121,7 @@ opus_int32 silk_NLSF_del_dec_quant( /* O Returns RD_Q25[ j + nStates ] = silk_SMLABB( silk_MLA( RD_tmp_Q25, silk_SMULBB( diff_Q10, diff_Q10 ), w_Q5[ i ] ), mu_Q20, rate1_Q5 ); } - if( nStates < NLSF_QUANT_DEL_DEC_STATES ) { + if( nStates <= ( NLSF_QUANT_DEL_DEC_STATES >> 1 ) ) { /* double number of states and copy */ for( j = 0; j < nStates; j++ ) { ind[ j + nStates ][ i ] = ind[ j ][ i ] + 1; diff --git a/code/opus-1.0.2/silk/NLSF_encode.c b/code/opus-1.1/silk/NLSF_encode.c similarity index 93% rename from code/opus-1.0.2/silk/NLSF_encode.c rename to code/opus-1.1/silk/NLSF_encode.c index 52a263d9..03a036fd 100644 --- a/code/opus-1.0.2/silk/NLSF_encode.c +++ b/code/opus-1.1/silk/NLSF_encode.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "stack_alloc.h" /***********************/ /* NLSF vector encoder */ @@ -46,10 +47,10 @@ opus_int32 silk_NLSF_encode( /* O Returns { opus_int i, s, ind1, bestIndex, prob_Q8, bits_q7; opus_int32 W_tmp_Q9; - opus_int32 err_Q26[ NLSF_VQ_MAX_VECTORS ]; - opus_int32 RD_Q25[ NLSF_VQ_MAX_SURVIVORS ]; - opus_int tempIndices1[ NLSF_VQ_MAX_SURVIVORS ]; - opus_int8 tempIndices2[ NLSF_VQ_MAX_SURVIVORS * MAX_LPC_ORDER ]; + VARDECL( opus_int32, err_Q26 ); + VARDECL( opus_int32, RD_Q25 ); + VARDECL( opus_int, tempIndices1 ); + VARDECL( opus_int8, tempIndices2 ); opus_int16 res_Q15[ MAX_LPC_ORDER ]; opus_int16 res_Q10[ MAX_LPC_ORDER ]; opus_int16 NLSF_tmp_Q15[ MAX_LPC_ORDER ]; @@ -58,6 +59,7 @@ opus_int32 silk_NLSF_encode( /* O Returns opus_uint8 pred_Q8[ MAX_LPC_ORDER ]; opus_int16 ec_ix[ MAX_LPC_ORDER ]; const opus_uint8 *pCB_element, *iCDF_ptr; + SAVE_STACK; silk_assert( nSurvivors <= NLSF_VQ_MAX_SURVIVORS ); silk_assert( signalType >= 0 && signalType <= 2 ); @@ -67,11 +69,16 @@ opus_int32 silk_NLSF_encode( /* O Returns silk_NLSF_stabilize( pNLSF_Q15, psNLSF_CB->deltaMin_Q15, psNLSF_CB->order ); /* First stage: VQ */ + ALLOC( err_Q26, psNLSF_CB->nVectors, opus_int32 ); silk_NLSF_VQ( err_Q26, pNLSF_Q15, psNLSF_CB->CB1_NLSF_Q8, psNLSF_CB->nVectors, psNLSF_CB->order ); /* Sort the quantization errors */ + ALLOC( tempIndices1, nSurvivors, opus_int ); silk_insertion_sort_increasing( err_Q26, tempIndices1, psNLSF_CB->nVectors, nSurvivors ); + ALLOC( RD_Q25, nSurvivors, opus_int32 ); + ALLOC( tempIndices2, nSurvivors * MAX_LPC_ORDER, opus_int8 ); + /* Loop over survivors */ for( s = 0; s < nSurvivors; s++ ) { ind1 = tempIndices1[ s ]; @@ -124,5 +131,6 @@ opus_int32 silk_NLSF_encode( /* O Returns /* Decode */ silk_NLSF_decode( pNLSF_Q15, NLSFIndices, psNLSF_CB ); + RESTORE_STACK; return RD_Q25[ 0 ]; } diff --git a/code/opus-1.0.2/silk/NLSF_stabilize.c b/code/opus-1.1/silk/NLSF_stabilize.c similarity index 99% rename from code/opus-1.0.2/silk/NLSF_stabilize.c rename to code/opus-1.1/silk/NLSF_stabilize.c index 7498b54a..1fa1ea37 100644 --- a/code/opus-1.0.2/silk/NLSF_stabilize.c +++ b/code/opus-1.1/silk/NLSF_stabilize.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/NLSF_unpack.c b/code/opus-1.1/silk/NLSF_unpack.c similarity index 99% rename from code/opus-1.0.2/silk/NLSF_unpack.c rename to code/opus-1.1/silk/NLSF_unpack.c index 47f6cfe8..17bd23f7 100644 --- a/code/opus-1.0.2/silk/NLSF_unpack.c +++ b/code/opus-1.1/silk/NLSF_unpack.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/NSQ.c b/code/opus-1.1/silk/NSQ.c similarity index 97% rename from code/opus-1.0.2/silk/NSQ.c rename to code/opus-1.1/silk/NSQ.c index b49cdf58..cf5b3fd5 100644 --- a/code/opus-1.0.2/silk/NSQ.c +++ b/code/opus-1.1/silk/NSQ.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,8 +30,9 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "stack_alloc.h" -static inline void silk_nsq_scale_states( +static OPUS_INLINE void silk_nsq_scale_states( const silk_encoder_state *psEncC, /* I Encoder State */ silk_nsq_state *NSQ, /* I/O NSQ state */ const opus_int32 x_Q3[], /* I input in Q3 */ @@ -45,7 +46,7 @@ static inline void silk_nsq_scale_states( const opus_int signal_type /* I Signal type */ ); -static inline void silk_noise_shape_quantizer( +static OPUS_INLINE void silk_noise_shape_quantizer( silk_nsq_state *NSQ, /* I/O NSQ state */ opus_int signalType, /* I Signal type */ const opus_int32 x_sc_Q10[], /* I */ @@ -88,11 +89,12 @@ void silk_NSQ( opus_int k, lag, start_idx, LSF_interpolation_flag; const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; opus_int16 *pxq; - opus_int32 sLTP_Q15[ 2 * MAX_FRAME_LENGTH ]; - opus_int16 sLTP[ 2 * MAX_FRAME_LENGTH ]; + VARDECL( opus_int32, sLTP_Q15 ); + VARDECL( opus_int16, sLTP ); opus_int32 HarmShapeFIRPacked_Q14; opus_int offset_Q10; - opus_int32 x_sc_Q10[ MAX_SUB_FRAME_LENGTH ]; + VARDECL( opus_int32, x_sc_Q10 ); + SAVE_STACK; NSQ->rand_seed = psIndices->Seed; @@ -109,6 +111,10 @@ void silk_NSQ( LSF_interpolation_flag = 1; } + ALLOC( sLTP_Q15, + psEncC->ltp_mem_length + psEncC->frame_length, opus_int32 ); + ALLOC( sLTP, psEncC->ltp_mem_length + psEncC->frame_length, opus_int16 ); + ALLOC( x_sc_Q10, psEncC->subfr_length, opus_int32 ); /* Set up pointers to start of sub frame */ NSQ->sLTP_shp_buf_idx = psEncC->ltp_mem_length; NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; @@ -160,12 +166,13 @@ void silk_NSQ( /* DEBUG_STORE_DATA( enc.pcm, &NSQ->xq[ psEncC->ltp_mem_length ], psEncC->frame_length * sizeof( opus_int16 ) ) */ silk_memmove( NSQ->xq, &NSQ->xq[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int16 ) ); silk_memmove( NSQ->sLTP_shp_Q14, &NSQ->sLTP_shp_Q14[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int32 ) ); + RESTORE_STACK; } /***********************************/ /* silk_noise_shape_quantizer */ /***********************************/ -static inline void silk_noise_shape_quantizer( +static OPUS_INLINE void silk_noise_shape_quantizer( silk_nsq_state *NSQ, /* I/O NSQ state */ opus_int signalType, /* I Signal type */ const opus_int32 x_sc_Q10[], /* I */ @@ -363,7 +370,7 @@ static inline void silk_noise_shape_quantizer( silk_memcpy( NSQ->sLPC_Q14, &NSQ->sLPC_Q14[ length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); } -static inline void silk_nsq_scale_states( +static OPUS_INLINE void silk_nsq_scale_states( const silk_encoder_state *psEncC, /* I Encoder State */ silk_nsq_state *NSQ, /* I/O NSQ state */ const opus_int32 x_Q3[], /* I input in Q3 */ diff --git a/code/opus-1.0.2/silk/NSQ_del_dec.c b/code/opus-1.1/silk/NSQ_del_dec.c similarity index 97% rename from code/opus-1.0.2/silk/NSQ_del_dec.c rename to code/opus-1.1/silk/NSQ_del_dec.c index b877fa96..522be406 100644 --- a/code/opus-1.0.2/silk/NSQ_del_dec.c +++ b/code/opus-1.1/silk/NSQ_del_dec.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "stack_alloc.h" typedef struct { opus_int32 sLPC_Q14[ MAX_SUB_FRAME_LENGTH + NSQ_LPC_BUF_LENGTH ]; @@ -54,7 +55,9 @@ typedef struct { opus_int32 LPC_exc_Q14; } NSQ_sample_struct; -static inline void silk_nsq_del_dec_scale_states( +typedef NSQ_sample_struct NSQ_sample_pair[ 2 ]; + +static OPUS_INLINE void silk_nsq_del_dec_scale_states( const silk_encoder_state *psEncC, /* I Encoder State */ silk_nsq_state *NSQ, /* I/O NSQ state */ NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ @@ -74,7 +77,7 @@ static inline void silk_nsq_del_dec_scale_states( /******************************************/ /* Noise shape quantizer for one subframe */ /******************************************/ -static inline void silk_noise_shape_quantizer_del_dec( +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec( silk_nsq_state *NSQ, /* I/O NSQ state */ NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ opus_int signalType, /* I Signal type */ @@ -123,17 +126,18 @@ void silk_NSQ_del_dec( { opus_int i, k, lag, start_idx, LSF_interpolation_flag, Winner_ind, subfr; opus_int last_smple_idx, smpl_buf_idx, decisionDelay; - const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; + const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; opus_int16 *pxq; - opus_int32 sLTP_Q15[ 2 * MAX_FRAME_LENGTH ]; - opus_int16 sLTP[ 2 * MAX_FRAME_LENGTH ]; + VARDECL( opus_int32, sLTP_Q15 ); + VARDECL( opus_int16, sLTP ); opus_int32 HarmShapeFIRPacked_Q14; opus_int offset_Q10; opus_int32 RDmin_Q10, Gain_Q10; - opus_int32 x_sc_Q10[ MAX_SUB_FRAME_LENGTH ]; - opus_int32 delayedGain_Q10[ DECISION_DELAY ]; - NSQ_del_dec_struct psDelDec[ MAX_DEL_DEC_STATES ]; + VARDECL( opus_int32, x_sc_Q10 ); + VARDECL( opus_int32, delayedGain_Q10 ); + VARDECL( NSQ_del_dec_struct, psDelDec ); NSQ_del_dec_struct *psDD; + SAVE_STACK; /* Set unvoiced lag to the previous one, overwrite later for voiced */ lag = NSQ->lagPrev; @@ -141,6 +145,7 @@ void silk_NSQ_del_dec( silk_assert( NSQ->prev_gain_Q16 != 0 ); /* Initialize delayed decision states */ + ALLOC( psDelDec, psEncC->nStatesDelayedDecision, NSQ_del_dec_struct ); silk_memset( psDelDec, 0, psEncC->nStatesDelayedDecision * sizeof( NSQ_del_dec_struct ) ); for( k = 0; k < psEncC->nStatesDelayedDecision; k++ ) { psDD = &psDelDec[ k ]; @@ -175,6 +180,11 @@ void silk_NSQ_del_dec( LSF_interpolation_flag = 1; } + ALLOC( sLTP_Q15, + psEncC->ltp_mem_length + psEncC->frame_length, opus_int32 ); + ALLOC( sLTP, psEncC->ltp_mem_length + psEncC->frame_length, opus_int16 ); + ALLOC( x_sc_Q10, psEncC->subfr_length, opus_int32 ); + ALLOC( delayedGain_Q10, DECISION_DELAY, opus_int32 ); /* Set up pointers to start of sub frame */ pxq = &NSQ->xq[ psEncC->ltp_mem_length ]; NSQ->sLTP_shp_buf_idx = psEncC->ltp_mem_length; @@ -287,12 +297,13 @@ void silk_NSQ_del_dec( /* DEBUG_STORE_DATA( enc.pcm, &NSQ->xq[psEncC->ltp_mem_length], psEncC->frame_length * sizeof( opus_int16 ) ) */ silk_memmove( NSQ->xq, &NSQ->xq[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int16 ) ); silk_memmove( NSQ->sLTP_shp_Q14, &NSQ->sLTP_shp_Q14[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int32 ) ); + RESTORE_STACK; } /******************************************/ /* Noise shape quantizer for one subframe */ /******************************************/ -static inline void silk_noise_shape_quantizer_del_dec( +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec( silk_nsq_state *NSQ, /* I/O NSQ state */ NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ opus_int signalType, /* I Signal type */ @@ -328,11 +339,13 @@ static inline void silk_noise_shape_quantizer_del_dec( opus_int32 q1_Q0, q1_Q10, q2_Q10, exc_Q14, LPC_exc_Q14, xq_Q14, Gain_Q10; opus_int32 tmp1, tmp2, sLF_AR_shp_Q14; opus_int32 *pred_lag_ptr, *shp_lag_ptr, *psLPC_Q14; - NSQ_sample_struct psSampleState[ MAX_DEL_DEC_STATES ][ 2 ]; + VARDECL( NSQ_sample_pair, psSampleState ); NSQ_del_dec_struct *psDD; NSQ_sample_struct *psSS; + SAVE_STACK; silk_assert( nStatesDelayedDecision > 0 ); + ALLOC( psSampleState, nStatesDelayedDecision, NSQ_sample_pair ); shp_lag_ptr = &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - lag + HARM_SHAPE_FIR_TAPS / 2 ]; pred_lag_ptr = &sLTP_Q15[ NSQ->sLTP_buf_idx - lag + LTP_ORDER / 2 ]; @@ -614,9 +627,10 @@ static inline void silk_noise_shape_quantizer_del_dec( psDD = &psDelDec[ k ]; silk_memcpy( psDD->sLPC_Q14, &psDD->sLPC_Q14[ length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); } + RESTORE_STACK; } -static inline void silk_nsq_del_dec_scale_states( +static OPUS_INLINE void silk_nsq_del_dec_scale_states( const silk_encoder_state *psEncC, /* I Encoder State */ silk_nsq_state *NSQ, /* I/O NSQ state */ NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ diff --git a/code/opus-1.0.2/silk/PLC.c b/code/opus-1.1/silk/PLC.c similarity index 99% rename from code/opus-1.0.2/silk/PLC.c rename to code/opus-1.1/silk/PLC.c index 8d547295..01f40014 100644 --- a/code/opus-1.0.2/silk/PLC.c +++ b/code/opus-1.1/silk/PLC.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -38,12 +38,12 @@ static const opus_int16 HARM_ATT_Q15[NB_ATT] = { 32440, 31130 }; /* static const opus_int16 PLC_RAND_ATTENUATE_V_Q15[NB_ATT] = { 31130, 26214 }; /* 0.95, 0.8 */ static const opus_int16 PLC_RAND_ATTENUATE_UV_Q15[NB_ATT] = { 32440, 29491 }; /* 0.99, 0.9 */ -static inline void silk_PLC_update( +static OPUS_INLINE void silk_PLC_update( silk_decoder_state *psDec, /* I/O Decoder state */ silk_decoder_control *psDecCtrl /* I/O Decoder control */ ); -static inline void silk_PLC_conceal( +static OPUS_INLINE void silk_PLC_conceal( silk_decoder_state *psDec, /* I/O Decoder state */ silk_decoder_control *psDecCtrl, /* I/O Decoder control */ opus_int16 frame[] /* O LPC residual signal */ @@ -92,7 +92,7 @@ void silk_PLC( /**************************************************/ /* Update state of PLC */ /**************************************************/ -static inline void silk_PLC_update( +static OPUS_INLINE void silk_PLC_update( silk_decoder_state *psDec, /* I/O Decoder state */ silk_decoder_control *psDecCtrl /* I/O Decoder control */ ) @@ -165,7 +165,7 @@ static inline void silk_PLC_update( psPLC->nb_subfr = psDec->nb_subfr; } -static inline void silk_PLC_conceal( +static OPUS_INLINE void silk_PLC_conceal( silk_decoder_state *psDec, /* I/O Decoder state */ silk_decoder_control *psDecCtrl, /* I/O Decoder control */ opus_int16 frame[] /* O LPC residual signal */ diff --git a/code/opus-1.0.2/silk/PLC.h b/code/opus-1.1/silk/PLC.h similarity index 99% rename from code/opus-1.0.2/silk/PLC.h rename to code/opus-1.1/silk/PLC.h index 1d2d9061..f1e2eccc 100644 --- a/code/opus-1.0.2/silk/PLC.h +++ b/code/opus-1.1/silk/PLC.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/SigProc_FIX.h b/code/opus-1.1/silk/SigProc_FIX.h similarity index 96% rename from code/opus-1.0.2/silk/SigProc_FIX.h rename to code/opus-1.1/silk/SigProc_FIX.h index daa5fd04..1b580579 100644 --- a/code/opus-1.0.2/silk/SigProc_FIX.h +++ b/code/opus-1.1/silk/SigProc_FIX.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -168,12 +168,6 @@ opus_int32 silk_log2lin( const opus_int32 inLog_Q7 /* I input on log scale */ ); -/* Function that returns the maximum absolut value of the input vector */ -opus_int16 silk_int16_array_maxabs( /* O Maximum absolute value, max: 2^15-1 */ - const opus_int16 *vec, /* I Input vector [len] */ - const opus_int32 len /* I Length of input vector */ -); - /* Compute number of bits to right shift the sum of squares of a vector */ /* of int16s to make it fit in an int32 */ void silk_sum_sqr_shift( @@ -233,7 +227,8 @@ void silk_autocorr( opus_int *scale, /* O Scaling of the correlation vector */ const opus_int16 *inputData, /* I Input data to correlate */ const opus_int inputDataSize, /* I Length of input */ - const opus_int correlationCount /* I Number of correlation taps to compute */ + const opus_int correlationCount, /* I Number of correlation taps to compute */ + int arch /* I Run-time architecture */ ); void silk_decode_pitch( @@ -252,10 +247,11 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 opus_int *LTPCorr_Q15, /* I/O Normalized correlation; input: value from previous frame */ opus_int prevLag, /* I Last lag of previous frame; set to zero is unvoiced */ const opus_int32 search_thres1_Q16, /* I First stage threshold for lag candidates 0 - 1 */ - const opus_int search_thres2_Q15, /* I Final threshold for lag candidates 0 - 1 */ + const opus_int search_thres2_Q13, /* I Final threshold for lag candidates 0 - 1 */ const opus_int Fs_kHz, /* I Sample frequency (kHz) */ const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ - const opus_int nb_subfr /* I number of 5 ms subframes */ + const opus_int nb_subfr, /* I number of 5 ms subframes */ + int arch /* I Run-time architecture */ ); /* Compute Normalized Line Spectral Frequencies (NLSFs) from whitening filter coefficients */ @@ -315,7 +311,8 @@ void silk_burg_modified( const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ const opus_int nb_subfr, /* I Number of subframes stacked in x */ - const opus_int D /* I Order */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */ ); /* Copy and multiply a vector by a constant */ @@ -364,8 +361,8 @@ opus_int64 silk_inner_prod16_aligned_64( /* Rotate a32 right by 'rot' bits. Negative rot values result in rotating left. Output is 32bit int. Note: contemporary compilers recognize the C expression below and - compile it into a 'ror' instruction if available. No need for inline ASM! */ -static inline opus_int32 silk_ROR32( opus_int32 a32, opus_int rot ) + compile it into a 'ror' instruction if available. No need for OPUS_INLINE ASM! */ +static OPUS_INLINE opus_int32 silk_ROR32( opus_int32 a32, opus_int rot ) { opus_uint32 x = (opus_uint32) a32; opus_uint32 r = (opus_uint32) rot; @@ -514,37 +511,37 @@ static inline opus_int32 silk_ROR32( opus_int32 a32, opus_int rot ) #define SILK_FIX_CONST( C, Q ) ((opus_int32)((C) * ((opus_int64)1 << (Q)) + 0.5)) /* silk_min() versions with typecast in the function call */ -static inline opus_int silk_min_int(opus_int a, opus_int b) +static OPUS_INLINE opus_int silk_min_int(opus_int a, opus_int b) { return (((a) < (b)) ? (a) : (b)); } -static inline opus_int16 silk_min_16(opus_int16 a, opus_int16 b) +static OPUS_INLINE opus_int16 silk_min_16(opus_int16 a, opus_int16 b) { return (((a) < (b)) ? (a) : (b)); } -static inline opus_int32 silk_min_32(opus_int32 a, opus_int32 b) +static OPUS_INLINE opus_int32 silk_min_32(opus_int32 a, opus_int32 b) { return (((a) < (b)) ? (a) : (b)); } -static inline opus_int64 silk_min_64(opus_int64 a, opus_int64 b) +static OPUS_INLINE opus_int64 silk_min_64(opus_int64 a, opus_int64 b) { return (((a) < (b)) ? (a) : (b)); } /* silk_min() versions with typecast in the function call */ -static inline opus_int silk_max_int(opus_int a, opus_int b) +static OPUS_INLINE opus_int silk_max_int(opus_int a, opus_int b) { return (((a) > (b)) ? (a) : (b)); } -static inline opus_int16 silk_max_16(opus_int16 a, opus_int16 b) +static OPUS_INLINE opus_int16 silk_max_16(opus_int16 a, opus_int16 b) { return (((a) > (b)) ? (a) : (b)); } -static inline opus_int32 silk_max_32(opus_int32 a, opus_int32 b) +static OPUS_INLINE opus_int32 silk_max_32(opus_int32 a, opus_int32 b) { return (((a) > (b)) ? (a) : (b)); } -static inline opus_int64 silk_max_64(opus_int64 a, opus_int64 b) +static OPUS_INLINE opus_int64 silk_max_64(opus_int64 a, opus_int64 b) { return (((a) > (b)) ? (a) : (b)); } @@ -582,6 +579,14 @@ static inline opus_int64 silk_max_64(opus_int64 a, opus_int64 b) #include "MacroCount.h" #include "MacroDebug.h" +#ifdef OPUS_ARM_INLINE_ASM +#include "arm/SigProc_FIX_armv4.h" +#endif + +#ifdef OPUS_ARM_INLINE_EDSP +#include "arm/SigProc_FIX_armv5e.h" +#endif + #ifdef __cplusplus } #endif diff --git a/code/opus-1.0.2/silk/VAD.c b/code/opus-1.1/silk/VAD.c similarity index 86% rename from code/opus-1.0.2/silk/VAD.c rename to code/opus-1.1/silk/VAD.c index bac89b44..a8090981 100644 --- a/code/opus-1.0.2/silk/VAD.c +++ b/code/opus-1.1/silk/VAD.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,9 +30,10 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "stack_alloc.h" /* Silk VAD noise level estimation */ -static inline void silk_VAD_GetNoiseLevels( +static OPUS_INLINE void silk_VAD_GetNoiseLevels( const opus_int32 pX[ VAD_N_BANDS ], /* I subband energies */ silk_VAD_state *psSilk_VAD /* I/O Pointer to Silk VAD state */ ); @@ -82,15 +83,19 @@ opus_int silk_VAD_GetSA_Q8( /* O Return v ) { opus_int SA_Q15, pSNR_dB_Q7, input_tilt; - opus_int decimated_framelength, dec_subframe_length, dec_subframe_offset, SNR_Q7, i, b, s; + opus_int decimated_framelength1, decimated_framelength2; + opus_int decimated_framelength; + opus_int dec_subframe_length, dec_subframe_offset, SNR_Q7, i, b, s; opus_int32 sumSquared, smooth_coef_Q16; opus_int16 HPstateTmp; - opus_int16 X[ VAD_N_BANDS ][ MAX_FRAME_LENGTH / 2 ]; + VARDECL( opus_int16, X ); opus_int32 Xnrg[ VAD_N_BANDS ]; opus_int32 NrgToNoiseRatio_Q8[ VAD_N_BANDS ]; opus_int32 speech_nrg, x_tmp; + opus_int X_offset[ VAD_N_BANDS ]; opus_int ret = 0; silk_VAD_state *psSilk_VAD = &psEncC->sVAD; + SAVE_STACK; /* Safety checks */ silk_assert( VAD_N_BANDS == 4 ); @@ -101,26 +106,46 @@ opus_int silk_VAD_GetSA_Q8( /* O Return v /***********************/ /* Filter and Decimate */ /***********************/ + decimated_framelength1 = silk_RSHIFT( psEncC->frame_length, 1 ); + decimated_framelength2 = silk_RSHIFT( psEncC->frame_length, 2 ); + decimated_framelength = silk_RSHIFT( psEncC->frame_length, 3 ); + /* Decimate into 4 bands: + 0 L 3L L 3L 5L + - -- - -- -- + 8 8 2 4 4 + + [0-1 kHz| temp. |1-2 kHz| 2-4 kHz | 4-8 kHz | + + They're arranged to allow the minimal ( frame_length / 4 ) extra + scratch space during the downsampling process */ + X_offset[ 0 ] = 0; + X_offset[ 1 ] = decimated_framelength + decimated_framelength2; + X_offset[ 2 ] = X_offset[ 1 ] + decimated_framelength; + X_offset[ 3 ] = X_offset[ 2 ] + decimated_framelength2; + ALLOC( X, X_offset[ 3 ] + decimated_framelength1, opus_int16 ); + /* 0-8 kHz to 0-4 kHz and 4-8 kHz */ - silk_ana_filt_bank_1( pIn, &psSilk_VAD->AnaState[ 0 ], &X[ 0 ][ 0 ], &X[ 3 ][ 0 ], psEncC->frame_length ); + silk_ana_filt_bank_1( pIn, &psSilk_VAD->AnaState[ 0 ], + X, &X[ X_offset[ 3 ] ], psEncC->frame_length ); /* 0-4 kHz to 0-2 kHz and 2-4 kHz */ - silk_ana_filt_bank_1( &X[ 0 ][ 0 ], &psSilk_VAD->AnaState1[ 0 ], &X[ 0 ][ 0 ], &X[ 2 ][ 0 ], silk_RSHIFT( psEncC->frame_length, 1 ) ); + silk_ana_filt_bank_1( X, &psSilk_VAD->AnaState1[ 0 ], + X, &X[ X_offset[ 2 ] ], decimated_framelength1 ); /* 0-2 kHz to 0-1 kHz and 1-2 kHz */ - silk_ana_filt_bank_1( &X[ 0 ][ 0 ], &psSilk_VAD->AnaState2[ 0 ], &X[ 0 ][ 0 ], &X[ 1 ][ 0 ], silk_RSHIFT( psEncC->frame_length, 2 ) ); + silk_ana_filt_bank_1( X, &psSilk_VAD->AnaState2[ 0 ], + X, &X[ X_offset[ 1 ] ], decimated_framelength2 ); /*********************************************/ /* HP filter on lowest band (differentiator) */ /*********************************************/ - decimated_framelength = silk_RSHIFT( psEncC->frame_length, 3 ); - X[ 0 ][ decimated_framelength - 1 ] = silk_RSHIFT( X[ 0 ][ decimated_framelength - 1 ], 1 ); - HPstateTmp = X[ 0 ][ decimated_framelength - 1 ]; + X[ decimated_framelength - 1 ] = silk_RSHIFT( X[ decimated_framelength - 1 ], 1 ); + HPstateTmp = X[ decimated_framelength - 1 ]; for( i = decimated_framelength - 1; i > 0; i-- ) { - X[ 0 ][ i - 1 ] = silk_RSHIFT( X[ 0 ][ i - 1 ], 1 ); - X[ 0 ][ i ] -= X[ 0 ][ i - 1 ]; + X[ i - 1 ] = silk_RSHIFT( X[ i - 1 ], 1 ); + X[ i ] -= X[ i - 1 ]; } - X[ 0 ][ 0 ] -= psSilk_VAD->HPstate; + X[ 0 ] -= psSilk_VAD->HPstate; psSilk_VAD->HPstate = HPstateTmp; /*************************************/ @@ -142,7 +167,8 @@ opus_int silk_VAD_GetSA_Q8( /* O Return v for( i = 0; i < dec_subframe_length; i++ ) { /* The energy will be less than dec_subframe_length * ( silk_int16_MIN / 8 ) ^ 2. */ /* Therefore we can accumulate with no risk of overflow (unless dec_subframe_length > 128) */ - x_tmp = silk_RSHIFT( X[ b ][ i + dec_subframe_offset ], 3 ); + x_tmp = silk_RSHIFT( + X[ X_offset[ b ] + i + dec_subframe_offset ], 3 ); sumSquared = silk_SMLABB( sumSquared, x_tmp, x_tmp ); /* Safety check */ @@ -263,13 +289,14 @@ opus_int silk_VAD_GetSA_Q8( /* O Return v psEncC->input_quality_bands_Q15[ b ] = silk_sigm_Q15( silk_RSHIFT( SNR_Q7 - 16 * 128, 4 ) ); } + RESTORE_STACK; return( ret ); } /**************************/ /* Noise level estimation */ /**************************/ -static inline void silk_VAD_GetNoiseLevels( +static OPUS_INLINE void silk_VAD_GetNoiseLevels( const opus_int32 pX[ VAD_N_BANDS ], /* I subband energies */ silk_VAD_state *psSilk_VAD /* I/O Pointer to Silk VAD state */ ) diff --git a/code/opus-1.0.2/silk/VQ_WMat_EC.c b/code/opus-1.1/silk/VQ_WMat_EC.c similarity index 90% rename from code/opus-1.0.2/silk/VQ_WMat_EC.c rename to code/opus-1.1/silk/VQ_WMat_EC.c index a308cfbf..13d5d34e 100644 --- a/code/opus-1.0.2/silk/VQ_WMat_EC.c +++ b/code/opus-1.1/silk/VQ_WMat_EC.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -35,15 +35,18 @@ POSSIBILITY OF SUCH DAMAGE. void silk_VQ_WMat_EC( opus_int8 *ind, /* O index of best codebook vector */ opus_int32 *rate_dist_Q14, /* O best weighted quant error + mu * rate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ const opus_int16 *in_Q14, /* I input vector to be quantized */ const opus_int32 *W_Q18, /* I weighting matrix */ const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ const opus_int mu_Q9, /* I tradeoff betw. weighted error and rate */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ opus_int L /* I number of vectors in codebook */ ) { - opus_int k; + opus_int k, gain_tmp_Q7; const opus_int8 *cb_row_Q7; opus_int16 diff_Q14[ 5 ]; opus_int32 sum1_Q14, sum2_Q16; @@ -52,6 +55,8 @@ void silk_VQ_WMat_EC( *rate_dist_Q14 = silk_int32_MAX; cb_row_Q7 = cb_Q7; for( k = 0; k < L; k++ ) { + gain_tmp_Q7 = cb_gain_Q7[k]; + diff_Q14[ 0 ] = in_Q14[ 0 ] - silk_LSHIFT( cb_row_Q7[ 0 ], 7 ); diff_Q14[ 1 ] = in_Q14[ 1 ] - silk_LSHIFT( cb_row_Q7[ 1 ], 7 ); diff_Q14[ 2 ] = in_Q14[ 2 ] - silk_LSHIFT( cb_row_Q7[ 2 ], 7 ); @@ -61,6 +66,9 @@ void silk_VQ_WMat_EC( /* Weighted rate */ sum1_Q14 = silk_SMULBB( mu_Q9, cl_Q5[ k ] ); + /* Penalty for too large gain */ + sum1_Q14 = silk_ADD_LSHIFT32( sum1_Q14, silk_max( silk_SUB32( gain_tmp_Q7, max_gain_Q7 ), 0 ), 10 ); + silk_assert( sum1_Q14 >= 0 ); /* first row of W_Q18 */ @@ -103,6 +111,7 @@ void silk_VQ_WMat_EC( if( sum1_Q14 < *rate_dist_Q14 ) { *rate_dist_Q14 = sum1_Q14; *ind = (opus_int8)k; + *gain_Q7 = gain_tmp_Q7; } /* Go to next cbk vector */ diff --git a/code/opus-1.0.2/silk/ana_filt_bank_1.c b/code/opus-1.1/silk/ana_filt_bank_1.c similarity index 99% rename from code/opus-1.0.2/silk/ana_filt_bank_1.c rename to code/opus-1.1/silk/ana_filt_bank_1.c index 4e04bef3..24cfb03f 100644 --- a/code/opus-1.0.2/silk/ana_filt_bank_1.c +++ b/code/opus-1.1/silk/ana_filt_bank_1.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.1/silk/arm/SigProc_FIX_armv4.h b/code/opus-1.1/silk/arm/SigProc_FIX_armv4.h new file mode 100644 index 00000000..ff62b1e5 --- /dev/null +++ b/code/opus-1.1/silk/arm/SigProc_FIX_armv4.h @@ -0,0 +1,47 @@ +/*********************************************************************** +Copyright (C) 2013 Xiph.Org Foundation and contributors +Copyright (c) 2013 Parrot +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_SIGPROC_FIX_ARMv4_H +#define SILK_SIGPROC_FIX_ARMv4_H + +#undef silk_MLA +static OPUS_INLINE opus_int32 silk_MLA_armv4(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + opus_int32 res; + __asm__( + "#silk_MLA\n\t" + "mla %0, %1, %2, %3\n\t" + : "=&r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_MLA(a, b, c) (silk_MLA_armv4(a, b, c)) + +#endif diff --git a/code/opus-1.1/silk/arm/SigProc_FIX_armv5e.h b/code/opus-1.1/silk/arm/SigProc_FIX_armv5e.h new file mode 100644 index 00000000..617a09ca --- /dev/null +++ b/code/opus-1.1/silk/arm/SigProc_FIX_armv5e.h @@ -0,0 +1,61 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Copyright (c) 2013 Parrot +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_SIGPROC_FIX_ARMv5E_H +#define SILK_SIGPROC_FIX_ARMv5E_H + +#undef silk_SMULTT +static OPUS_INLINE opus_int32 silk_SMULTT_armv5e(opus_int32 a, opus_int32 b) +{ + opus_int32 res; + __asm__( + "#silk_SMULTT\n\t" + "smultt %0, %1, %2\n\t" + : "=r"(res) + : "%r"(a), "r"(b) + ); + return res; +} +#define silk_SMULTT(a, b) (silk_SMULTT_armv5e(a, b)) + +#undef silk_SMLATT +static OPUS_INLINE opus_int32 silk_SMLATT_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + opus_int32 res; + __asm__( + "#silk_SMLATT\n\t" + "smlatt %0, %1, %2, %3\n\t" + : "=r"(res) + : "%r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLATT(a, b, c) (silk_SMLATT_armv5e(a, b, c)) + +#endif diff --git a/code/opus-1.1/silk/arm/macros_armv4.h b/code/opus-1.1/silk/arm/macros_armv4.h new file mode 100644 index 00000000..3f30e972 --- /dev/null +++ b/code/opus-1.1/silk/arm/macros_armv4.h @@ -0,0 +1,103 @@ +/*********************************************************************** +Copyright (C) 2013 Xiph.Org Foundation and contributors. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MACROS_ARMv4_H +#define SILK_MACROS_ARMv4_H + +/* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */ +#undef silk_SMULWB +static OPUS_INLINE opus_int32 silk_SMULWB_armv4(opus_int32 a, opus_int16 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMULWB\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(a), "r"(b<<16) + ); + return rd_hi; +} +#define silk_SMULWB(a, b) (silk_SMULWB_armv4(a, b)) + +/* a32 + (b32 * (opus_int32)((opus_int16)(c32))) >> 16 output have to be 32bit int */ +#undef silk_SMLAWB +#define silk_SMLAWB(a, b, c) ((a) + silk_SMULWB(b, c)) + +/* (a32 * (b32 >> 16)) >> 16 */ +#undef silk_SMULWT +static OPUS_INLINE opus_int32 silk_SMULWT_armv4(opus_int32 a, opus_int32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMULWT\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(a), "r"(b&~0xFFFF) + ); + return rd_hi; +} +#define silk_SMULWT(a, b) (silk_SMULWT_armv4(a, b)) + +/* a32 + (b32 * (c32 >> 16)) >> 16 */ +#undef silk_SMLAWT +#define silk_SMLAWT(a, b, c) ((a) + silk_SMULWT(b, c)) + +/* (a32 * b32) >> 16 */ +#undef silk_SMULWW +static OPUS_INLINE opus_int32 silk_SMULWW_armv4(opus_int32 a, opus_int32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMULWW\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(a), "r"(b) + ); + return (rd_hi<<16)+(rd_lo>>16); +} +#define silk_SMULWW(a, b) (silk_SMULWW_armv4(a, b)) + +#undef silk_SMLAWW +static OPUS_INLINE opus_int32 silk_SMLAWW_armv4(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMLAWW\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(b), "r"(c) + ); + return a+(rd_hi<<16)+(rd_lo>>16); +} +#define silk_SMLAWW(a, b, c) (silk_SMLAWW_armv4(a, b, c)) + +#endif /* SILK_MACROS_ARMv4_H */ diff --git a/code/opus-1.1/silk/arm/macros_armv5e.h b/code/opus-1.1/silk/arm/macros_armv5e.h new file mode 100644 index 00000000..aad4117e --- /dev/null +++ b/code/opus-1.1/silk/arm/macros_armv5e.h @@ -0,0 +1,213 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Copyright (c) 2013 Parrot +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MACROS_ARMv5E_H +#define SILK_MACROS_ARMv5E_H + +/* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */ +#undef silk_SMULWB +static OPUS_INLINE opus_int32 silk_SMULWB_armv5e(opus_int32 a, opus_int16 b) +{ + int res; + __asm__( + "#silk_SMULWB\n\t" + "smulwb %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SMULWB(a, b) (silk_SMULWB_armv5e(a, b)) + +/* a32 + (b32 * (opus_int32)((opus_int16)(c32))) >> 16 output have to be 32bit int */ +#undef silk_SMLAWB +static OPUS_INLINE opus_int32 silk_SMLAWB_armv5e(opus_int32 a, opus_int32 b, + opus_int16 c) +{ + int res; + __asm__( + "#silk_SMLAWB\n\t" + "smlawb %0, %1, %2, %3\n\t" + : "=r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLAWB(a, b, c) (silk_SMLAWB_armv5e(a, b, c)) + +/* (a32 * (b32 >> 16)) >> 16 */ +#undef silk_SMULWT +static OPUS_INLINE opus_int32 silk_SMULWT_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SMULWT\n\t" + "smulwt %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SMULWT(a, b) (silk_SMULWT_armv5e(a, b)) + +/* a32 + (b32 * (c32 >> 16)) >> 16 */ +#undef silk_SMLAWT +static OPUS_INLINE opus_int32 silk_SMLAWT_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + int res; + __asm__( + "#silk_SMLAWT\n\t" + "smlawt %0, %1, %2, %3\n\t" + : "=r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLAWT(a, b, c) (silk_SMLAWT_armv5e(a, b, c)) + +/* (opus_int32)((opus_int16)(a3))) * (opus_int32)((opus_int16)(b32)) output have to be 32bit int */ +#undef silk_SMULBB +static OPUS_INLINE opus_int32 silk_SMULBB_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SMULBB\n\t" + "smulbb %0, %1, %2\n\t" + : "=r"(res) + : "%r"(a), "r"(b) + ); + return res; +} +#define silk_SMULBB(a, b) (silk_SMULBB_armv5e(a, b)) + +/* a32 + (opus_int32)((opus_int16)(b32)) * (opus_int32)((opus_int16)(c32)) output have to be 32bit int */ +#undef silk_SMLABB +static OPUS_INLINE opus_int32 silk_SMLABB_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + int res; + __asm__( + "#silk_SMLABB\n\t" + "smlabb %0, %1, %2, %3\n\t" + : "=r"(res) + : "%r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLABB(a, b, c) (silk_SMLABB_armv5e(a, b, c)) + +/* (opus_int32)((opus_int16)(a32)) * (b32 >> 16) */ +#undef silk_SMULBT +static OPUS_INLINE opus_int32 silk_SMULBT_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SMULBT\n\t" + "smulbt %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SMULBT(a, b) (silk_SMULBT_armv5e(a, b)) + +/* a32 + (opus_int32)((opus_int16)(b32)) * (c32 >> 16) */ +#undef silk_SMLABT +static OPUS_INLINE opus_int32 silk_SMLABT_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + int res; + __asm__( + "#silk_SMLABT\n\t" + "smlabt %0, %1, %2, %3\n\t" + : "=r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLABT(a, b, c) (silk_SMLABT_armv5e(a, b, c)) + +/* add/subtract with output saturated */ +#undef silk_ADD_SAT32 +static OPUS_INLINE opus_int32 silk_ADD_SAT32_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_ADD_SAT32\n\t" + "qadd %0, %1, %2\n\t" + : "=r"(res) + : "%r"(a), "r"(b) + ); + return res; +} +#define silk_ADD_SAT32(a, b) (silk_ADD_SAT32_armv5e(a, b)) + +#undef silk_SUB_SAT32 +static OPUS_INLINE opus_int32 silk_SUB_SAT32_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SUB_SAT32\n\t" + "qsub %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SUB_SAT32(a, b) (silk_SUB_SAT32_armv5e(a, b)) + +#undef silk_CLZ16 +static OPUS_INLINE opus_int32 silk_CLZ16_armv5(opus_int16 in16) +{ + int res; + __asm__( + "#silk_CLZ16\n\t" + "clz %0, %1;\n" + : "=r"(res) + : "r"(in16<<16|0x8000) + ); + return res; +} +#define silk_CLZ16(in16) (silk_CLZ16_armv5(in16)) + +#undef silk_CLZ32 +static OPUS_INLINE opus_int32 silk_CLZ32_armv5(opus_int32 in32) +{ + int res; + __asm__( + "#silk_CLZ32\n\t" + "clz %0, %1\n\t" + : "=r"(res) + : "r"(in32) + ); + return res; +} +#define silk_CLZ32(in32) (silk_CLZ32_armv5(in32)) + +#endif /* SILK_MACROS_ARMv5E_H */ diff --git a/code/opus-1.0.2/silk/biquad_alt.c b/code/opus-1.1/silk/biquad_alt.c similarity index 99% rename from code/opus-1.0.2/silk/biquad_alt.c rename to code/opus-1.1/silk/biquad_alt.c index a639e21a..d55f5ee9 100644 --- a/code/opus-1.0.2/silk/biquad_alt.c +++ b/code/opus-1.1/silk/biquad_alt.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/bwexpander.c b/code/opus-1.1/silk/bwexpander.c similarity index 99% rename from code/opus-1.0.2/silk/bwexpander.c rename to code/opus-1.1/silk/bwexpander.c index 77ea1163..2eb44566 100644 --- a/code/opus-1.0.2/silk/bwexpander.c +++ b/code/opus-1.1/silk/bwexpander.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/bwexpander_32.c b/code/opus-1.1/silk/bwexpander_32.c similarity index 99% rename from code/opus-1.0.2/silk/bwexpander_32.c rename to code/opus-1.1/silk/bwexpander_32.c index 5ad92dd4..d0010f73 100644 --- a/code/opus-1.0.2/silk/bwexpander_32.c +++ b/code/opus-1.1/silk/bwexpander_32.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/check_control_input.c b/code/opus-1.1/silk/check_control_input.c similarity index 99% rename from code/opus-1.0.2/silk/check_control_input.c rename to code/opus-1.1/silk/check_control_input.c index 972a480d..b5de9ce4 100644 --- a/code/opus-1.0.2/silk/check_control_input.c +++ b/code/opus-1.1/silk/check_control_input.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/code_signs.c b/code/opus-1.1/silk/code_signs.c similarity index 99% rename from code/opus-1.0.2/silk/code_signs.c rename to code/opus-1.1/silk/code_signs.c index 9893cdd8..0419ea26 100644 --- a/code/opus-1.0.2/silk/code_signs.c +++ b/code/opus-1.1/silk/code_signs.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/control.h b/code/opus-1.1/silk/control.h similarity index 97% rename from code/opus-1.0.2/silk/control.h rename to code/opus-1.1/silk/control.h index c52ec3fe..747e5426 100644 --- a/code/opus-1.0.2/silk/control.h +++ b/code/opus-1.1/silk/control.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -92,6 +92,9 @@ typedef struct { /* I: Opus encoder is allowing us to switch bandwidth */ opus_int opusCanSwitch; + /* I: Make frames as independent as possible (but still use LPC) */ + opus_int reducedDependency; + /* O: Internal sampling rate used, in Hertz; 8000/12000/16000 */ opus_int32 internalSampleRate; diff --git a/code/opus-1.0.2/silk/control_SNR.c b/code/opus-1.1/silk/control_SNR.c similarity index 99% rename from code/opus-1.0.2/silk/control_SNR.c rename to code/opus-1.1/silk/control_SNR.c index 08e4e1a1..f04e69fc 100644 --- a/code/opus-1.0.2/silk/control_SNR.c +++ b/code/opus-1.1/silk/control_SNR.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/control_audio_bandwidth.c b/code/opus-1.1/silk/control_audio_bandwidth.c similarity index 97% rename from code/opus-1.0.2/silk/control_audio_bandwidth.c rename to code/opus-1.1/silk/control_audio_bandwidth.c index b645dd57..4f9bc5cb 100644 --- a/code/opus-1.0.2/silk/control_audio_bandwidth.c +++ b/code/opus-1.1/silk/control_audio_bandwidth.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -115,6 +115,9 @@ opus_int silk_control_audio_bandwidth( psEncC->sLP.mode = 1; } } + } else { + if (psEncC->sLP.mode<0) + psEncC->sLP.mode = 1; } } } diff --git a/code/opus-1.0.2/silk/control_codec.c b/code/opus-1.1/silk/control_codec.c similarity index 92% rename from code/opus-1.0.2/silk/control_codec.c rename to code/opus-1.1/silk/control_codec.c index ecc338ce..1f674bdd 100644 --- a/code/opus-1.0.2/silk/control_codec.c +++ b/code/opus-1.1/silk/control_codec.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -35,6 +35,7 @@ POSSIBILITY OF SUCH DAMAGE. #include "main_FLP.h" #define silk_encoder_state_Fxx silk_encoder_state_FLP #endif +#include "stack_alloc.h" #include "tuning_parameters.h" #include "pitch_est_defines.h" @@ -54,7 +55,7 @@ static opus_int silk_setup_complexity( opus_int Complexity /* I */ ); -static inline opus_int silk_setup_LBRR( +static OPUS_INLINE opus_int silk_setup_LBRR( silk_encoder_state *psEncC, /* I/O */ const opus_int32 TargetRate_bps /* I */ ); @@ -137,7 +138,7 @@ static opus_int silk_setup_resamplers( ) { opus_int ret = SILK_NO_ERROR; - opus_int32 nSamples_temp; + SAVE_STACK; if( psEnc->sCmn.fs_kHz != fs_kHz || psEnc->sCmn.prev_API_fs_Hz != psEnc->sCmn.API_fs_Hz ) { @@ -145,44 +146,54 @@ static opus_int silk_setup_resamplers( /* Initialize the resampler for enc_API.c preparing resampling from API_fs_Hz to fs_kHz */ ret += silk_resampler_init( &psEnc->sCmn.resampler_state, psEnc->sCmn.API_fs_Hz, fs_kHz * 1000, 1 ); } else { - /* Allocate worst case space for temporary upsampling, 8 to 48 kHz, so a factor 6 */ - opus_int16 x_buf_API_fs_Hz[ ( 2 * MAX_FRAME_LENGTH_MS + LA_SHAPE_MS ) * MAX_API_FS_KHZ ]; - silk_resampler_state_struct temp_resampler_state; + VARDECL( opus_int16, x_buf_API_fs_Hz ); + VARDECL( silk_resampler_state_struct, temp_resampler_state ); #ifdef FIXED_POINT opus_int16 *x_bufFIX = psEnc->x_buf; #else - opus_int16 x_bufFIX[ 2 * MAX_FRAME_LENGTH + LA_SHAPE_MAX ]; + VARDECL( opus_int16, x_bufFIX ); + opus_int32 new_buf_samples; #endif + opus_int32 api_buf_samples; + opus_int32 old_buf_samples; + opus_int32 buf_length_ms; - nSamples_temp = silk_LSHIFT( psEnc->sCmn.frame_length, 1 ) + LA_SHAPE_MS * psEnc->sCmn.fs_kHz; + buf_length_ms = silk_LSHIFT( psEnc->sCmn.nb_subfr * 5, 1 ) + LA_SHAPE_MS; + old_buf_samples = buf_length_ms * psEnc->sCmn.fs_kHz; #ifndef FIXED_POINT - silk_float2short_array( x_bufFIX, psEnc->x_buf, nSamples_temp ); + new_buf_samples = buf_length_ms * fs_kHz; + ALLOC( x_bufFIX, silk_max( old_buf_samples, new_buf_samples ), + opus_int16 ); + silk_float2short_array( x_bufFIX, psEnc->x_buf, old_buf_samples ); #endif /* Initialize resampler for temporary resampling of x_buf data to API_fs_Hz */ - ret += silk_resampler_init( &temp_resampler_state, silk_SMULBB( psEnc->sCmn.fs_kHz, 1000 ), psEnc->sCmn.API_fs_Hz, 0 ); + ALLOC( temp_resampler_state, 1, silk_resampler_state_struct ); + ret += silk_resampler_init( temp_resampler_state, silk_SMULBB( psEnc->sCmn.fs_kHz, 1000 ), psEnc->sCmn.API_fs_Hz, 0 ); + + /* Calculate number of samples to temporarily upsample */ + api_buf_samples = buf_length_ms * silk_DIV32_16( psEnc->sCmn.API_fs_Hz, 1000 ); /* Temporary resampling of x_buf data to API_fs_Hz */ - ret += silk_resampler( &temp_resampler_state, x_buf_API_fs_Hz, x_bufFIX, nSamples_temp ); - - /* Calculate number of samples that has been temporarily upsampled */ - nSamples_temp = silk_DIV32_16( nSamples_temp * psEnc->sCmn.API_fs_Hz, silk_SMULBB( psEnc->sCmn.fs_kHz, 1000 ) ); + ALLOC( x_buf_API_fs_Hz, api_buf_samples, opus_int16 ); + ret += silk_resampler( temp_resampler_state, x_buf_API_fs_Hz, x_bufFIX, old_buf_samples ); /* Initialize the resampler for enc_API.c preparing resampling from API_fs_Hz to fs_kHz */ ret += silk_resampler_init( &psEnc->sCmn.resampler_state, psEnc->sCmn.API_fs_Hz, silk_SMULBB( fs_kHz, 1000 ), 1 ); /* Correct resampler state by resampling buffered data from API_fs_Hz to fs_kHz */ - ret += silk_resampler( &psEnc->sCmn.resampler_state, x_bufFIX, x_buf_API_fs_Hz, nSamples_temp ); + ret += silk_resampler( &psEnc->sCmn.resampler_state, x_bufFIX, x_buf_API_fs_Hz, api_buf_samples ); #ifndef FIXED_POINT - silk_short2float_array( psEnc->x_buf, x_bufFIX, ( 2 * MAX_FRAME_LENGTH_MS + LA_SHAPE_MS ) * fs_kHz ); + silk_short2float_array( psEnc->x_buf, x_bufFIX, new_buf_samples); #endif } } psEnc->sCmn.prev_API_fs_Hz = psEnc->sCmn.API_fs_Hz; + RESTORE_STACK; return ret; } @@ -381,7 +392,7 @@ static opus_int silk_setup_complexity( return ret; } -static inline opus_int silk_setup_LBRR( +static OPUS_INLINE opus_int silk_setup_LBRR( silk_encoder_state *psEncC, /* I/O */ const opus_int32 TargetRate_bps /* I */ ) diff --git a/code/opus-1.0.2/silk/debug.c b/code/opus-1.1/silk/debug.c similarity index 99% rename from code/opus-1.0.2/silk/debug.c rename to code/opus-1.1/silk/debug.c index 9aa16cc8..9253faf7 100644 --- a/code/opus-1.0.2/silk/debug.c +++ b/code/opus-1.1/silk/debug.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/debug.h b/code/opus-1.1/silk/debug.h similarity index 98% rename from code/opus-1.0.2/silk/debug.h rename to code/opus-1.1/silk/debug.h index 8ae7094f..efb6d3e9 100644 --- a/code/opus-1.0.2/silk/debug.h +++ b/code/opus-1.1/silk/debug.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -28,10 +28,6 @@ POSSIBILITY OF SUCH DAMAGE. #ifndef SILK_DEBUG_H #define SILK_DEBUG_H -#ifdef _WIN32 -#define _CRT_SECURE_NO_DEPRECATE 1 -#endif - #include "typedef.h" #include /* file writing */ #include /* strcpy, strcmp */ @@ -69,7 +65,6 @@ unsigned long GetHighResolutionTime(void); /* O time in usec*/ #if (defined(_WIN32) || defined(_WINCE)) #include /* timer */ -#pragma warning( disable : 4996 ) /* stop bitching about strcpy in TIC()*/ #else /* Linux or Mac*/ #include #endif diff --git a/code/opus-1.0.2/silk/dec_API.c b/code/opus-1.1/silk/dec_API.c similarity index 97% rename from code/opus-1.0.2/silk/dec_API.c rename to code/opus-1.1/silk/dec_API.c index 68403b7c..4cbcf715 100644 --- a/code/opus-1.0.2/silk/dec_API.c +++ b/code/opus-1.1/silk/dec_API.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -69,6 +69,9 @@ opus_int silk_InitDecoder( /* O Returns error co for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { ret = silk_init_decoder( &channel_state[ n ] ); } + silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo)); + /* Not strictly needed, but it's cleaner that way */ + ((silk_decoder *)decState)->prev_decode_only_middle = 0; return ret; } @@ -97,6 +100,8 @@ opus_int silk_Decode( /* O Returns error co opus_int stereo_to_mono; SAVE_STACK; + silk_assert( decControl->nChannelsInternal == 1 || decControl->nChannelsInternal == 2 ); + /**********************************/ /* Test if first frame in payload */ /**********************************/ @@ -300,7 +305,7 @@ opus_int silk_Decode( /* O Returns error co /* Set up pointers to temp buffers */ ALLOC( samplesOut2_tmp, - decControl->nChannelsAPI == 2 ? *nSamplesOut : 0, opus_int16 ); + decControl->nChannelsAPI == 2 ? *nSamplesOut : ALLOC_NONE, opus_int16 ); if( decControl->nChannelsAPI == 2 ) { resample_out_ptr = samplesOut2_tmp; } else { diff --git a/code/opus-1.0.2/silk/decode_core.c b/code/opus-1.1/silk/decode_core.c similarity index 99% rename from code/opus-1.0.2/silk/decode_core.c rename to code/opus-1.1/silk/decode_core.c index 0365ffdf..a820bf11 100644 --- a/code/opus-1.0.2/silk/decode_core.c +++ b/code/opus-1.1/silk/decode_core.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/decode_frame.c b/code/opus-1.1/silk/decode_frame.c similarity index 99% rename from code/opus-1.0.2/silk/decode_frame.c rename to code/opus-1.1/silk/decode_frame.c index 3e4a6e2b..abc00a3d 100644 --- a/code/opus-1.0.2/silk/decode_frame.c +++ b/code/opus-1.1/silk/decode_frame.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/decode_indices.c b/code/opus-1.1/silk/decode_indices.c similarity index 99% rename from code/opus-1.0.2/silk/decode_indices.c rename to code/opus-1.1/silk/decode_indices.c index 69172102..7afe5c26 100644 --- a/code/opus-1.0.2/silk/decode_indices.c +++ b/code/opus-1.1/silk/decode_indices.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/decode_parameters.c b/code/opus-1.1/silk/decode_parameters.c similarity index 99% rename from code/opus-1.0.2/silk/decode_parameters.c rename to code/opus-1.1/silk/decode_parameters.c index e4c7e7a4..e345b1dc 100644 --- a/code/opus-1.0.2/silk/decode_parameters.c +++ b/code/opus-1.1/silk/decode_parameters.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/decode_pitch.c b/code/opus-1.1/silk/decode_pitch.c similarity index 99% rename from code/opus-1.0.2/silk/decode_pitch.c rename to code/opus-1.1/silk/decode_pitch.c index 80fb4d9f..fedbc6a5 100644 --- a/code/opus-1.0.2/silk/decode_pitch.c +++ b/code/opus-1.1/silk/decode_pitch.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/decode_pulses.c b/code/opus-1.1/silk/decode_pulses.c similarity index 99% rename from code/opus-1.0.2/silk/decode_pulses.c rename to code/opus-1.1/silk/decode_pulses.c index 1c781a0b..e8a87c2a 100644 --- a/code/opus-1.0.2/silk/decode_pulses.c +++ b/code/opus-1.1/silk/decode_pulses.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/decoder_set_fs.c b/code/opus-1.1/silk/decoder_set_fs.c similarity index 99% rename from code/opus-1.0.2/silk/decoder_set_fs.c rename to code/opus-1.1/silk/decoder_set_fs.c index 38ac249c..eef0fd25 100644 --- a/code/opus-1.0.2/silk/decoder_set_fs.c +++ b/code/opus-1.1/silk/decoder_set_fs.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/define.h b/code/opus-1.1/silk/define.h similarity index 99% rename from code/opus-1.0.2/silk/define.h rename to code/opus-1.1/silk/define.h index f74f4869..c47aca9f 100644 --- a/code/opus-1.0.2/silk/define.h +++ b/code/opus-1.1/silk/define.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/enc_API.c b/code/opus-1.1/silk/enc_API.c similarity index 94% rename from code/opus-1.0.2/silk/enc_API.c rename to code/opus-1.1/silk/enc_API.c index ec7915ce..43739efc 100644 --- a/code/opus-1.0.2/silk/enc_API.c +++ b/code/opus-1.1/silk/enc_API.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -32,6 +32,7 @@ POSSIBILITY OF SUCH DAMAGE. #include "API.h" #include "control.h" #include "typedef.h" +#include "stack_alloc.h" #include "structs.h" #include "tuning_parameters.h" #ifdef FIXED_POINT @@ -68,6 +69,7 @@ opus_int silk_Get_Encoder_Size( /* O Returns error co /*************************/ opus_int silk_InitEncoder( /* O Returns error code */ void *encState, /* I/O State */ + int arch, /* I Run-time architecture */ silk_EncControlStruct *encStatus /* O Encoder Status */ ) { @@ -79,7 +81,7 @@ opus_int silk_InitEncoder( /* O Returns error co /* Reset encoder */ silk_memset( psEnc, 0, sizeof( silk_encoder ) ); for( n = 0; n < ENCODER_NUM_CHANNELS; n++ ) { - if( ret += silk_init_encoder( &psEnc->state_Fxx[ n ] ) ) { + if( ret += silk_init_encoder( &psEnc->state_Fxx[ n ], arch ) ) { silk_assert( 0 ); } } @@ -146,18 +148,26 @@ opus_int silk_Encode( /* O Returns error co ) { opus_int n, i, nBits, flags, tmp_payloadSize_ms = 0, tmp_complexity = 0, ret = 0; - opus_int nSamplesToBuffer, nBlocksOf10ms, nSamplesFromInput = 0; + opus_int nSamplesToBuffer, nSamplesToBufferMax, nBlocksOf10ms; + opus_int nSamplesFromInput = 0, nSamplesFromInputMax; opus_int speech_act_thr_for_switch_Q8; opus_int32 TargetRate_bps, MStargetRates_bps[ 2 ], channelRate_bps, LBRR_symbol, sum; silk_encoder *psEnc = ( silk_encoder * )encState; - opus_int16 buf[ MAX_FRAME_LENGTH_MS * MAX_API_FS_KHZ ]; + VARDECL( opus_int16, buf ); opus_int transition, curr_block, tot_blocks; + SAVE_STACK; + if (encControl->reducedDependency) + { + psEnc->state_Fxx[0].sCmn.first_frame_after_reset = 1; + psEnc->state_Fxx[1].sCmn.first_frame_after_reset = 1; + } psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded = psEnc->state_Fxx[ 1 ].sCmn.nFramesEncoded = 0; /* Check values in encoder control structure */ if( ( ret = check_control_input( encControl ) != 0 ) ) { silk_assert( 0 ); + RESTORE_STACK; return ret; } @@ -165,7 +175,7 @@ opus_int silk_Encode( /* O Returns error co if( encControl->nChannelsInternal > psEnc->nChannelsInternal ) { /* Mono -> Stereo transition: init state of second channel and stereo state */ - ret += silk_init_encoder( &psEnc->state_Fxx[ 1 ] ); + ret += silk_init_encoder( &psEnc->state_Fxx[ 1 ], psEnc->state_Fxx[ 0 ].sCmn.arch ); silk_memset( psEnc->sStereo.pred_prev_Q13, 0, sizeof( psEnc->sStereo.pred_prev_Q13 ) ); silk_memset( psEnc->sStereo.sSide, 0, sizeof( psEnc->sStereo.sSide ) ); psEnc->sStereo.mid_side_amp_Q0[ 0 ] = 0; @@ -191,15 +201,14 @@ opus_int silk_Encode( /* O Returns error co if( prefillFlag ) { /* Only accept input length of 10 ms */ if( nBlocksOf10ms != 1 ) { - ret = SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; silk_assert( 0 ); - return ret; + RESTORE_STACK; + return SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; } /* Reset Encoder */ for( n = 0; n < encControl->nChannelsInternal; n++ ) { - if( (ret = silk_init_encoder( &psEnc->state_Fxx[ n ] ) ) != 0 ) { - silk_assert( 0 ); - } + ret = silk_init_encoder( &psEnc->state_Fxx[ n ], psEnc->state_Fxx[ n ].sCmn.arch ); + silk_assert( !ret ); } tmp_payloadSize_ms = encControl->payloadSize_ms; encControl->payloadSize_ms = 10; @@ -212,15 +221,15 @@ opus_int silk_Encode( /* O Returns error co } else { /* Only accept input lengths that are a multiple of 10 ms */ if( nBlocksOf10ms * encControl->API_sampleRate != 100 * nSamplesIn || nSamplesIn < 0 ) { - ret = SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; silk_assert( 0 ); - return ret; + RESTORE_STACK; + return SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; } /* Make sure no more than one packet can be produced */ if( 1000 * (opus_int32)nSamplesIn > encControl->payloadSize_ms * encControl->API_sampleRate ) { - ret = SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; silk_assert( 0 ); - return ret; + RESTORE_STACK; + return SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; } } @@ -230,6 +239,7 @@ opus_int silk_Encode( /* O Returns error co opus_int force_fs_kHz = (n==1) ? psEnc->state_Fxx[0].sCmn.fs_kHz : 0; if( ( ret = silk_control_encoder( &psEnc->state_Fxx[ n ], encControl, TargetRate_bps, psEnc->allowBandwidthSwitch, n, force_fs_kHz ) ) != 0 ) { silk_assert( 0 ); + RESTORE_STACK; return ret; } if( psEnc->state_Fxx[n].sCmn.first_frame_after_reset || transition ) { @@ -242,9 +252,16 @@ opus_int silk_Encode( /* O Returns error co silk_assert( encControl->nChannelsInternal == 1 || psEnc->state_Fxx[ 0 ].sCmn.fs_kHz == psEnc->state_Fxx[ 1 ].sCmn.fs_kHz ); /* Input buffering/resampling and encoding */ + nSamplesToBufferMax = + 10 * nBlocksOf10ms * psEnc->state_Fxx[ 0 ].sCmn.fs_kHz; + nSamplesFromInputMax = + silk_DIV32_16( nSamplesToBufferMax * + psEnc->state_Fxx[ 0 ].sCmn.API_fs_Hz, + psEnc->state_Fxx[ 0 ].sCmn.fs_kHz * 1000 ); + ALLOC( buf, nSamplesFromInputMax, opus_int16 ); while( 1 ) { nSamplesToBuffer = psEnc->state_Fxx[ 0 ].sCmn.frame_length - psEnc->state_Fxx[ 0 ].sCmn.inputBufIx; - nSamplesToBuffer = silk_min( nSamplesToBuffer, 10 * nBlocksOf10ms * psEnc->state_Fxx[ 0 ].sCmn.fs_kHz ); + nSamplesToBuffer = silk_min( nSamplesToBuffer, nSamplesToBufferMax ); nSamplesFromInput = silk_DIV32_16( nSamplesToBuffer * psEnc->state_Fxx[ 0 ].sCmn.API_fs_Hz, psEnc->state_Fxx[ 0 ].sCmn.fs_kHz * 1000 ); /* Resample and write to buffer */ if( encControl->nChannelsAPI == 2 && encControl->nChannelsInternal == 2 ) { @@ -533,6 +550,7 @@ opus_int silk_Encode( /* O Returns error co } } + RESTORE_STACK; return ret; } diff --git a/code/opus-1.0.2/silk/encode_indices.c b/code/opus-1.1/silk/encode_indices.c similarity index 99% rename from code/opus-1.0.2/silk/encode_indices.c rename to code/opus-1.1/silk/encode_indices.c index 91e28aa9..666c8c0b 100644 --- a/code/opus-1.0.2/silk/encode_indices.c +++ b/code/opus-1.1/silk/encode_indices.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/encode_pulses.c b/code/opus-1.1/silk/encode_pulses.c similarity index 94% rename from code/opus-1.0.2/silk/encode_pulses.c rename to code/opus-1.1/silk/encode_pulses.c index b01b5853..a4501438 100644 --- a/code/opus-1.0.2/silk/encode_pulses.c +++ b/code/opus-1.1/silk/encode_pulses.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,12 +30,13 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "stack_alloc.h" /*********************************************/ /* Encode quantization indices of excitation */ /*********************************************/ -static inline opus_int combine_and_check( /* return ok */ +static OPUS_INLINE opus_int combine_and_check( /* return ok */ opus_int *pulses_comb, /* O */ const opus_int *pulses_in, /* I */ opus_int max_pulses, /* I max value for sum of pulses */ @@ -66,14 +67,15 @@ void silk_encode_pulses( { opus_int i, k, j, iter, bit, nLS, scale_down, RateLevelIndex = 0; opus_int32 abs_q, minSumBits_Q5, sumBits_Q5; - opus_int abs_pulses[ MAX_FRAME_LENGTH ]; - opus_int sum_pulses[ MAX_NB_SHELL_BLOCKS ]; - opus_int nRshifts[ MAX_NB_SHELL_BLOCKS ]; + VARDECL( opus_int, abs_pulses ); + VARDECL( opus_int, sum_pulses ); + VARDECL( opus_int, nRshifts ); opus_int pulses_comb[ 8 ]; opus_int *abs_pulses_ptr; const opus_int8 *pulses_ptr; const opus_uint8 *cdf_ptr; const opus_uint8 *nBits_ptr; + SAVE_STACK; silk_memset( pulses_comb, 0, 8 * sizeof( opus_int ) ); /* Fixing Valgrind reported problem*/ @@ -90,6 +92,8 @@ void silk_encode_pulses( } /* Take the absolute value of the pulses */ + ALLOC( abs_pulses, iter * SHELL_CODEC_FRAME_LENGTH, opus_int ); + silk_assert( !( SHELL_CODEC_FRAME_LENGTH & 3 ) ); for( i = 0; i < iter * SHELL_CODEC_FRAME_LENGTH; i+=4 ) { abs_pulses[i+0] = ( opus_int )silk_abs( pulses[ i + 0 ] ); abs_pulses[i+1] = ( opus_int )silk_abs( pulses[ i + 1 ] ); @@ -98,6 +102,8 @@ void silk_encode_pulses( } /* Calc sum pulses per shell code frame */ + ALLOC( sum_pulses, iter, opus_int ); + ALLOC( nRshifts, iter, opus_int ); abs_pulses_ptr = abs_pulses; for( i = 0; i < iter; i++ ) { nRshifts[ i ] = 0; @@ -196,4 +202,5 @@ void silk_encode_pulses( /* Encode signs */ /****************/ silk_encode_signs( psRangeEnc, pulses, frame_length, signalType, quantOffsetType, sum_pulses ); + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/errors.h b/code/opus-1.1/silk/errors.h similarity index 99% rename from code/opus-1.0.2/silk/errors.h rename to code/opus-1.1/silk/errors.h index 0591e009..45070800 100644 --- a/code/opus-1.0.2/silk/errors.h +++ b/code/opus-1.1/silk/errors.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/LTP_analysis_filter_FIX.c b/code/opus-1.1/silk/fixed/LTP_analysis_filter_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/LTP_analysis_filter_FIX.c rename to code/opus-1.1/silk/fixed/LTP_analysis_filter_FIX.c index a8fee555..a9419080 100644 --- a/code/opus-1.0.2/silk/fixed/LTP_analysis_filter_FIX.c +++ b/code/opus-1.1/silk/fixed/LTP_analysis_filter_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/LTP_scale_ctrl_FIX.c b/code/opus-1.1/silk/fixed/LTP_scale_ctrl_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/LTP_scale_ctrl_FIX.c rename to code/opus-1.1/silk/fixed/LTP_scale_ctrl_FIX.c index ac2fba17..3dcedef8 100644 --- a/code/opus-1.0.2/silk/fixed/LTP_scale_ctrl_FIX.c +++ b/code/opus-1.1/silk/fixed/LTP_scale_ctrl_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/apply_sine_window_FIX.c b/code/opus-1.1/silk/fixed/apply_sine_window_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/apply_sine_window_FIX.c rename to code/opus-1.1/silk/fixed/apply_sine_window_FIX.c index 897fdc30..4502b713 100644 --- a/code/opus-1.0.2/silk/fixed/apply_sine_window_FIX.c +++ b/code/opus-1.1/silk/fixed/apply_sine_window_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/autocorr_FIX.c b/code/opus-1.1/silk/fixed/autocorr_FIX.c similarity index 64% rename from code/opus-1.0.2/silk/fixed/autocorr_FIX.c rename to code/opus-1.1/silk/fixed/autocorr_FIX.c index c2ebb6a9..de95c986 100644 --- a/code/opus-1.0.2/silk/fixed/autocorr_FIX.c +++ b/code/opus-1.1/silk/fixed/autocorr_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "SigProc_FIX.h" +#include "celt_lpc.h" /* Compute autocorrelation */ void silk_autocorr( @@ -37,40 +38,11 @@ void silk_autocorr( opus_int *scale, /* O Scaling of the correlation vector */ const opus_int16 *inputData, /* I Input data to correlate */ const opus_int inputDataSize, /* I Length of input */ - const opus_int correlationCount /* I Number of correlation taps to compute */ + const opus_int correlationCount, /* I Number of correlation taps to compute */ + int arch /* I Run-time architecture */ ) { - opus_int i, lz, nRightShifts, corrCount; - opus_int64 corr64; - + opus_int corrCount; corrCount = silk_min_int( inputDataSize, correlationCount ); - - /* compute energy (zero-lag correlation) */ - corr64 = silk_inner_prod16_aligned_64( inputData, inputData, inputDataSize ); - - /* deal with all-zero input data */ - corr64 += 1; - - /* number of leading zeros */ - lz = silk_CLZ64( corr64 ); - - /* scaling: number of right shifts applied to correlations */ - nRightShifts = 35 - lz; - *scale = nRightShifts; - - if( nRightShifts <= 0 ) { - results[ 0 ] = silk_LSHIFT( (opus_int32)silk_CHECK_FIT32( corr64 ), -nRightShifts ); - - /* compute remaining correlations based on int32 inner product */ - for( i = 1; i < corrCount; i++ ) { - results[ i ] = silk_LSHIFT( silk_inner_prod_aligned( inputData, inputData + i, inputDataSize - i ), -nRightShifts ); - } - } else { - results[ 0 ] = (opus_int32)silk_CHECK_FIT32( silk_RSHIFT64( corr64, nRightShifts ) ); - - /* compute remaining correlations based on int64 inner product */ - for( i = 1; i < corrCount; i++ ) { - results[ i ] = (opus_int32)silk_CHECK_FIT32( silk_RSHIFT64( silk_inner_prod16_aligned_64( inputData, inputData + i, inputDataSize - i ), nRightShifts ) ); - } - } + *scale = _celt_autocorr(inputData, results, NULL, 0, corrCount-1, inputDataSize, arch); } diff --git a/code/opus-1.0.2/silk/fixed/burg_modified_FIX.c b/code/opus-1.1/silk/fixed/burg_modified_FIX.c similarity index 94% rename from code/opus-1.0.2/silk/fixed/burg_modified_FIX.c rename to code/opus-1.1/silk/fixed/burg_modified_FIX.c index 26a66b1c..db348295 100644 --- a/code/opus-1.0.2/silk/fixed/burg_modified_FIX.c +++ b/code/opus-1.1/silk/fixed/burg_modified_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -32,6 +32,7 @@ POSSIBILITY OF SUCH DAMAGE. #include "SigProc_FIX.h" #include "define.h" #include "tuning_parameters.h" +#include "pitch.h" #define MAX_FRAME_SIZE 384 /* subfr_length * nb_subfr = ( 0.005 * 16000 + 16 ) * 4 = 384 */ @@ -49,7 +50,8 @@ void silk_burg_modified( const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ const opus_int nb_subfr, /* I Number of subframes stacked in x */ - const opus_int D /* I Order */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */ ) { opus_int k, n, s, lz, rshifts, rshifts_extra, reached_max_gain; @@ -60,6 +62,7 @@ void silk_burg_modified( opus_int32 Af_QA[ SILK_MAX_ORDER_LPC ]; opus_int32 CAf[ SILK_MAX_ORDER_LPC + 1 ]; opus_int32 CAb[ SILK_MAX_ORDER_LPC + 1 ]; + opus_int32 xcorr[ SILK_MAX_ORDER_LPC ]; silk_assert( subfr_length * nb_subfr <= MAX_FRAME_SIZE ); @@ -93,10 +96,17 @@ void silk_burg_modified( } } else { for( s = 0; s < nb_subfr; s++ ) { + int i; + opus_int32 d; x_ptr = x + s * subfr_length; + celt_pitch_xcorr(x_ptr, x_ptr + 1, xcorr, subfr_length - D, D, arch ); for( n = 1; n < D + 1; n++ ) { - C_first_row[ n - 1 ] += silk_LSHIFT32( - silk_inner_prod_aligned( x_ptr, x_ptr + n, subfr_length - n ), -rshifts ); + for ( i = n + subfr_length - D, d = 0; i < subfr_length; i++ ) + d = MAC16_16( d, x_ptr[ i ], x_ptr[ i - n ] ); + xcorr[ n - 1 ] += d; + } + for( n = 1; n < D + 1; n++ ) { + C_first_row[ n - 1 ] += silk_LSHIFT32( xcorr[ n - 1 ], -rshifts ); } } } @@ -263,7 +273,7 @@ void silk_burg_modified( tmp1 = silk_SMLAWW( tmp1, Atmp1, Atmp1 ); /* Q16 */ A_Q16[ k ] = -Atmp1; } - *res_nrg = silk_SMLAWW( nrg, silk_SMMUL( FIND_LPC_COND_FAC, C0 ), -tmp1 ); /* Q( -rshifts ) */ + *res_nrg = silk_SMLAWW( nrg, silk_SMMUL( SILK_FIX_CONST( FIND_LPC_COND_FAC, 32 ), C0 ), -tmp1 );/* Q( -rshifts ) */ *res_nrg_Q = -rshifts; - } + } } diff --git a/code/opus-1.0.2/silk/fixed/corrMatrix_FIX.c b/code/opus-1.1/silk/fixed/corrMatrix_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/corrMatrix_FIX.c rename to code/opus-1.1/silk/fixed/corrMatrix_FIX.c index 21502499..c6172705 100644 --- a/code/opus-1.0.2/silk/fixed/corrMatrix_FIX.c +++ b/code/opus-1.1/silk/fixed/corrMatrix_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/encode_frame_FIX.c b/code/opus-1.1/silk/fixed/encode_frame_FIX.c similarity index 95% rename from code/opus-1.0.2/silk/fixed/encode_frame_FIX.c rename to code/opus-1.1/silk/fixed/encode_frame_FIX.c index a37a9f21..b490986b 100644 --- a/code/opus-1.0.2/silk/fixed/encode_frame_FIX.c +++ b/code/opus-1.1/silk/fixed/encode_frame_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,10 +30,11 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" #include "tuning_parameters.h" /* Low Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode with lower bitrate */ -static inline void silk_LBRR_encode_FIX( +static OPUS_INLINE void silk_LBRR_encode_FIX( silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ silk_encoder_control_FIX *psEncCtrl, /* I/O Pointer to Silk FIX encoder control struct */ const opus_int32 xfw_Q3[], /* I Input signal */ @@ -84,9 +85,7 @@ opus_int silk_encode_frame_FIX( { silk_encoder_control_FIX sEncCtrl; opus_int i, iter, maxIter, found_upper, found_lower, ret = 0; - opus_int16 *x_frame, *res_pitch_frame; - opus_int32 xfw_Q3[ MAX_FRAME_LENGTH ]; - opus_int16 res_pitch[ 2 * MAX_FRAME_LENGTH + LA_PITCH_MAX ]; + opus_int16 *x_frame; ec_enc sRangeEnc_copy, sRangeEnc_copy2; silk_nsq_state sNSQ_copy, sNSQ_copy2; opus_int32 seed_copy, nBits, nBits_lower, nBits_upper, gainMult_lower, gainMult_upper; @@ -95,7 +94,7 @@ opus_int silk_encode_frame_FIX( opus_int16 ec_prevLagIndex_copy; opus_int ec_prevSignalType_copy; opus_int8 LastGainIndex_copy2; - opus_uint8 ec_buf_copy[ 1275 ]; + SAVE_STACK; /* This is totally unnecessary but many compilers (including gcc) are too dumb to realise it */ LastGainIndex_copy2 = nBits_lower = nBits_upper = gainMult_lower = gainMult_upper = 0; @@ -105,9 +104,8 @@ opus_int silk_encode_frame_FIX( /**************************************************************/ /* Set up Input Pointers, and insert frame in input buffer */ /*************************************************************/ - /* pointers aligned with start of frame to encode */ - x_frame = psEnc->x_buf + psEnc->sCmn.ltp_mem_length; /* start of frame to encode */ - res_pitch_frame = res_pitch + psEnc->sCmn.ltp_mem_length; /* start of pitch LPC residual frame */ + /* start of frame to encode */ + x_frame = psEnc->x_buf + psEnc->sCmn.ltp_mem_length; /***************************************/ /* Ensure smooth bandwidth transitions */ @@ -120,15 +118,26 @@ opus_int silk_encode_frame_FIX( silk_memcpy( x_frame + LA_SHAPE_MS * psEnc->sCmn.fs_kHz, psEnc->sCmn.inputBuf + 1, psEnc->sCmn.frame_length * sizeof( opus_int16 ) ); if( !psEnc->sCmn.prefillFlag ) { + VARDECL( opus_int32, xfw_Q3 ); + VARDECL( opus_int16, res_pitch ); + VARDECL( opus_uint8, ec_buf_copy ); + opus_int16 *res_pitch_frame; + + ALLOC( res_pitch, + psEnc->sCmn.la_pitch + psEnc->sCmn.frame_length + + psEnc->sCmn.ltp_mem_length, opus_int16 ); + /* start of pitch LPC residual frame */ + res_pitch_frame = res_pitch + psEnc->sCmn.ltp_mem_length; + /*****************************************/ /* Find pitch lags, initial LPC analysis */ /*****************************************/ - silk_find_pitch_lags_FIX( psEnc, &sEncCtrl, res_pitch, x_frame ); + silk_find_pitch_lags_FIX( psEnc, &sEncCtrl, res_pitch, x_frame, psEnc->sCmn.arch ); /************************/ /* Noise shape analysis */ /************************/ - silk_noise_shape_analysis_FIX( psEnc, &sEncCtrl, res_pitch_frame, x_frame ); + silk_noise_shape_analysis_FIX( psEnc, &sEncCtrl, res_pitch_frame, x_frame, psEnc->sCmn.arch ); /***************************************************/ /* Find linear prediction coefficients (LPC + LTP) */ @@ -143,6 +152,7 @@ opus_int silk_encode_frame_FIX( /*****************************************/ /* Prefiltering for noise shaper */ /*****************************************/ + ALLOC( xfw_Q3, psEnc->sCmn.frame_length, opus_int32 ); silk_prefilter_FIX( psEnc, &sEncCtrl, xfw_Q3, x_frame ); /****************************************/ @@ -164,6 +174,7 @@ opus_int silk_encode_frame_FIX( seed_copy = psEnc->sCmn.indices.Seed; ec_prevLagIndex_copy = psEnc->sCmn.ec_prevLagIndex; ec_prevSignalType_copy = psEnc->sCmn.ec_prevSignalType; + ALLOC( ec_buf_copy, 1275, opus_uint8 ); for( iter = 0; ; iter++ ) { if( gainsID == gainsID_lower ) { nBits = nBits_lower; @@ -291,17 +302,18 @@ opus_int silk_encode_frame_FIX( silk_memmove( psEnc->x_buf, &psEnc->x_buf[ psEnc->sCmn.frame_length ], ( psEnc->sCmn.ltp_mem_length + LA_SHAPE_MS * psEnc->sCmn.fs_kHz ) * sizeof( opus_int16 ) ); - /* Parameters needed for next frame */ - psEnc->sCmn.prevLag = sEncCtrl.pitchL[ psEnc->sCmn.nb_subfr - 1 ]; - psEnc->sCmn.prevSignalType = psEnc->sCmn.indices.signalType; - /* Exit without entropy coding */ if( psEnc->sCmn.prefillFlag ) { /* No payload */ *pnBytesOut = 0; + RESTORE_STACK; return ret; } + /* Parameters needed for next frame */ + psEnc->sCmn.prevLag = sEncCtrl.pitchL[ psEnc->sCmn.nb_subfr - 1 ]; + psEnc->sCmn.prevSignalType = psEnc->sCmn.indices.signalType; + /****************************************/ /* Finalize payload */ /****************************************/ @@ -309,11 +321,12 @@ opus_int silk_encode_frame_FIX( /* Payload size */ *pnBytesOut = silk_RSHIFT( ec_tell( psRangeEnc ) + 7, 3 ); + RESTORE_STACK; return ret; } /* Low-Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode excitation at lower bitrate */ -static inline void silk_LBRR_encode_FIX( +static OPUS_INLINE void silk_LBRR_encode_FIX( silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ silk_encoder_control_FIX *psEncCtrl, /* I/O Pointer to Silk FIX encoder control struct */ const opus_int32 xfw_Q3[], /* I Input signal */ diff --git a/code/opus-1.0.2/silk/fixed/find_LPC_FIX.c b/code/opus-1.1/silk/fixed/find_LPC_FIX.c similarity index 96% rename from code/opus-1.0.2/silk/fixed/find_LPC_FIX.c rename to code/opus-1.1/silk/fixed/find_LPC_FIX.c index 0ed7e846..783d32e2 100644 --- a/code/opus-1.0.2/silk/fixed/find_LPC_FIX.c +++ b/code/opus-1.1/silk/fixed/find_LPC_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" #include "tuning_parameters.h" /* Finds LPC vector from correlations, and converts to NLSF */ @@ -51,7 +52,7 @@ void silk_find_LPC_FIX( opus_int res_nrg_interp_Q, res_nrg_Q, res_tmp_nrg_Q; opus_int16 a_tmp_Q12[ MAX_LPC_ORDER ]; opus_int16 NLSF0_Q15[ MAX_LPC_ORDER ]; - opus_int16 LPC_res[ MAX_FRAME_LENGTH + MAX_NB_SUBFR * MAX_LPC_ORDER ]; + SAVE_STACK; subfr_length = psEncC->subfr_length + psEncC->predictLPCOrder; @@ -59,11 +60,13 @@ void silk_find_LPC_FIX( psEncC->indices.NLSFInterpCoef_Q2 = 4; /* Burg AR analysis for the full frame */ - silk_burg_modified( &res_nrg, &res_nrg_Q, a_Q16, x, minInvGain_Q30, subfr_length, psEncC->nb_subfr, psEncC->predictLPCOrder ); + silk_burg_modified( &res_nrg, &res_nrg_Q, a_Q16, x, minInvGain_Q30, subfr_length, psEncC->nb_subfr, psEncC->predictLPCOrder, psEncC->arch ); if( psEncC->useInterpolatedNLSFs && !psEncC->first_frame_after_reset && psEncC->nb_subfr == MAX_NB_SUBFR ) { + VARDECL( opus_int16, LPC_res ); + /* Optimal solution for last 10 ms */ - silk_burg_modified( &res_tmp_nrg, &res_tmp_nrg_Q, a_tmp_Q16, x + 2 * subfr_length, minInvGain_Q30, subfr_length, 2, psEncC->predictLPCOrder ); + silk_burg_modified( &res_tmp_nrg, &res_tmp_nrg_Q, a_tmp_Q16, x + 2 * subfr_length, minInvGain_Q30, subfr_length, 2, psEncC->predictLPCOrder, psEncC->arch ); /* subtract residual energy here, as that's easier than adding it to the */ /* residual energy of the first 10 ms in each iteration of the search below */ @@ -81,6 +84,8 @@ void silk_find_LPC_FIX( /* Convert to NLSFs */ silk_A2NLSF( NLSF_Q15, a_tmp_Q16, psEncC->predictLPCOrder ); + ALLOC( LPC_res, 2 * subfr_length, opus_int16 ); + /* Search over interpolation indices to find the one with lowest residual energy */ for( k = 3; k >= 0; k-- ) { /* Interpolate NLSFs for first half */ @@ -142,4 +147,5 @@ void silk_find_LPC_FIX( } silk_assert( psEncC->indices.NLSFInterpCoef_Q2 == 4 || ( psEncC->useInterpolatedNLSFs && !psEncC->first_frame_after_reset && psEncC->nb_subfr == MAX_NB_SUBFR ) ); + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/fixed/find_LTP_FIX.c b/code/opus-1.1/silk/fixed/find_LTP_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/find_LTP_FIX.c rename to code/opus-1.1/silk/fixed/find_LTP_FIX.c index bd210874..8c4d7037 100644 --- a/code/opus-1.0.2/silk/fixed/find_LTP_FIX.c +++ b/code/opus-1.1/silk/fixed/find_LTP_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/find_pitch_lags_FIX.c b/code/opus-1.1/silk/fixed/find_pitch_lags_FIX.c similarity index 83% rename from code/opus-1.0.2/silk/fixed/find_pitch_lags_FIX.c rename to code/opus-1.1/silk/fixed/find_pitch_lags_FIX.c index 39c30487..620f8dcd 100644 --- a/code/opus-1.0.2/silk/fixed/find_pitch_lags_FIX.c +++ b/code/opus-1.1/silk/fixed/find_pitch_lags_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" #include "tuning_parameters.h" /* Find pitch lags */ @@ -37,17 +38,20 @@ void silk_find_pitch_lags_FIX( silk_encoder_state_FIX *psEnc, /* I/O encoder state */ silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ opus_int16 res[], /* O residual */ - const opus_int16 x[] /* I Speech signal */ + const opus_int16 x[], /* I Speech signal */ + int arch /* I Run-time architecture */ ) { opus_int buf_len, i, scale; - opus_int32 thrhld_Q15, res_nrg; + opus_int32 thrhld_Q13, res_nrg; const opus_int16 *x_buf, *x_buf_ptr; - opus_int16 Wsig[ FIND_PITCH_LPC_WIN_MAX ], *Wsig_ptr; + VARDECL( opus_int16, Wsig ); + opus_int16 *Wsig_ptr; opus_int32 auto_corr[ MAX_FIND_PITCH_LPC_ORDER + 1 ]; opus_int16 rc_Q15[ MAX_FIND_PITCH_LPC_ORDER ]; opus_int32 A_Q24[ MAX_FIND_PITCH_LPC_ORDER ]; opus_int16 A_Q12[ MAX_FIND_PITCH_LPC_ORDER ]; + SAVE_STACK; /******************************************/ /* Set up buffer lengths etc based on Fs */ @@ -65,6 +69,8 @@ void silk_find_pitch_lags_FIX( /* Calculate windowed signal */ + ALLOC( Wsig, psEnc->sCmn.pitch_LPC_win_length, opus_int16 ); + /* First LA_LTP samples */ x_buf_ptr = x_buf + buf_len - psEnc->sCmn.pitch_LPC_win_length; Wsig_ptr = Wsig; @@ -81,7 +87,7 @@ void silk_find_pitch_lags_FIX( silk_apply_sine_window( Wsig_ptr, x_buf_ptr, 2, psEnc->sCmn.la_pitch ); /* Calculate autocorrelation sequence */ - silk_autocorr( auto_corr, &scale, Wsig, psEnc->sCmn.pitch_LPC_win_length, psEnc->sCmn.pitchEstimationLPCOrder + 1 ); + silk_autocorr( auto_corr, &scale, Wsig, psEnc->sCmn.pitch_LPC_win_length, psEnc->sCmn.pitchEstimationLPCOrder + 1, arch ); /* Add white noise, as fraction of energy */ auto_corr[ 0 ] = silk_SMLAWB( auto_corr[ 0 ], auto_corr[ 0 ], SILK_FIX_CONST( FIND_PITCH_WHITE_NOISE_FRACTION, 16 ) ) + 1; @@ -110,19 +116,20 @@ void silk_find_pitch_lags_FIX( if( psEnc->sCmn.indices.signalType != TYPE_NO_VOICE_ACTIVITY && psEnc->sCmn.first_frame_after_reset == 0 ) { /* Threshold for pitch estimator */ - thrhld_Q15 = SILK_FIX_CONST( 0.6, 15 ); - thrhld_Q15 = silk_SMLABB( thrhld_Q15, SILK_FIX_CONST( -0.004, 15 ), psEnc->sCmn.pitchEstimationLPCOrder ); - thrhld_Q15 = silk_SMLABB( thrhld_Q15, SILK_FIX_CONST( -0.1, 7 ), psEnc->sCmn.speech_activity_Q8 ); - thrhld_Q15 = silk_SMLABB( thrhld_Q15, SILK_FIX_CONST( -0.15, 15 ), silk_RSHIFT( psEnc->sCmn.prevSignalType, 1 ) ); - thrhld_Q15 = silk_SMLAWB( thrhld_Q15, SILK_FIX_CONST( -0.1, 16 ), psEnc->sCmn.input_tilt_Q15 ); - thrhld_Q15 = silk_SAT16( thrhld_Q15 ); + thrhld_Q13 = SILK_FIX_CONST( 0.6, 13 ); + thrhld_Q13 = silk_SMLABB( thrhld_Q13, SILK_FIX_CONST( -0.004, 13 ), psEnc->sCmn.pitchEstimationLPCOrder ); + thrhld_Q13 = silk_SMLAWB( thrhld_Q13, SILK_FIX_CONST( -0.1, 21 ), psEnc->sCmn.speech_activity_Q8 ); + thrhld_Q13 = silk_SMLABB( thrhld_Q13, SILK_FIX_CONST( -0.15, 13 ), silk_RSHIFT( psEnc->sCmn.prevSignalType, 1 ) ); + thrhld_Q13 = silk_SMLAWB( thrhld_Q13, SILK_FIX_CONST( -0.1, 14 ), psEnc->sCmn.input_tilt_Q15 ); + thrhld_Q13 = silk_SAT16( thrhld_Q13 ); /*****************************************/ /* Call pitch estimator */ /*****************************************/ if( silk_pitch_analysis_core( res, psEncCtrl->pitchL, &psEnc->sCmn.indices.lagIndex, &psEnc->sCmn.indices.contourIndex, &psEnc->LTPCorr_Q15, psEnc->sCmn.prevLag, psEnc->sCmn.pitchEstimationThreshold_Q16, - (opus_int16)thrhld_Q15, psEnc->sCmn.fs_kHz, psEnc->sCmn.pitchEstimationComplexity, psEnc->sCmn.nb_subfr ) == 0 ) + (opus_int)thrhld_Q13, psEnc->sCmn.fs_kHz, psEnc->sCmn.pitchEstimationComplexity, psEnc->sCmn.nb_subfr, + psEnc->sCmn.arch) == 0 ) { psEnc->sCmn.indices.signalType = TYPE_VOICED; } else { @@ -134,4 +141,5 @@ void silk_find_pitch_lags_FIX( psEnc->sCmn.indices.contourIndex = 0; psEnc->LTPCorr_Q15 = 0; } + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/fixed/find_pred_coefs_FIX.c b/code/opus-1.1/silk/fixed/find_pred_coefs_FIX.c similarity index 92% rename from code/opus-1.0.2/silk/fixed/find_pred_coefs_FIX.c rename to code/opus-1.1/silk/fixed/find_pred_coefs_FIX.c index 997989b5..5c22f828 100644 --- a/code/opus-1.0.2/silk/fixed/find_pred_coefs_FIX.c +++ b/code/opus-1.1/silk/fixed/find_pred_coefs_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" void silk_find_pred_coefs_FIX( silk_encoder_state_FIX *psEnc, /* I/O encoder state */ @@ -40,13 +41,14 @@ void silk_find_pred_coefs_FIX( ) { opus_int i; - opus_int32 WLTP[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ]; opus_int32 invGains_Q16[ MAX_NB_SUBFR ], local_gains[ MAX_NB_SUBFR ], Wght_Q15[ MAX_NB_SUBFR ]; opus_int16 NLSF_Q15[ MAX_LPC_ORDER ]; const opus_int16 *x_ptr; - opus_int16 *x_pre_ptr, LPC_in_pre[ MAX_NB_SUBFR * MAX_LPC_ORDER + MAX_FRAME_LENGTH ]; + opus_int16 *x_pre_ptr; + VARDECL( opus_int16, LPC_in_pre ); opus_int32 tmp, min_gain_Q16, minInvGain_Q30; opus_int LTP_corrs_rshift[ MAX_NB_SUBFR ]; + SAVE_STACK; /* weighting for weighted least squares */ min_gain_Q16 = silk_int32_MAX >> 6; @@ -71,12 +73,19 @@ void silk_find_pred_coefs_FIX( local_gains[ i ] = silk_DIV32( ( (opus_int32)1 << 16 ), invGains_Q16[ i ] ); } + ALLOC( LPC_in_pre, + psEnc->sCmn.nb_subfr * psEnc->sCmn.predictLPCOrder + + psEnc->sCmn.frame_length, opus_int16 ); if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + VARDECL( opus_int32, WLTP ); + /**********/ /* VOICED */ /**********/ silk_assert( psEnc->sCmn.ltp_mem_length - psEnc->sCmn.predictLPCOrder >= psEncCtrl->pitchL[ 0 ] + LTP_ORDER / 2 ); + ALLOC( WLTP, psEnc->sCmn.nb_subfr * LTP_ORDER * LTP_ORDER, opus_int32 ); + /* LTP analysis */ silk_find_LTP_FIX( psEncCtrl->LTPCoef_Q14, WLTP, &psEncCtrl->LTPredCodGain_Q7, res_pitch, psEncCtrl->pitchL, Wght_Q15, psEnc->sCmn.subfr_length, @@ -84,7 +93,7 @@ void silk_find_pred_coefs_FIX( /* Quantize LTP gain parameters */ silk_quant_LTP_gains( psEncCtrl->LTPCoef_Q14, psEnc->sCmn.indices.LTPIndex, &psEnc->sCmn.indices.PERIndex, - WLTP, psEnc->sCmn.mu_LTP_Q9, psEnc->sCmn.LTPQuantLowComplexity, psEnc->sCmn.nb_subfr); + &psEnc->sCmn.sum_log_gain_Q7, WLTP, psEnc->sCmn.mu_LTP_Q9, psEnc->sCmn.LTPQuantLowComplexity, psEnc->sCmn.nb_subfr); /* Control LTP scaling */ silk_LTP_scale_ctrl_FIX( psEnc, psEncCtrl, condCoding ); @@ -109,6 +118,7 @@ void silk_find_pred_coefs_FIX( silk_memset( psEncCtrl->LTPCoef_Q14, 0, psEnc->sCmn.nb_subfr * LTP_ORDER * sizeof( opus_int16 ) ); psEncCtrl->LTPredCodGain_Q7 = 0; + psEnc->sCmn.sum_log_gain_Q7 = 0; } /* Limit on total predictive coding gain */ @@ -133,4 +143,5 @@ void silk_find_pred_coefs_FIX( /* Copy to prediction struct for use in next frame for interpolation */ silk_memcpy( psEnc->sCmn.prev_NLSFq_Q15, NLSF_Q15, sizeof( psEnc->sCmn.prev_NLSFq_Q15 ) ); + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/fixed/k2a_FIX.c b/code/opus-1.1/silk/fixed/k2a_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/k2a_FIX.c rename to code/opus-1.1/silk/fixed/k2a_FIX.c index cadc9274..5fee599b 100644 --- a/code/opus-1.0.2/silk/fixed/k2a_FIX.c +++ b/code/opus-1.1/silk/fixed/k2a_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/k2a_Q16_FIX.c b/code/opus-1.1/silk/fixed/k2a_Q16_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/k2a_Q16_FIX.c rename to code/opus-1.1/silk/fixed/k2a_Q16_FIX.c index f96f3064..3b039875 100644 --- a/code/opus-1.0.2/silk/fixed/k2a_Q16_FIX.c +++ b/code/opus-1.1/silk/fixed/k2a_Q16_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/main_FIX.h b/code/opus-1.1/silk/fixed/main_FIX.h similarity index 96% rename from code/opus-1.0.2/silk/fixed/main_FIX.h rename to code/opus-1.1/silk/fixed/main_FIX.h index 369b31ee..a56ca07a 100644 --- a/code/opus-1.0.2/silk/fixed/main_FIX.h +++ b/code/opus-1.1/silk/fixed/main_FIX.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -73,7 +73,8 @@ opus_int silk_encode_frame_FIX( /* Initializes the Silk encoder state */ opus_int silk_init_encoder( - silk_encoder_state_Fxx *psEnc /* I/O Pointer to Silk FIX encoder state */ + silk_encoder_state_Fxx *psEnc, /* I/O Pointer to Silk FIX encoder state */ + int arch /* I Run-time architecture */ ); /* Control the Silk encoder */ @@ -104,7 +105,8 @@ void silk_noise_shape_analysis_FIX( silk_encoder_state_FIX *psEnc, /* I/O Encoder state FIX */ silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control FIX */ const opus_int16 *pitch_res, /* I LPC residual from pitch analysis */ - const opus_int16 *x /* I Input signal [ frame_length + la_shape ] */ + const opus_int16 *x, /* I Input signal [ frame_length + la_shape ] */ + int arch /* I Run-time architecture */ ); /* Autocorrelations for a warped frequency axis */ @@ -132,7 +134,8 @@ void silk_find_pitch_lags_FIX( silk_encoder_state_FIX *psEnc, /* I/O encoder state */ silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ opus_int16 res[], /* O residual */ - const opus_int16 x[] /* I Speech signal */ + const opus_int16 x[], /* I Speech signal */ + int arch /* I Run-time architecture */ ); /* Find LPC and LTP coefficients */ diff --git a/code/opus-1.0.2/silk/fixed/noise_shape_analysis_FIX.c b/code/opus-1.1/silk/fixed/noise_shape_analysis_FIX.c similarity index 97% rename from code/opus-1.0.2/silk/fixed/noise_shape_analysis_FIX.c rename to code/opus-1.1/silk/fixed/noise_shape_analysis_FIX.c index d230e48d..e24d2e9d 100644 --- a/code/opus-1.0.2/silk/fixed/noise_shape_analysis_FIX.c +++ b/code/opus-1.1/silk/fixed/noise_shape_analysis_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,13 +30,14 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" #include "tuning_parameters.h" /* Compute gain to make warped filter coefficients have a zero mean log frequency response on a */ /* non-warped frequency scale. (So that it can be implemented with a minimum-phase monic filter.) */ /* Note: A monic filter is one with the first coefficient equal to 1.0. In Silk we omit the first */ /* coefficient in an array of coefficients, for monic filters. */ -static inline opus_int32 warped_gain( /* gain in Q16*/ +static OPUS_INLINE opus_int32 warped_gain( /* gain in Q16*/ const opus_int32 *coefs_Q24, opus_int lambda_Q16, opus_int order @@ -55,7 +56,7 @@ static inline opus_int32 warped_gain( /* gain in Q16*/ /* Convert warped filter coefficients to monic pseudo-warped coefficients and limit maximum */ /* amplitude of monic warped coefficients by using bandwidth expansion on the true coefficients */ -static inline void limit_warped_coefs( +static OPUS_INLINE void limit_warped_coefs( opus_int32 *coefs_syn_Q24, opus_int32 *coefs_ana_Q24, opus_int lambda_Q16, @@ -144,7 +145,8 @@ void silk_noise_shape_analysis_FIX( silk_encoder_state_FIX *psEnc, /* I/O Encoder state FIX */ silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control FIX */ const opus_int16 *pitch_res, /* I LPC residual from pitch analysis */ - const opus_int16 *x /* I Input signal [ frame_length + la_shape ] */ + const opus_int16 *x, /* I Input signal [ frame_length + la_shape ] */ + int arch /* I Run-time architecture */ ) { silk_shape_state_FIX *psShapeSt = &psEnc->sShape; @@ -156,8 +158,9 @@ void silk_noise_shape_analysis_FIX( opus_int32 refl_coef_Q16[ MAX_SHAPE_LPC_ORDER ]; opus_int32 AR1_Q24[ MAX_SHAPE_LPC_ORDER ]; opus_int32 AR2_Q24[ MAX_SHAPE_LPC_ORDER ]; - opus_int16 x_windowed[ SHAPE_LPC_WIN_MAX ]; + VARDECL( opus_int16, x_windowed ); const opus_int16 *x_ptr, *pitch_res_ptr; + SAVE_STACK; /* Point to start of first LPC analysis block */ x_ptr = x - psEnc->sCmn.la_shape; @@ -258,6 +261,7 @@ void silk_noise_shape_analysis_FIX( /********************************************/ /* Compute noise shaping AR coefs and gains */ /********************************************/ + ALLOC( x_windowed, psEnc->sCmn.shapeWinLength, opus_int16 ); for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { /* Apply window: sine slope followed by flat part followed by cosine slope */ opus_int shift, slope_part, flat_part; @@ -278,7 +282,7 @@ void silk_noise_shape_analysis_FIX( silk_warped_autocorrelation_FIX( auto_corr, &scale, x_windowed, warping_Q16, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder ); } else { /* Calculate regular auto correlation */ - silk_autocorr( auto_corr, &scale, x_windowed, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder + 1 ); + silk_autocorr( auto_corr, &scale, x_windowed, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder + 1, arch ); } /* Add white noise, as a fraction of energy */ @@ -437,4 +441,5 @@ void silk_noise_shape_analysis_FIX( psEncCtrl->HarmShapeGain_Q14[ k ] = ( opus_int )silk_RSHIFT_ROUND( psShapeSt->HarmShapeGain_smth_Q16, 2 ); psEncCtrl->Tilt_Q14[ k ] = ( opus_int )silk_RSHIFT_ROUND( psShapeSt->Tilt_smth_Q16, 2 ); } + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/fixed/pitch_analysis_core_FIX.c b/code/opus-1.1/silk/fixed/pitch_analysis_core_FIX.c similarity index 63% rename from code/opus-1.0.2/silk/fixed/pitch_analysis_core_FIX.c rename to code/opus-1.1/silk/fixed/pitch_analysis_core_FIX.c index d43f444d..1641a0fb 100644 --- a/code/opus-1.0.2/silk/fixed/pitch_analysis_core_FIX.c +++ b/code/opus-1.1/silk/fixed/pitch_analysis_core_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -34,24 +34,40 @@ POSSIBILITY OF SUCH DAMAGE. ********************************************************** */ #include "SigProc_FIX.h" #include "pitch_est_defines.h" +#include "stack_alloc.h" #include "debug.h" +#include "pitch.h" #define SCRATCH_SIZE 22 +#define SF_LENGTH_4KHZ ( PE_SUBFR_LENGTH_MS * 4 ) +#define SF_LENGTH_8KHZ ( PE_SUBFR_LENGTH_MS * 8 ) +#define MIN_LAG_4KHZ ( PE_MIN_LAG_MS * 4 ) +#define MIN_LAG_8KHZ ( PE_MIN_LAG_MS * 8 ) +#define MAX_LAG_4KHZ ( PE_MAX_LAG_MS * 4 ) +#define MAX_LAG_8KHZ ( PE_MAX_LAG_MS * 8 - 1 ) +#define CSTRIDE_4KHZ ( MAX_LAG_4KHZ + 1 - MIN_LAG_4KHZ ) +#define CSTRIDE_8KHZ ( MAX_LAG_8KHZ + 3 - ( MIN_LAG_8KHZ - 2 ) ) +#define D_COMP_MIN ( MIN_LAG_8KHZ - 3 ) +#define D_COMP_MAX ( MAX_LAG_8KHZ + 4 ) +#define D_COMP_STRIDE ( D_COMP_MAX - D_COMP_MIN ) + +typedef opus_int32 silk_pe_stage3_vals[ PE_NB_STAGE3_LAGS ]; /************************************************************/ /* Internally used functions */ /************************************************************/ -void silk_P_Ana_calc_corr_st3( - opus_int32 cross_corr_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ],/* (O) 3 DIM correlation array */ +static void silk_P_Ana_calc_corr_st3( + silk_pe_stage3_vals cross_corr_st3[], /* O 3 DIM correlation array */ const opus_int16 frame[], /* I vector to correlate */ opus_int start_lag, /* I lag offset to search around */ opus_int sf_length, /* I length of a 5 ms subframe */ opus_int nb_subfr, /* I number of subframes */ - opus_int complexity /* I Complexity setting */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ ); -void silk_P_Ana_calc_energy_st3( - opus_int32 energies_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ],/* (O) 3 DIM energy array */ +static void silk_P_Ana_calc_energy_st3( + silk_pe_stage3_vals energies_st3[], /* O 3 DIM energy array */ const opus_int16 frame[], /* I vector to calc energy in */ opus_int start_lag, /* I lag offset to search around */ opus_int sf_length, /* I length of one 5 ms subframe */ @@ -59,12 +75,6 @@ void silk_P_Ana_calc_energy_st3( opus_int complexity /* I Complexity setting */ ); -opus_int32 silk_P_Ana_find_scaling( - const opus_int16 *frame, - const opus_int frame_length, - const opus_int sum_sqr_len -); - /*************************************************************/ /* FIXED POINT CORE PITCH ANALYSIS FUNCTION */ /*************************************************************/ @@ -76,36 +86,38 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 opus_int *LTPCorr_Q15, /* I/O Normalized correlation; input: value from previous frame */ opus_int prevLag, /* I Last lag of previous frame; set to zero is unvoiced */ const opus_int32 search_thres1_Q16, /* I First stage threshold for lag candidates 0 - 1 */ - const opus_int search_thres2_Q15, /* I Final threshold for lag candidates 0 - 1 */ + const opus_int search_thres2_Q13, /* I Final threshold for lag candidates 0 - 1 */ const opus_int Fs_kHz, /* I Sample frequency (kHz) */ const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ - const opus_int nb_subfr /* I number of 5 ms subframes */ + const opus_int nb_subfr, /* I number of 5 ms subframes */ + int arch /* I Run-time architecture */ ) { - opus_int16 frame_8kHz[ PE_MAX_FRAME_LENGTH_ST_2 ]; - opus_int16 frame_4kHz[ PE_MAX_FRAME_LENGTH_ST_1 ]; + VARDECL( opus_int16, frame_8kHz ); + VARDECL( opus_int16, frame_4kHz ); opus_int32 filt_state[ 6 ]; - opus_int32 scratch_mem[ 3 * PE_MAX_FRAME_LENGTH ]; - opus_int16 *input_frame_ptr; + const opus_int16 *input_frame_ptr; opus_int i, k, d, j; - opus_int16 C[ PE_MAX_NB_SUBFR ][ ( PE_MAX_LAG >> 1 ) + 5 ]; + VARDECL( opus_int16, C ); + VARDECL( opus_int32, xcorr32 ); const opus_int16 *target_ptr, *basis_ptr; opus_int32 cross_corr, normalizer, energy, shift, energy_basis, energy_target; opus_int d_srch[ PE_D_SRCH_LENGTH ], Cmax, length_d_srch, length_d_comp; - opus_int16 d_comp[ ( PE_MAX_LAG >> 1 ) + 5 ]; - opus_int32 sum, threshold, temp32, lag_counter; + VARDECL( opus_int16, d_comp ); + opus_int32 sum, threshold, lag_counter; opus_int CBimax, CBimax_new, CBimax_old, lag, start_lag, end_lag, lag_new; opus_int32 CC[ PE_NB_CBKS_STAGE2_EXT ], CCmax, CCmax_b, CCmax_new_b, CCmax_new; - opus_int32 energies_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ]; - opus_int32 crosscorr_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ]; - opus_int frame_length, frame_length_8kHz, frame_length_4kHz, max_sum_sq_length; - opus_int sf_length, sf_length_8kHz, sf_length_4kHz; - opus_int min_lag, min_lag_8kHz, min_lag_4kHz; - opus_int max_lag, max_lag_8kHz, max_lag_4kHz; - opus_int32 contour_bias_Q20, diff, lz, lshift; + VARDECL( silk_pe_stage3_vals, energies_st3 ); + VARDECL( silk_pe_stage3_vals, cross_corr_st3 ); + opus_int frame_length, frame_length_8kHz, frame_length_4kHz; + opus_int sf_length; + opus_int min_lag; + opus_int max_lag; + opus_int32 contour_bias_Q15, diff; opus_int nb_cbk_search, cbk_size; - opus_int32 delta_lag_log2_sqr_Q7, lag_log2_Q7, prevLag_log2_Q7, prev_lag_bias_Q15, corr_thres_Q15; + opus_int32 delta_lag_log2_sqr_Q7, lag_log2_Q7, prevLag_log2_Q7, prev_lag_bias_Q13; const opus_int8 *Lag_CB_ptr; + SAVE_STACK; /* Check for valid sampling frequency */ silk_assert( Fs_kHz == 8 || Fs_kHz == 12 || Fs_kHz == 16 ); @@ -114,25 +126,18 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 silk_assert( complexity <= SILK_PE_MAX_COMPLEX ); silk_assert( search_thres1_Q16 >= 0 && search_thres1_Q16 <= (1<<16) ); - silk_assert( search_thres2_Q15 >= 0 && search_thres2_Q15 <= (1<<15) ); + silk_assert( search_thres2_Q13 >= 0 && search_thres2_Q13 <= (1<<13) ); /* Set up frame lengths max / min lag for the sampling frequency */ frame_length = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * Fs_kHz; frame_length_4kHz = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * 4; frame_length_8kHz = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * 8; sf_length = PE_SUBFR_LENGTH_MS * Fs_kHz; - sf_length_4kHz = PE_SUBFR_LENGTH_MS * 4; - sf_length_8kHz = PE_SUBFR_LENGTH_MS * 8; min_lag = PE_MIN_LAG_MS * Fs_kHz; - min_lag_4kHz = PE_MIN_LAG_MS * 4; - min_lag_8kHz = PE_MIN_LAG_MS * 8; max_lag = PE_MAX_LAG_MS * Fs_kHz - 1; - max_lag_4kHz = PE_MAX_LAG_MS * 4; - max_lag_8kHz = PE_MAX_LAG_MS * 8 - 1; - - silk_memset( C, 0, sizeof( opus_int16 ) * nb_subfr * ( ( PE_MAX_LAG >> 1 ) + 5) ); /* Resample from input sampled at Fs_kHz to 8 kHz */ + ALLOC( frame_8kHz, frame_length_8kHz, opus_int16 ); if( Fs_kHz == 16 ) { silk_memset( filt_state, 0, 2 * sizeof( opus_int32 ) ); silk_resampler_down2( filt_state, frame_8kHz, frame, frame_length ); @@ -146,6 +151,7 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 /* Decimate again to 4 kHz */ silk_memset( filt_state, 0, 2 * sizeof( opus_int32 ) );/* Set state to zero */ + ALLOC( frame_4kHz, frame_length_4kHz, opus_int16 ); silk_resampler_down2( filt_state, frame_4kHz, frame_8kHz, frame_length_8kHz ); /* Low-pass filter */ @@ -159,9 +165,9 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 *******************************************************************************/ /* Inner product is calculated with different lengths, so scale for the worst case */ - max_sum_sq_length = silk_max_32( sf_length_8kHz, silk_LSHIFT( sf_length_4kHz, 2 ) ); - shift = silk_P_Ana_find_scaling( frame_4kHz, frame_length_4kHz, max_sum_sq_length ); + silk_sum_sqr_shift( &energy, &shift, frame_4kHz, frame_length_4kHz ); if( shift > 0 ) { + shift = silk_RSHIFT( shift, 1 ); for( i = 0; i < frame_length_4kHz; i++ ) { frame_4kHz[ i ] = silk_RSHIFT( frame_4kHz[ i ], shift ); } @@ -170,94 +176,93 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 /****************************************************************************** * FIRST STAGE, operating in 4 khz ******************************************************************************/ - target_ptr = &frame_4kHz[ silk_LSHIFT( sf_length_4kHz, 2 ) ]; + ALLOC( C, nb_subfr * CSTRIDE_8KHZ, opus_int16 ); + ALLOC( xcorr32, MAX_LAG_4KHZ-MIN_LAG_4KHZ+1, opus_int32 ); + silk_memset( C, 0, (nb_subfr >> 1) * CSTRIDE_4KHZ * sizeof( opus_int16 ) ); + target_ptr = &frame_4kHz[ silk_LSHIFT( SF_LENGTH_4KHZ, 2 ) ]; for( k = 0; k < nb_subfr >> 1; k++ ) { /* Check that we are within range of the array */ silk_assert( target_ptr >= frame_4kHz ); - silk_assert( target_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); + silk_assert( target_ptr + SF_LENGTH_8KHZ <= frame_4kHz + frame_length_4kHz ); - basis_ptr = target_ptr - min_lag_4kHz; + basis_ptr = target_ptr - MIN_LAG_4KHZ; /* Check that we are within range of the array */ silk_assert( basis_ptr >= frame_4kHz ); - silk_assert( basis_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); + silk_assert( basis_ptr + SF_LENGTH_8KHZ <= frame_4kHz + frame_length_4kHz ); + + celt_pitch_xcorr( target_ptr, target_ptr - MAX_LAG_4KHZ, xcorr32, SF_LENGTH_8KHZ, MAX_LAG_4KHZ - MIN_LAG_4KHZ + 1, arch ); /* Calculate first vector products before loop */ - cross_corr = silk_inner_prod_aligned( target_ptr, basis_ptr, sf_length_8kHz ); - normalizer = silk_inner_prod_aligned( basis_ptr, basis_ptr, sf_length_8kHz ); - normalizer = silk_ADD_SAT32( normalizer, silk_SMULBB( sf_length_8kHz, 4000 ) ); + cross_corr = xcorr32[ MAX_LAG_4KHZ - MIN_LAG_4KHZ ]; + normalizer = silk_inner_prod_aligned( target_ptr, target_ptr, SF_LENGTH_8KHZ ); + normalizer = silk_ADD32( normalizer, silk_inner_prod_aligned( basis_ptr, basis_ptr, SF_LENGTH_8KHZ ) ); + normalizer = silk_ADD32( normalizer, silk_SMULBB( SF_LENGTH_8KHZ, 4000 ) ); - temp32 = silk_DIV32( cross_corr, silk_SQRT_APPROX( normalizer ) + 1 ); - C[ k ][ min_lag_4kHz ] = (opus_int16)silk_SAT16( temp32 ); /* Q0 */ + matrix_ptr( C, k, 0, CSTRIDE_4KHZ ) = + (opus_int16)silk_DIV32_varQ( cross_corr, normalizer, 13 + 1 ); /* Q13 */ /* From now on normalizer is computed recursively */ - for( d = min_lag_4kHz + 1; d <= max_lag_4kHz; d++ ) { + for( d = MIN_LAG_4KHZ + 1; d <= MAX_LAG_4KHZ; d++ ) { basis_ptr--; /* Check that we are within range of the array */ silk_assert( basis_ptr >= frame_4kHz ); - silk_assert( basis_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); + silk_assert( basis_ptr + SF_LENGTH_8KHZ <= frame_4kHz + frame_length_4kHz ); - cross_corr = silk_inner_prod_aligned( target_ptr, basis_ptr, sf_length_8kHz ); + cross_corr = xcorr32[ MAX_LAG_4KHZ - d ]; /* Add contribution of new sample and remove contribution from oldest sample */ - normalizer += + normalizer = silk_ADD32( normalizer, silk_SMULBB( basis_ptr[ 0 ], basis_ptr[ 0 ] ) - - silk_SMULBB( basis_ptr[ sf_length_8kHz ], basis_ptr[ sf_length_8kHz ] ); + silk_SMULBB( basis_ptr[ SF_LENGTH_8KHZ ], basis_ptr[ SF_LENGTH_8KHZ ] ) ); - temp32 = silk_DIV32( cross_corr, silk_SQRT_APPROX( normalizer ) + 1 ); - C[ k ][ d ] = (opus_int16)silk_SAT16( temp32 ); /* Q0 */ + matrix_ptr( C, k, d - MIN_LAG_4KHZ, CSTRIDE_4KHZ) = + (opus_int16)silk_DIV32_varQ( cross_corr, normalizer, 13 + 1 ); /* Q13 */ } /* Update target pointer */ - target_ptr += sf_length_8kHz; + target_ptr += SF_LENGTH_8KHZ; } /* Combine two subframes into single correlation measure and apply short-lag bias */ if( nb_subfr == PE_MAX_NB_SUBFR ) { - for( i = max_lag_4kHz; i >= min_lag_4kHz; i-- ) { - sum = (opus_int32)C[ 0 ][ i ] + (opus_int32)C[ 1 ][ i ]; /* Q0 */ - silk_assert( silk_RSHIFT( sum, 1 ) == silk_SAT16( silk_RSHIFT( sum, 1 ) ) ); - sum = silk_RSHIFT( sum, 1 ); /* Q-1 */ - silk_assert( silk_LSHIFT( (opus_int32)-i, 4 ) == silk_SAT16( silk_LSHIFT( (opus_int32)-i, 4 ) ) ); - sum = silk_SMLAWB( sum, sum, silk_LSHIFT( -i, 4 ) ); /* Q-1 */ - silk_assert( sum == silk_SAT16( sum ) ); - C[ 0 ][ i ] = (opus_int16)sum; /* Q-1 */ + for( i = MAX_LAG_4KHZ; i >= MIN_LAG_4KHZ; i-- ) { + sum = (opus_int32)matrix_ptr( C, 0, i - MIN_LAG_4KHZ, CSTRIDE_4KHZ ) + + (opus_int32)matrix_ptr( C, 1, i - MIN_LAG_4KHZ, CSTRIDE_4KHZ ); /* Q14 */ + sum = silk_SMLAWB( sum, sum, silk_LSHIFT( -i, 4 ) ); /* Q14 */ + C[ i - MIN_LAG_4KHZ ] = (opus_int16)sum; /* Q14 */ } } else { /* Only short-lag bias */ - for( i = max_lag_4kHz; i >= min_lag_4kHz; i-- ) { - sum = (opus_int32)C[ 0 ][ i ]; - sum = silk_SMLAWB( sum, sum, silk_LSHIFT( -i, 4 ) ); /* Q-1 */ - C[ 0 ][ i ] = (opus_int16)sum; /* Q-1 */ + for( i = MAX_LAG_4KHZ; i >= MIN_LAG_4KHZ; i-- ) { + sum = silk_LSHIFT( (opus_int32)C[ i - MIN_LAG_4KHZ ], 1 ); /* Q14 */ + sum = silk_SMLAWB( sum, sum, silk_LSHIFT( -i, 4 ) ); /* Q14 */ + C[ i - MIN_LAG_4KHZ ] = (opus_int16)sum; /* Q14 */ } } /* Sort */ length_d_srch = silk_ADD_LSHIFT32( 4, complexity, 1 ); silk_assert( 3 * length_d_srch <= PE_D_SRCH_LENGTH ); - silk_insertion_sort_decreasing_int16( &C[ 0 ][ min_lag_4kHz ], d_srch, max_lag_4kHz - min_lag_4kHz + 1, length_d_srch ); + silk_insertion_sort_decreasing_int16( C, d_srch, CSTRIDE_4KHZ, + length_d_srch ); /* Escape if correlation is very low already here */ - target_ptr = &frame_4kHz[ silk_SMULBB( sf_length_4kHz, nb_subfr ) ]; - energy = silk_inner_prod_aligned( target_ptr, target_ptr, silk_LSHIFT( sf_length_4kHz, 2 ) ); - energy = silk_ADD_SAT32( energy, 1000 ); /* Q0 */ - Cmax = (opus_int)C[ 0 ][ min_lag_4kHz ]; /* Q-1 */ - threshold = silk_SMULBB( Cmax, Cmax ); /* Q-2 */ - - /* Compare in Q-2 domain */ - if( silk_RSHIFT( energy, 4 + 2 ) > threshold ) { + Cmax = (opus_int)C[ 0 ]; /* Q14 */ + if( Cmax < SILK_FIX_CONST( 0.2, 14 ) ) { silk_memset( pitch_out, 0, nb_subfr * sizeof( opus_int ) ); *LTPCorr_Q15 = 0; *lagIndex = 0; *contourIndex = 0; + RESTORE_STACK; return 1; } threshold = silk_SMULWB( search_thres1_Q16, Cmax ); for( i = 0; i < length_d_srch; i++ ) { /* Convert to 8 kHz indices for the sorted correlation that exceeds the threshold */ - if( C[ 0 ][ min_lag_4kHz + i ] > threshold ) { - d_srch[ i ] = silk_LSHIFT( d_srch[ i ] + min_lag_4kHz, 1 ); + if( C[ i ] > threshold ) { + d_srch[ i ] = silk_LSHIFT( d_srch[ i ] + MIN_LAG_4KHZ, 1 ); } else { length_d_srch = i; break; @@ -265,34 +270,37 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 } silk_assert( length_d_srch > 0 ); - for( i = min_lag_8kHz - 5; i < max_lag_8kHz + 5; i++ ) { - d_comp[ i ] = 0; + ALLOC( d_comp, D_COMP_STRIDE, opus_int16 ); + for( i = D_COMP_MIN; i < D_COMP_MAX; i++ ) { + d_comp[ i - D_COMP_MIN ] = 0; } for( i = 0; i < length_d_srch; i++ ) { - d_comp[ d_srch[ i ] ] = 1; + d_comp[ d_srch[ i ] - D_COMP_MIN ] = 1; } /* Convolution */ - for( i = max_lag_8kHz + 3; i >= min_lag_8kHz; i-- ) { - d_comp[ i ] += d_comp[ i - 1 ] + d_comp[ i - 2 ]; + for( i = D_COMP_MAX - 1; i >= MIN_LAG_8KHZ; i-- ) { + d_comp[ i - D_COMP_MIN ] += + d_comp[ i - 1 - D_COMP_MIN ] + d_comp[ i - 2 - D_COMP_MIN ]; } length_d_srch = 0; - for( i = min_lag_8kHz; i < max_lag_8kHz + 1; i++ ) { - if( d_comp[ i + 1 ] > 0 ) { + for( i = MIN_LAG_8KHZ; i < MAX_LAG_8KHZ + 1; i++ ) { + if( d_comp[ i + 1 - D_COMP_MIN ] > 0 ) { d_srch[ length_d_srch ] = i; length_d_srch++; } } /* Convolution */ - for( i = max_lag_8kHz + 3; i >= min_lag_8kHz; i-- ) { - d_comp[ i ] += d_comp[ i - 1 ] + d_comp[ i - 2 ] + d_comp[ i - 3 ]; + for( i = D_COMP_MAX - 1; i >= MIN_LAG_8KHZ; i-- ) { + d_comp[ i - D_COMP_MIN ] += d_comp[ i - 1 - D_COMP_MIN ] + + d_comp[ i - 2 - D_COMP_MIN ] + d_comp[ i - 3 - D_COMP_MIN ]; } length_d_comp = 0; - for( i = min_lag_8kHz; i < max_lag_8kHz + 4; i++ ) { - if( d_comp[ i ] > 0 ) { + for( i = MIN_LAG_8KHZ; i < D_COMP_MAX; i++ ) { + if( d_comp[ i - D_COMP_MIN ] > 0 ) { d_comp[ length_d_comp ] = i - 2; length_d_comp++; } @@ -306,8 +314,9 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 ** Scale signal down to avoid correlations measures from overflowing *******************************************************************************/ /* find scaling as max scaling for each subframe */ - shift = silk_P_Ana_find_scaling( frame_8kHz, frame_length_8kHz, sf_length_8kHz ); + silk_sum_sqr_shift( &energy, &shift, frame_8kHz, frame_length_8kHz ); if( shift > 0 ) { + shift = silk_RSHIFT( shift, 1 ); for( i = 0; i < frame_length_8kHz; i++ ) { frame_8kHz[ i ] = silk_RSHIFT( frame_8kHz[ i ], shift ); } @@ -316,43 +325,37 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 /********************************************************************************* * Find energy of each subframe projected onto its history, for a range of delays *********************************************************************************/ - silk_memset( C, 0, PE_MAX_NB_SUBFR * ( ( PE_MAX_LAG >> 1 ) + 5 ) * sizeof( opus_int16 ) ); + silk_memset( C, 0, nb_subfr * CSTRIDE_8KHZ * sizeof( opus_int16 ) ); target_ptr = &frame_8kHz[ PE_LTP_MEM_LENGTH_MS * 8 ]; for( k = 0; k < nb_subfr; k++ ) { /* Check that we are within range of the array */ silk_assert( target_ptr >= frame_8kHz ); - silk_assert( target_ptr + sf_length_8kHz <= frame_8kHz + frame_length_8kHz ); + silk_assert( target_ptr + SF_LENGTH_8KHZ <= frame_8kHz + frame_length_8kHz ); - energy_target = silk_inner_prod_aligned( target_ptr, target_ptr, sf_length_8kHz ); + energy_target = silk_ADD32( silk_inner_prod_aligned( target_ptr, target_ptr, SF_LENGTH_8KHZ ), 1 ); for( j = 0; j < length_d_comp; j++ ) { d = d_comp[ j ]; basis_ptr = target_ptr - d; /* Check that we are within range of the array */ silk_assert( basis_ptr >= frame_8kHz ); - silk_assert( basis_ptr + sf_length_8kHz <= frame_8kHz + frame_length_8kHz ); + silk_assert( basis_ptr + SF_LENGTH_8KHZ <= frame_8kHz + frame_length_8kHz ); - cross_corr = silk_inner_prod_aligned( target_ptr, basis_ptr, sf_length_8kHz ); - energy_basis = silk_inner_prod_aligned( basis_ptr, basis_ptr, sf_length_8kHz ); + cross_corr = silk_inner_prod_aligned( target_ptr, basis_ptr, SF_LENGTH_8KHZ ); if( cross_corr > 0 ) { - energy = silk_max( energy_target, energy_basis ); /* Find max to make sure first division < 1.0 */ - lz = silk_CLZ32( cross_corr ); - lshift = silk_LIMIT_32( lz - 1, 0, 15 ); - temp32 = silk_DIV32( silk_LSHIFT( cross_corr, lshift ), silk_RSHIFT( energy, 15 - lshift ) + 1 ); /* Q15 */ - silk_assert( temp32 == silk_SAT16( temp32 ) ); - temp32 = silk_SMULWB( cross_corr, temp32 ); /* Q(-1), cc * ( cc / max(b, t) ) */ - temp32 = silk_ADD_SAT32( temp32, temp32 ); /* Q(0) */ - lz = silk_CLZ32( temp32 ); - lshift = silk_LIMIT_32( lz - 1, 0, 15 ); - energy = silk_min( energy_target, energy_basis ); - C[ k ][ d ] = silk_DIV32( silk_LSHIFT( temp32, lshift ), silk_RSHIFT( energy, 15 - lshift ) + 1 ); /* Q15*/ + energy_basis = silk_inner_prod_aligned( basis_ptr, basis_ptr, SF_LENGTH_8KHZ ); + matrix_ptr( C, k, d - ( MIN_LAG_8KHZ - 2 ), CSTRIDE_8KHZ ) = + (opus_int16)silk_DIV32_varQ( cross_corr, + silk_ADD32( energy_target, + energy_basis ), + 13 + 1 ); /* Q13 */ } else { - C[ k ][ d ] = 0; + matrix_ptr( C, k, d - ( MIN_LAG_8KHZ - 2 ), CSTRIDE_8KHZ ) = 0; } } - target_ptr += sf_length_8kHz; + target_ptr += SF_LENGTH_8KHZ; } /* search over lag range and lags codebook */ @@ -374,7 +377,7 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 } else { prevLag_log2_Q7 = 0; } - silk_assert( search_thres2_Q15 == silk_SAT16( search_thres2_Q15 ) ); + silk_assert( search_thres2_Q13 == silk_SAT16( search_thres2_Q13 ) ); /* Set up stage 2 codebook based on number of subframes */ if( nb_subfr == PE_MAX_NB_SUBFR ) { cbk_size = PE_NB_CBKS_STAGE2_EXT; @@ -385,12 +388,10 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 } else { nb_cbk_search = PE_NB_CBKS_STAGE2; } - corr_thres_Q15 = silk_RSHIFT( silk_SMULBB( search_thres2_Q15, search_thres2_Q15 ), 13 ); } else { cbk_size = PE_NB_CBKS_STAGE2_10MS; Lag_CB_ptr = &silk_CB_lags_stage2_10_ms[ 0 ][ 0 ]; nb_cbk_search = PE_NB_CBKS_STAGE2_10MS; - corr_thres_Q15 = silk_RSHIFT( silk_SMULBB( search_thres2_Q15, search_thres2_Q15 ), 14 ); } for( k = 0; k < length_d_srch; k++ ) { @@ -398,8 +399,13 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 for( j = 0; j < nb_cbk_search; j++ ) { CC[ j ] = 0; for( i = 0; i < nb_subfr; i++ ) { + opus_int d_subfr; /* Try all codebooks */ - CC[ j ] = CC[ j ] + (opus_int32)C[ i ][ d + matrix_ptr( Lag_CB_ptr, i, j, cbk_size )]; + d_subfr = d + matrix_ptr( Lag_CB_ptr, i, j, cbk_size ); + CC[ j ] = CC[ j ] + + (opus_int32)matrix_ptr( C, i, + d_subfr - ( MIN_LAG_8KHZ - 2 ), + CSTRIDE_8KHZ ); } } /* Find best codebook */ @@ -413,25 +419,25 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 } /* Bias towards shorter lags */ - lag_log2_Q7 = silk_lin2log( (opus_int32)d ); /* Q7 */ + lag_log2_Q7 = silk_lin2log( d ); /* Q7 */ silk_assert( lag_log2_Q7 == silk_SAT16( lag_log2_Q7 ) ); - silk_assert( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 15 ) == silk_SAT16( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 15 ) ) ); - CCmax_new_b = CCmax_new - silk_RSHIFT( silk_SMULBB( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 15 ), lag_log2_Q7 ), 7 ); /* Q15 */ + silk_assert( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 13 ) == silk_SAT16( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 13 ) ) ); + CCmax_new_b = CCmax_new - silk_RSHIFT( silk_SMULBB( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 13 ), lag_log2_Q7 ), 7 ); /* Q13 */ /* Bias towards previous lag */ - silk_assert( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 15 ) == silk_SAT16( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 15 ) ) ); + silk_assert( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 13 ) == silk_SAT16( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 13 ) ) ); if( prevLag > 0 ) { delta_lag_log2_sqr_Q7 = lag_log2_Q7 - prevLag_log2_Q7; silk_assert( delta_lag_log2_sqr_Q7 == silk_SAT16( delta_lag_log2_sqr_Q7 ) ); delta_lag_log2_sqr_Q7 = silk_RSHIFT( silk_SMULBB( delta_lag_log2_sqr_Q7, delta_lag_log2_sqr_Q7 ), 7 ); - prev_lag_bias_Q15 = silk_RSHIFT( silk_SMULBB( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 15 ), *LTPCorr_Q15 ), 15 ); /* Q15 */ - prev_lag_bias_Q15 = silk_DIV32( silk_MUL( prev_lag_bias_Q15, delta_lag_log2_sqr_Q7 ), delta_lag_log2_sqr_Q7 + ( 1 << 6 ) ); - CCmax_new_b -= prev_lag_bias_Q15; /* Q15 */ + prev_lag_bias_Q13 = silk_RSHIFT( silk_SMULBB( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 13 ), *LTPCorr_Q15 ), 15 ); /* Q13 */ + prev_lag_bias_Q13 = silk_DIV32( silk_MUL( prev_lag_bias_Q13, delta_lag_log2_sqr_Q7 ), delta_lag_log2_sqr_Q7 + SILK_FIX_CONST( 0.5, 7 ) ); + CCmax_new_b -= prev_lag_bias_Q13; /* Q13 */ } if( CCmax_new_b > CCmax_b && /* Find maximum biased correlation */ - CCmax_new > corr_thres_Q15 && /* Correlation needs to be high enough to be voiced */ - silk_CB_lags_stage2[ 0 ][ CBimax_new ] <= min_lag_8kHz /* Lag must be in range */ + CCmax_new > silk_SMULBB( nb_subfr, search_thres2_Q13 ) && /* Correlation needs to be high enough to be voiced */ + silk_CB_lags_stage2[ 0 ][ CBimax_new ] <= MIN_LAG_8KHZ /* Lag must be in range */ ) { CCmax_b = CCmax_new_b; CCmax = CCmax_new; @@ -446,24 +452,31 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 *LTPCorr_Q15 = 0; *lagIndex = 0; *contourIndex = 0; + RESTORE_STACK; return 1; } + /* Output normalized correlation */ + *LTPCorr_Q15 = (opus_int)silk_LSHIFT( silk_DIV32_16( CCmax, nb_subfr ), 2 ); + silk_assert( *LTPCorr_Q15 >= 0 ); + if( Fs_kHz > 8 ) { + VARDECL( opus_int16, scratch_mem ); /***************************************************************************/ /* Scale input signal down to avoid correlations measures from overflowing */ /***************************************************************************/ /* find scaling as max scaling for each subframe */ - shift = silk_P_Ana_find_scaling( frame, frame_length, sf_length ); + silk_sum_sqr_shift( &energy, &shift, frame, frame_length ); + ALLOC( scratch_mem, shift > 0 ? frame_length : ALLOC_NONE, opus_int16 ); if( shift > 0 ) { /* Move signal to scratch mem because the input signal should be unchanged */ - /* Reuse the 32 bit scratch mem vector, use a 16 bit pointer from now */ - input_frame_ptr = (opus_int16*)scratch_mem; + shift = silk_RSHIFT( shift, 1 ); for( i = 0; i < frame_length; i++ ) { - input_frame_ptr[ i ] = silk_RSHIFT( frame[ i ], shift ); + scratch_mem[ i ] = silk_RSHIFT( frame[ i ], shift ); } + input_frame_ptr = scratch_mem; } else { - input_frame_ptr = (opus_int16*)frame; + input_frame_ptr = frame; } /* Search in original signal */ @@ -483,22 +496,13 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 start_lag = silk_max_int( lag - 2, min_lag ); end_lag = silk_min_int( lag + 2, max_lag ); lag_new = lag; /* to avoid undefined lag */ - CBimax = 0; /* to avoid undefined lag */ - silk_assert( silk_LSHIFT( CCmax, 13 ) >= 0 ); - *LTPCorr_Q15 = (opus_int)silk_SQRT_APPROX( silk_LSHIFT( CCmax, 13 ) ); /* Output normalized correlation */ + CBimax = 0; /* to avoid undefined lag */ CCmax = silk_int32_MIN; /* pitch lags according to second stage */ for( k = 0; k < nb_subfr; k++ ) { pitch_out[ k ] = lag + 2 * silk_CB_lags_stage2[ k ][ CBimax_old ]; } - /* Calculate the correlations and energies needed in stage 3 */ - silk_P_Ana_calc_corr_st3( crosscorr_st3, input_frame_ptr, start_lag, sf_length, nb_subfr, complexity ); - silk_P_Ana_calc_energy_st3( energies_st3, input_frame_ptr, start_lag, sf_length, nb_subfr, complexity ); - - lag_counter = 0; - silk_assert( lag == silk_SAT16( lag ) ); - contour_bias_Q20 = silk_DIV32_16( SILK_FIX_CONST( PE_FLATCONTOUR_BIAS, 20 ), lag ); /* Set up codebook parameters according to complexity setting and frame length */ if( nb_subfr == PE_MAX_NB_SUBFR ) { @@ -510,41 +514,43 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 cbk_size = PE_NB_CBKS_STAGE3_10MS; Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; } + + /* Calculate the correlations and energies needed in stage 3 */ + ALLOC( energies_st3, nb_subfr * nb_cbk_search, silk_pe_stage3_vals ); + ALLOC( cross_corr_st3, nb_subfr * nb_cbk_search, silk_pe_stage3_vals ); + silk_P_Ana_calc_corr_st3( cross_corr_st3, input_frame_ptr, start_lag, sf_length, nb_subfr, complexity, arch ); + silk_P_Ana_calc_energy_st3( energies_st3, input_frame_ptr, start_lag, sf_length, nb_subfr, complexity ); + + lag_counter = 0; + silk_assert( lag == silk_SAT16( lag ) ); + contour_bias_Q15 = silk_DIV32_16( SILK_FIX_CONST( PE_FLATCONTOUR_BIAS, 15 ), lag ); + + target_ptr = &input_frame_ptr[ PE_LTP_MEM_LENGTH_MS * Fs_kHz ]; + energy_target = silk_ADD32( silk_inner_prod_aligned( target_ptr, target_ptr, nb_subfr * sf_length ), 1 ); for( d = start_lag; d <= end_lag; d++ ) { for( j = 0; j < nb_cbk_search; j++ ) { cross_corr = 0; - energy = 0; + energy = energy_target; for( k = 0; k < nb_subfr; k++ ) { - silk_assert( PE_MAX_NB_SUBFR == 4 ); - energy += silk_RSHIFT( energies_st3[ k ][ j ][ lag_counter ], 2 ); /* use mean, to avoid overflow */ + cross_corr = silk_ADD32( cross_corr, + matrix_ptr( cross_corr_st3, k, j, + nb_cbk_search )[ lag_counter ] ); + energy = silk_ADD32( energy, + matrix_ptr( energies_st3, k, j, + nb_cbk_search )[ lag_counter ] ); silk_assert( energy >= 0 ); - cross_corr += silk_RSHIFT( crosscorr_st3[ k ][ j ][ lag_counter ], 2 ); /* use mean, to avoid overflow */ } if( cross_corr > 0 ) { - /* Divide cross_corr / energy and get result in Q15 */ - lz = silk_CLZ32( cross_corr ); - /* Divide with result in Q13, cross_corr could be larger than energy */ - lshift = silk_LIMIT_32( lz - 1, 0, 13 ); - CCmax_new = silk_DIV32( silk_LSHIFT( cross_corr, lshift ), silk_RSHIFT( energy, 13 - lshift ) + 1 ); - CCmax_new = silk_SAT16( CCmax_new ); - CCmax_new = silk_SMULWB( cross_corr, CCmax_new ); - /* Saturate */ - if( CCmax_new > silk_RSHIFT( silk_int32_MAX, 3 ) ) { - CCmax_new = silk_int32_MAX; - } else { - CCmax_new = silk_LSHIFT( CCmax_new, 3 ); - } + CCmax_new = silk_DIV32_varQ( cross_corr, energy, 13 + 1 ); /* Q13 */ /* Reduce depending on flatness of contour */ - diff = silk_int16_MAX - silk_RSHIFT( silk_MUL( contour_bias_Q20, j ), 5 ); /* Q20 -> Q15 */ + diff = silk_int16_MAX - silk_MUL( contour_bias_Q15, j ); /* Q15 */ silk_assert( diff == silk_SAT16( diff ) ); - CCmax_new = silk_LSHIFT( silk_SMULWB( CCmax_new, diff ), 1 ); + CCmax_new = silk_SMULWB( CCmax_new, diff ); /* Q14 */ } else { CCmax_new = 0; } - if( CCmax_new > CCmax && - ( d + silk_CB_lags_stage3[ 0 ][ j ] ) <= max_lag - ) { + if( CCmax_new > CCmax && ( d + silk_CB_lags_stage3[ 0 ][ j ] ) <= max_lag ) { CCmax = CCmax_new; lag_new = d; CBimax = j; @@ -560,40 +566,50 @@ opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 *lagIndex = (opus_int16)( lag_new - min_lag); *contourIndex = (opus_int8)CBimax; } else { /* Fs_kHz == 8 */ - /* Save Lags and correlation */ - CCmax = silk_max( CCmax, 0 ); - *LTPCorr_Q15 = (opus_int)silk_SQRT_APPROX( silk_LSHIFT( CCmax, 13 ) ); /* Output normalized correlation */ + /* Save Lags */ for( k = 0; k < nb_subfr; k++ ) { pitch_out[ k ] = lag + matrix_ptr( Lag_CB_ptr, k, CBimax, cbk_size ); - pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], min_lag_8kHz, PE_MAX_LAG_MS * Fs_kHz ); + pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], MIN_LAG_8KHZ, PE_MAX_LAG_MS * 8 ); } - *lagIndex = (opus_int16)( lag - min_lag_8kHz ); + *lagIndex = (opus_int16)( lag - MIN_LAG_8KHZ ); *contourIndex = (opus_int8)CBimax; } silk_assert( *lagIndex >= 0 ); /* return as voiced */ + RESTORE_STACK; return 0; } -/*************************************************************************/ -/* Calculates the correlations used in stage 3 search. In order to cover */ -/* the whole lag codebook for all the searched offset lags (lag +- 2), */ -/*************************************************************************/ -void silk_P_Ana_calc_corr_st3( - opus_int32 cross_corr_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ],/* (O) 3 DIM correlation array */ +/*********************************************************************** + * Calculates the correlations used in stage 3 search. In order to cover + * the whole lag codebook for all the searched offset lags (lag +- 2), + * the following correlations are needed in each sub frame: + * + * sf1: lag range [-8,...,7] total 16 correlations + * sf2: lag range [-4,...,4] total 9 correlations + * sf3: lag range [-3,....4] total 8 correltions + * sf4: lag range [-6,....8] total 15 correlations + * + * In total 48 correlations. The direct implementation computed in worst + * case 4*12*5 = 240 correlations, but more likely around 120. + ***********************************************************************/ +static void silk_P_Ana_calc_corr_st3( + silk_pe_stage3_vals cross_corr_st3[], /* O 3 DIM correlation array */ const opus_int16 frame[], /* I vector to correlate */ opus_int start_lag, /* I lag offset to search around */ opus_int sf_length, /* I length of a 5 ms subframe */ opus_int nb_subfr, /* I number of subframes */ - opus_int complexity /* I Complexity setting */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ ) { - const opus_int16 *target_ptr, *basis_ptr; - opus_int32 cross_corr; + const opus_int16 *target_ptr; opus_int i, j, k, lag_counter, lag_low, lag_high; opus_int nb_cbk_search, delta, idx, cbk_size; - opus_int32 scratch_mem[ SCRATCH_SIZE ]; + VARDECL( opus_int32, scratch_mem ); + VARDECL( opus_int32, xcorr32 ); const opus_int8 *Lag_range_ptr, *Lag_CB_ptr; + SAVE_STACK; silk_assert( complexity >= SILK_PE_MIN_COMPLEX ); silk_assert( complexity <= SILK_PE_MAX_COMPLEX ); @@ -610,6 +626,8 @@ void silk_P_Ana_calc_corr_st3( nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; cbk_size = PE_NB_CBKS_STAGE3_10MS; } + ALLOC( scratch_mem, SCRATCH_SIZE, opus_int32 ); + ALLOC( xcorr32, SCRATCH_SIZE, opus_int32 ); target_ptr = &frame[ silk_LSHIFT( sf_length, 2 ) ]; /* Pointer to middle of frame */ for( k = 0; k < nb_subfr; k++ ) { @@ -618,11 +636,11 @@ void silk_P_Ana_calc_corr_st3( /* Calculate the correlations for each subframe */ lag_low = matrix_ptr( Lag_range_ptr, k, 0, 2 ); lag_high = matrix_ptr( Lag_range_ptr, k, 1, 2 ); + silk_assert(lag_high-lag_low+1 <= SCRATCH_SIZE); + celt_pitch_xcorr( target_ptr, target_ptr - start_lag - lag_high, xcorr32, sf_length, lag_high - lag_low + 1, arch ); for( j = lag_low; j <= lag_high; j++ ) { - basis_ptr = target_ptr - ( start_lag + j ); - cross_corr = silk_inner_prod_aligned( (opus_int16*)target_ptr, (opus_int16*)basis_ptr, sf_length ); silk_assert( lag_counter < SCRATCH_SIZE ); - scratch_mem[ lag_counter ] = cross_corr; + scratch_mem[ lag_counter ] = xcorr32[ lag_high - j ]; lag_counter++; } @@ -634,32 +652,35 @@ void silk_P_Ana_calc_corr_st3( for( j = 0; j < PE_NB_STAGE3_LAGS; j++ ) { silk_assert( idx + j < SCRATCH_SIZE ); silk_assert( idx + j < lag_counter ); - cross_corr_st3[ k ][ i ][ j ] = scratch_mem[ idx + j ]; + matrix_ptr( cross_corr_st3, k, i, nb_cbk_search )[ j ] = + scratch_mem[ idx + j ]; } } target_ptr += sf_length; } + RESTORE_STACK; } /********************************************************************/ /* Calculate the energies for first two subframes. The energies are */ /* calculated recursively. */ /********************************************************************/ -void silk_P_Ana_calc_energy_st3( - opus_int32 energies_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ],/* (O) 3 DIM energy array */ - const opus_int16 frame[], /* I vector to calc energy in */ - opus_int start_lag, /* I lag offset to search around */ - opus_int sf_length, /* I length of one 5 ms subframe */ - opus_int nb_subfr, /* I number of subframes */ - opus_int complexity /* I Complexity setting */ +static void silk_P_Ana_calc_energy_st3( + silk_pe_stage3_vals energies_st3[], /* O 3 DIM energy array */ + const opus_int16 frame[], /* I vector to calc energy in */ + opus_int start_lag, /* I lag offset to search around */ + opus_int sf_length, /* I length of one 5 ms subframe */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity /* I Complexity setting */ ) { const opus_int16 *target_ptr, *basis_ptr; opus_int32 energy; opus_int k, i, j, lag_counter; opus_int nb_cbk_search, delta, idx, cbk_size, lag_diff; - opus_int32 scratch_mem[ SCRATCH_SIZE ]; + VARDECL( opus_int32, scratch_mem ); const opus_int8 *Lag_range_ptr, *Lag_CB_ptr; + SAVE_STACK; silk_assert( complexity >= SILK_PE_MIN_COMPLEX ); silk_assert( complexity <= SILK_PE_MAX_COMPLEX ); @@ -676,6 +697,8 @@ void silk_P_Ana_calc_energy_st3( nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; cbk_size = PE_NB_CBKS_STAGE3_10MS; } + ALLOC( scratch_mem, SCRATCH_SIZE, opus_int32 ); + target_ptr = &frame[ silk_LSHIFT( sf_length, 2 ) ]; for( k = 0; k < nb_subfr; k++ ) { lag_counter = 0; @@ -709,37 +732,13 @@ void silk_P_Ana_calc_energy_st3( for( j = 0; j < PE_NB_STAGE3_LAGS; j++ ) { silk_assert( idx + j < SCRATCH_SIZE ); silk_assert( idx + j < lag_counter ); - energies_st3[ k ][ i ][ j ] = scratch_mem[ idx + j ]; - silk_assert( energies_st3[ k ][ i ][ j ] >= 0 ); + matrix_ptr( energies_st3, k, i, nb_cbk_search )[ j ] = + scratch_mem[ idx + j ]; + silk_assert( + matrix_ptr( energies_st3, k, i, nb_cbk_search )[ j ] >= 0 ); } } target_ptr += sf_length; } -} - -opus_int32 silk_P_Ana_find_scaling( - const opus_int16 *frame, - const opus_int frame_length, - const opus_int sum_sqr_len -) -{ - opus_int32 nbits, x_max; - - x_max = silk_int16_array_maxabs( frame, frame_length ); - - if( x_max < silk_int16_MAX ) { - /* Number of bits needed for the sum of the squares */ - nbits = 32 - silk_CLZ32( silk_SMULBB( x_max, x_max ) ); - } else { - /* Here we don't know if x_max should have been silk_int16_MAX + 1, so we expect the worst case */ - nbits = 30; - } - nbits += 17 - silk_CLZ16( sum_sqr_len ); - - /* Without a guarantee of saturation, we need to keep the 31st bit free */ - if( nbits < 31 ) { - return 0; - } else { - return( nbits - 30 ); - } + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/fixed/prefilter_FIX.c b/code/opus-1.1/silk/fixed/prefilter_FIX.c similarity index 96% rename from code/opus-1.0.2/silk/fixed/prefilter_FIX.c rename to code/opus-1.1/silk/fixed/prefilter_FIX.c index a96f5118..d381730c 100644 --- a/code/opus-1.0.2/silk/fixed/prefilter_FIX.c +++ b/code/opus-1.1/silk/fixed/prefilter_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,10 +30,11 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" #include "tuning_parameters.h" /* Prefilter for finding Quantizer input signal */ -static inline void silk_prefilt_FIX( +static OPUS_INLINE void silk_prefilt_FIX( silk_prefilter_state_FIX *P, /* I/O state */ opus_int32 st_res_Q12[], /* I short term residual signal */ opus_int32 xw_Q3[], /* O prefiltered signal */ @@ -101,14 +102,17 @@ void silk_prefilter_FIX( opus_int32 *pxw_Q3; opus_int HarmShapeGain_Q12, Tilt_Q14; opus_int32 HarmShapeFIRPacked_Q12, LF_shp_Q14; - opus_int32 x_filt_Q12[ MAX_SUB_FRAME_LENGTH ]; - opus_int32 st_res_Q2[ MAX_SUB_FRAME_LENGTH + MAX_LPC_ORDER ]; + VARDECL( opus_int32, x_filt_Q12 ); + VARDECL( opus_int32, st_res_Q2 ); opus_int16 B_Q10[ 2 ]; + SAVE_STACK; /* Set up pointers */ px = x; pxw_Q3 = xw_Q3; lag = P->lagPrev; + ALLOC( x_filt_Q12, psEnc->sCmn.subfr_length, opus_int32 ); + ALLOC( st_res_Q2, psEnc->sCmn.subfr_length, opus_int32 ); for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { /* Update Variables that change per sub frame */ if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { @@ -148,10 +152,11 @@ void silk_prefilter_FIX( } P->lagPrev = psEncCtrl->pitchL[ psEnc->sCmn.nb_subfr - 1 ]; + RESTORE_STACK; } /* Prefilter for finding Quantizer input signal */ -static inline void silk_prefilt_FIX( +static OPUS_INLINE void silk_prefilt_FIX( silk_prefilter_state_FIX *P, /* I/O state */ opus_int32 st_res_Q12[], /* I short term residual signal */ opus_int32 xw_Q3[], /* O prefiltered signal */ diff --git a/code/opus-1.0.2/silk/fixed/process_gains_FIX.c b/code/opus-1.1/silk/fixed/process_gains_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/process_gains_FIX.c rename to code/opus-1.1/silk/fixed/process_gains_FIX.c index 22d3a71a..05aba317 100644 --- a/code/opus-1.0.2/silk/fixed/process_gains_FIX.c +++ b/code/opus-1.1/silk/fixed/process_gains_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/regularize_correlations_FIX.c b/code/opus-1.1/silk/fixed/regularize_correlations_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/regularize_correlations_FIX.c rename to code/opus-1.1/silk/fixed/regularize_correlations_FIX.c index 098c1509..a2836b05 100644 --- a/code/opus-1.0.2/silk/fixed/regularize_correlations_FIX.c +++ b/code/opus-1.1/silk/fixed/regularize_correlations_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/residual_energy16_FIX.c b/code/opus-1.1/silk/fixed/residual_energy16_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/residual_energy16_FIX.c rename to code/opus-1.1/silk/fixed/residual_energy16_FIX.c index d61e8493..ebffb2a6 100644 --- a/code/opus-1.0.2/silk/fixed/residual_energy16_FIX.c +++ b/code/opus-1.1/silk/fixed/residual_energy16_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/residual_energy_FIX.c b/code/opus-1.1/silk/fixed/residual_energy_FIX.c similarity index 94% rename from code/opus-1.0.2/silk/fixed/residual_energy_FIX.c rename to code/opus-1.1/silk/fixed/residual_energy_FIX.c index f284e51f..105ae318 100644 --- a/code/opus-1.0.2/silk/fixed/residual_energy_FIX.c +++ b/code/opus-1.1/silk/fixed/residual_energy_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" /* Calculates residual energies of input subframes where all subframes have LPC_order */ /* of preceding samples */ @@ -45,14 +46,18 @@ void silk_residual_energy_FIX( ) { opus_int offset, i, j, rshift, lz1, lz2; - opus_int16 *LPC_res_ptr, LPC_res[ ( MAX_FRAME_LENGTH + MAX_NB_SUBFR * MAX_LPC_ORDER ) / 2 ]; + opus_int16 *LPC_res_ptr; + VARDECL( opus_int16, LPC_res ); const opus_int16 *x_ptr; opus_int32 tmp32; + SAVE_STACK; x_ptr = x; offset = LPC_order + subfr_length; /* Filter input to create the LPC residual for each frame half, and measure subframe energies */ + ALLOC( LPC_res, ( MAX_NB_SUBFR >> 1 ) * offset, opus_int16 ); + silk_assert( ( nb_subfr >> 1 ) * ( MAX_NB_SUBFR >> 1 ) == nb_subfr ); for( i = 0; i < nb_subfr >> 1; i++ ) { /* Calculate half frame LPC residual signal including preceding samples */ silk_LPC_analysis_filter( LPC_res, x_ptr, a_Q12[ i ], ( MAX_NB_SUBFR >> 1 ) * offset, LPC_order ); @@ -88,4 +93,5 @@ void silk_residual_energy_FIX( nrgs[ i ] = silk_SMMUL( tmp32, silk_LSHIFT32( nrgs[ i ], lz1 ) ); /* Q( nrgsQ[ i ] + lz1 + 2 * lz2 - 32 - 32 )*/ nrgsQ[ i ] += lz1 + 2 * lz2 - 32 - 32; } + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/fixed/schur64_FIX.c b/code/opus-1.1/silk/fixed/schur64_FIX.c similarity index 87% rename from code/opus-1.0.2/silk/fixed/schur64_FIX.c rename to code/opus-1.1/silk/fixed/schur64_FIX.c index 5ff27567..764a10ef 100644 --- a/code/opus-1.0.2/silk/fixed/schur64_FIX.c +++ b/code/opus-1.1/silk/fixed/schur64_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -56,6 +56,17 @@ opus_int32 silk_schur64( /* O returns residual ene } for( k = 0; k < order; k++ ) { + /* Check that we won't be getting an unstable rc, otherwise stop here. */ + if (silk_abs_int32(C[ k + 1 ][ 0 ]) >= C[ 0 ][ 1 ]) { + if ( C[ k + 1 ][ 0 ] > 0 ) { + rc_Q16[ k ] = -SILK_FIX_CONST( .99f, 16 ); + } else { + rc_Q16[ k ] = SILK_FIX_CONST( .99f, 16 ); + } + k++; + break; + } + /* Get reflection coefficient: divide two Q30 values and get result in Q31 */ rc_tmp_Q31 = silk_DIV32_varQ( -C[ k + 1 ][ 0 ], C[ 0 ][ 1 ], 31 ); @@ -73,5 +84,9 @@ opus_int32 silk_schur64( /* O returns residual ene } } - return( C[ 0 ][ 1 ] ); + for(; k < order; k++ ) { + rc_Q16[ k ] = 0; + } + + return silk_max_32( 1, C[ 0 ][ 1 ] ); } diff --git a/code/opus-1.0.2/silk/fixed/schur_FIX.c b/code/opus-1.1/silk/fixed/schur_FIX.c similarity index 89% rename from code/opus-1.0.2/silk/fixed/schur_FIX.c rename to code/opus-1.1/silk/fixed/schur_FIX.c index 43db5018..c4c0ef23 100644 --- a/code/opus-1.0.2/silk/fixed/schur_FIX.c +++ b/code/opus-1.1/silk/fixed/schur_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -68,6 +68,16 @@ opus_int32 silk_schur( /* O Returns residual ene } for( k = 0; k < order; k++ ) { + /* Check that we won't be getting an unstable rc, otherwise stop here. */ + if (silk_abs_int32(C[ k + 1 ][ 0 ]) >= C[ 0 ][ 1 ]) { + if ( C[ k + 1 ][ 0 ] > 0 ) { + rc_Q15[ k ] = -SILK_FIX_CONST( .99f, 15 ); + } else { + rc_Q15[ k ] = SILK_FIX_CONST( .99f, 15 ); + } + k++; + break; + } /* Get reflection coefficient */ rc_tmp_Q15 = -silk_DIV32_16( C[ k + 1 ][ 0 ], silk_max_32( silk_RSHIFT( C[ 0 ][ 1 ], 15 ), 1 ) ); @@ -87,6 +97,10 @@ opus_int32 silk_schur( /* O Returns residual ene } } + for(; k < order; k++ ) { + rc_Q15[ k ] = 0; + } + /* return residual energy */ - return C[ 0 ][ 1 ]; + return silk_max_32( 1, C[ 0 ][ 1 ] ); } diff --git a/code/opus-1.0.2/silk/fixed/solve_LS_FIX.c b/code/opus-1.1/silk/fixed/solve_LS_FIX.c similarity index 95% rename from code/opus-1.0.2/silk/fixed/solve_LS_FIX.c rename to code/opus-1.1/silk/fixed/solve_LS_FIX.c index fb913abe..51d7d49d 100644 --- a/code/opus-1.0.2/silk/fixed/solve_LS_FIX.c +++ b/code/opus-1.1/silk/fixed/solve_LS_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main_FIX.h" +#include "stack_alloc.h" #include "tuning_parameters.h" /*****************************/ @@ -42,7 +43,7 @@ typedef struct { } inv_D_t; /* Factorize square matrix A into LDL form */ -static inline void silk_LDL_factorize_FIX( +static OPUS_INLINE void silk_LDL_factorize_FIX( opus_int32 *A, /* I/O Pointer to Symetric Square Matrix */ opus_int M, /* I Size of Matrix */ opus_int32 *L_Q16, /* I/O Pointer to Square Upper triangular Matrix */ @@ -50,7 +51,7 @@ static inline void silk_LDL_factorize_FIX( ); /* Solve Lx = b, when L is lower triangular and has ones on the diagonal */ -static inline void silk_LS_SolveFirst_FIX( +static OPUS_INLINE void silk_LS_SolveFirst_FIX( const opus_int32 *L_Q16, /* I Pointer to Lower Triangular Matrix */ opus_int M, /* I Dim of Matrix equation */ const opus_int32 *b, /* I b Vector */ @@ -58,14 +59,14 @@ static inline void silk_LS_SolveFirst_FIX( ); /* Solve L^t*x = b, where L is lower triangular with ones on the diagonal */ -static inline void silk_LS_SolveLast_FIX( +static OPUS_INLINE void silk_LS_SolveLast_FIX( const opus_int32 *L_Q16, /* I Pointer to Lower Triangular Matrix */ const opus_int M, /* I Dim of Matrix equation */ const opus_int32 *b, /* I b Vector */ opus_int32 *x_Q16 /* O x Vector */ ); -static inline void silk_LS_divide_Q16_FIX( +static OPUS_INLINE void silk_LS_divide_Q16_FIX( opus_int32 T[], /* I/O Numenator vector */ inv_D_t *inv_D, /* I 1 / D vector */ opus_int M /* I dimension */ @@ -79,11 +80,13 @@ void silk_solve_LDL_FIX( opus_int32 *x_Q16 /* O Pointer to x solution vector */ ) { - opus_int32 L_Q16[ MAX_MATRIX_SIZE * MAX_MATRIX_SIZE ]; + VARDECL( opus_int32, L_Q16 ); opus_int32 Y[ MAX_MATRIX_SIZE ]; inv_D_t inv_D[ MAX_MATRIX_SIZE ]; + SAVE_STACK; silk_assert( M <= MAX_MATRIX_SIZE ); + ALLOC( L_Q16, M * M, opus_int32 ); /*************************************************** Factorize A by LDL such that A = L*D*L', @@ -107,9 +110,10 @@ void silk_solve_LDL_FIX( x = inv(L') * inv(D) * Y *****************************************************/ silk_LS_SolveLast_FIX( L_Q16, M, Y, x_Q16 ); + RESTORE_STACK; } -static inline void silk_LDL_factorize_FIX( +static OPUS_INLINE void silk_LDL_factorize_FIX( opus_int32 *A, /* I/O Pointer to Symetric Square Matrix */ opus_int M, /* I Size of Matrix */ opus_int32 *L_Q16, /* I/O Pointer to Square Upper triangular Matrix */ @@ -181,7 +185,7 @@ static inline void silk_LDL_factorize_FIX( silk_assert( status == 0 ); } -static inline void silk_LS_divide_Q16_FIX( +static OPUS_INLINE void silk_LS_divide_Q16_FIX( opus_int32 T[], /* I/O Numenator vector */ inv_D_t *inv_D, /* I 1 / D vector */ opus_int M /* I dimension */ @@ -201,7 +205,7 @@ static inline void silk_LS_divide_Q16_FIX( } /* Solve Lx = b, when L is lower triangular and has ones on the diagonal */ -static inline void silk_LS_SolveFirst_FIX( +static OPUS_INLINE void silk_LS_SolveFirst_FIX( const opus_int32 *L_Q16, /* I Pointer to Lower Triangular Matrix */ opus_int M, /* I Dim of Matrix equation */ const opus_int32 *b, /* I b Vector */ @@ -223,7 +227,7 @@ static inline void silk_LS_SolveFirst_FIX( } /* Solve L^t*x = b, where L is lower triangular with ones on the diagonal */ -static inline void silk_LS_SolveLast_FIX( +static OPUS_INLINE void silk_LS_SolveLast_FIX( const opus_int32 *L_Q16, /* I Pointer to Lower Triangular Matrix */ const opus_int M, /* I Dim of Matrix equation */ const opus_int32 *b, /* I b Vector */ diff --git a/code/opus-1.0.2/silk/fixed/structs_FIX.h b/code/opus-1.1/silk/fixed/structs_FIX.h similarity index 99% rename from code/opus-1.0.2/silk/fixed/structs_FIX.h rename to code/opus-1.1/silk/fixed/structs_FIX.h index 4162608b..244b4793 100644 --- a/code/opus-1.0.2/silk/fixed/structs_FIX.h +++ b/code/opus-1.1/silk/fixed/structs_FIX.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/fixed/vector_ops_FIX.c b/code/opus-1.1/silk/fixed/vector_ops_FIX.c similarity index 80% rename from code/opus-1.0.2/silk/fixed/vector_ops_FIX.c rename to code/opus-1.1/silk/fixed/vector_ops_FIX.c index d6206024..509c8b35 100644 --- a/code/opus-1.0.2/silk/fixed/vector_ops_FIX.c +++ b/code/opus-1.1/silk/fixed/vector_ops_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -94,34 +94,3 @@ opus_int64 silk_inner_prod16_aligned_64( } return sum; } - -/* Function that returns the maximum absolut value of the input vector */ -opus_int16 silk_int16_array_maxabs( /* O Maximum absolute value, max: 2^15-1 */ - const opus_int16 *vec, /* I Input vector [len] */ - const opus_int32 len /* I Length of input vector */ -) -{ - opus_int32 max = 0, i, lvl = 0, ind; - if( len == 0 ) return 0; - - ind = len - 1; - max = silk_SMULBB( vec[ ind ], vec[ ind ] ); - for( i = len - 2; i >= 0; i-- ) { - lvl = silk_SMULBB( vec[ i ], vec[ i ] ); - if( lvl > max ) { - max = lvl; - ind = i; - } - } - - /* Do not return 32768, as it will not fit in an int16 so may lead to problems later on */ - if( max >= 1073676289 ) { /* (2^15-1)^2 = 1073676289 */ - return( silk_int16_MAX ); - } else { - if( vec[ ind ] < 0 ) { - return( -vec[ ind ] ); - } else { - return( vec[ ind ] ); - } - } -} diff --git a/code/opus-1.0.2/silk/fixed/warped_autocorrelation_FIX.c b/code/opus-1.1/silk/fixed/warped_autocorrelation_FIX.c similarity index 99% rename from code/opus-1.0.2/silk/fixed/warped_autocorrelation_FIX.c rename to code/opus-1.1/silk/fixed/warped_autocorrelation_FIX.c index d7a3944b..a4a579b1 100644 --- a/code/opus-1.0.2/silk/fixed/warped_autocorrelation_FIX.c +++ b/code/opus-1.1/silk/fixed/warped_autocorrelation_FIX.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/LPC_analysis_filter_FLP.c b/code/opus-1.1/silk/float/LPC_analysis_filter_FLP.c similarity index 97% rename from code/opus-1.0.2/silk/float/LPC_analysis_filter_FLP.c rename to code/opus-1.1/silk/float/LPC_analysis_filter_FLP.c index 9845655b..cae89a0a 100644 --- a/code/opus-1.0.2/silk/float/LPC_analysis_filter_FLP.c +++ b/code/opus-1.1/silk/float/LPC_analysis_filter_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -40,7 +40,7 @@ POSSIBILITY OF SUCH DAMAGE. /************************************************/ /* 16th order LPC analysis filter, does not write first 16 samples */ -static inline void silk_LPC_analysis_filter16_FLP( +static OPUS_INLINE void silk_LPC_analysis_filter16_FLP( silk_float r_LPC[], /* O LPC residual signal */ const silk_float PredCoef[], /* I LPC coefficients */ const silk_float s[], /* I Input signal */ @@ -78,7 +78,7 @@ static inline void silk_LPC_analysis_filter16_FLP( } /* 12th order LPC analysis filter, does not write first 12 samples */ -static inline void silk_LPC_analysis_filter12_FLP( +static OPUS_INLINE void silk_LPC_analysis_filter12_FLP( silk_float r_LPC[], /* O LPC residual signal */ const silk_float PredCoef[], /* I LPC coefficients */ const silk_float s[], /* I Input signal */ @@ -112,7 +112,7 @@ static inline void silk_LPC_analysis_filter12_FLP( } /* 10th order LPC analysis filter, does not write first 10 samples */ -static inline void silk_LPC_analysis_filter10_FLP( +static OPUS_INLINE void silk_LPC_analysis_filter10_FLP( silk_float r_LPC[], /* O LPC residual signal */ const silk_float PredCoef[], /* I LPC coefficients */ const silk_float s[], /* I Input signal */ @@ -144,7 +144,7 @@ static inline void silk_LPC_analysis_filter10_FLP( } /* 8th order LPC analysis filter, does not write first 8 samples */ -static inline void silk_LPC_analysis_filter8_FLP( +static OPUS_INLINE void silk_LPC_analysis_filter8_FLP( silk_float r_LPC[], /* O LPC residual signal */ const silk_float PredCoef[], /* I LPC coefficients */ const silk_float s[], /* I Input signal */ @@ -174,7 +174,7 @@ static inline void silk_LPC_analysis_filter8_FLP( } /* 6th order LPC analysis filter, does not write first 6 samples */ -static inline void silk_LPC_analysis_filter6_FLP( +static OPUS_INLINE void silk_LPC_analysis_filter6_FLP( silk_float r_LPC[], /* O LPC residual signal */ const silk_float PredCoef[], /* I LPC coefficients */ const silk_float s[], /* I Input signal */ diff --git a/code/opus-1.0.2/silk/float/LPC_inv_pred_gain_FLP.c b/code/opus-1.1/silk/float/LPC_inv_pred_gain_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/LPC_inv_pred_gain_FLP.c rename to code/opus-1.1/silk/float/LPC_inv_pred_gain_FLP.c index 8645f77f..25178bac 100644 --- a/code/opus-1.0.2/silk/float/LPC_inv_pred_gain_FLP.c +++ b/code/opus-1.1/silk/float/LPC_inv_pred_gain_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/LTP_analysis_filter_FLP.c b/code/opus-1.1/silk/float/LTP_analysis_filter_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/LTP_analysis_filter_FLP.c rename to code/opus-1.1/silk/float/LTP_analysis_filter_FLP.c index d3a6a5ae..849b7c1c 100644 --- a/code/opus-1.0.2/silk/float/LTP_analysis_filter_FLP.c +++ b/code/opus-1.1/silk/float/LTP_analysis_filter_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/LTP_scale_ctrl_FLP.c b/code/opus-1.1/silk/float/LTP_scale_ctrl_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/LTP_scale_ctrl_FLP.c rename to code/opus-1.1/silk/float/LTP_scale_ctrl_FLP.c index f3f0c572..8dbe29d0 100644 --- a/code/opus-1.0.2/silk/float/LTP_scale_ctrl_FLP.c +++ b/code/opus-1.1/silk/float/LTP_scale_ctrl_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/SigProc_FLP.h b/code/opus-1.1/silk/float/SigProc_FLP.h similarity index 95% rename from code/opus-1.0.2/silk/float/SigProc_FLP.h rename to code/opus-1.1/silk/float/SigProc_FLP.h index 036b46da..f0cb3733 100644 --- a/code/opus-1.0.2/silk/float/SigProc_FLP.h +++ b/code/opus-1.1/silk/float/SigProc_FLP.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -94,7 +94,8 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, const silk_float search_thres2, /* I Final threshold for lag candidates 0 - 1 */ const opus_int Fs_kHz, /* I sample frequency (kHz) */ const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ - const opus_int nb_subfr /* I Number of 5 ms subframes */ + const opus_int nb_subfr, /* I Number of 5 ms subframes */ + int arch /* I Run-time architecture */ ); void silk_insertion_sort_decreasing_FLP( @@ -153,19 +154,19 @@ double silk_energy_FLP( #define silk_abs_float( a ) ((silk_float)fabs(a)) /* sigmoid function */ -static inline silk_float silk_sigmoid( silk_float x ) +static OPUS_INLINE silk_float silk_sigmoid( silk_float x ) { return (silk_float)(1.0 / (1.0 + exp(-x))); } /* floating-point to integer conversion (rounding) */ -static inline opus_int32 silk_float2int( silk_float x ) +static OPUS_INLINE opus_int32 silk_float2int( silk_float x ) { return (opus_int32)float2int( x ); } /* floating-point to integer conversion (rounding) */ -static inline void silk_float2short_array( +static OPUS_INLINE void silk_float2short_array( opus_int16 *out, const silk_float *in, opus_int32 length @@ -178,7 +179,7 @@ static inline void silk_float2short_array( } /* integer to floating-point conversion */ -static inline void silk_short2float_array( +static OPUS_INLINE void silk_short2float_array( silk_float *out, const opus_int16 *in, opus_int32 length @@ -191,7 +192,7 @@ static inline void silk_short2float_array( } /* using log2() helps the fixed-point conversion */ -static inline silk_float silk_log2( double x ) +static OPUS_INLINE silk_float silk_log2( double x ) { return ( silk_float )( 3.32192809488736 * log10( x ) ); } diff --git a/code/opus-1.0.2/silk/float/apply_sine_window_FLP.c b/code/opus-1.1/silk/float/apply_sine_window_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/apply_sine_window_FLP.c rename to code/opus-1.1/silk/float/apply_sine_window_FLP.c index e06333f7..6aae57c0 100644 --- a/code/opus-1.0.2/silk/float/apply_sine_window_FLP.c +++ b/code/opus-1.1/silk/float/apply_sine_window_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/autocorrelation_FLP.c b/code/opus-1.1/silk/float/autocorrelation_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/autocorrelation_FLP.c rename to code/opus-1.1/silk/float/autocorrelation_FLP.c index 9ce709e2..8b8a9e65 100644 --- a/code/opus-1.0.2/silk/float/autocorrelation_FLP.c +++ b/code/opus-1.1/silk/float/autocorrelation_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/burg_modified_FLP.c b/code/opus-1.1/silk/float/burg_modified_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/burg_modified_FLP.c rename to code/opus-1.1/silk/float/burg_modified_FLP.c index 31c9b228..ea5dc25a 100644 --- a/code/opus-1.0.2/silk/float/burg_modified_FLP.c +++ b/code/opus-1.1/silk/float/burg_modified_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/bwexpander_FLP.c b/code/opus-1.1/silk/float/bwexpander_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/bwexpander_FLP.c rename to code/opus-1.1/silk/float/bwexpander_FLP.c index 59ca4eaf..d55a4d79 100644 --- a/code/opus-1.0.2/silk/float/bwexpander_FLP.c +++ b/code/opus-1.1/silk/float/bwexpander_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/corrMatrix_FLP.c b/code/opus-1.1/silk/float/corrMatrix_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/corrMatrix_FLP.c rename to code/opus-1.1/silk/float/corrMatrix_FLP.c index c59f73c3..eae6a1cf 100644 --- a/code/opus-1.0.2/silk/float/corrMatrix_FLP.c +++ b/code/opus-1.1/silk/float/corrMatrix_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/encode_frame_FLP.c b/code/opus-1.1/silk/float/encode_frame_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/encode_frame_FLP.c rename to code/opus-1.1/silk/float/encode_frame_FLP.c index 23260bc7..d54e2686 100644 --- a/code/opus-1.0.2/silk/float/encode_frame_FLP.c +++ b/code/opus-1.1/silk/float/encode_frame_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -33,7 +33,7 @@ POSSIBILITY OF SUCH DAMAGE. #include "tuning_parameters.h" /* Low Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode with lower bitrate */ -static inline void silk_LBRR_encode_FLP( +static OPUS_INLINE void silk_LBRR_encode_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ const silk_float xfw[], /* I Input signal */ @@ -129,7 +129,7 @@ opus_int silk_encode_frame_FLP( /*****************************************/ /* Find pitch lags, initial LPC analysis */ /*****************************************/ - silk_find_pitch_lags_FLP( psEnc, &sEncCtrl, res_pitch, x_frame ); + silk_find_pitch_lags_FLP( psEnc, &sEncCtrl, res_pitch, x_frame, psEnc->sCmn.arch ); /************************/ /* Noise shape analysis */ @@ -294,10 +294,6 @@ opus_int silk_encode_frame_FLP( silk_memmove( psEnc->x_buf, &psEnc->x_buf[ psEnc->sCmn.frame_length ], ( psEnc->sCmn.ltp_mem_length + LA_SHAPE_MS * psEnc->sCmn.fs_kHz ) * sizeof( silk_float ) ); - /* Parameters needed for next frame */ - psEnc->sCmn.prevLag = sEncCtrl.pitchL[ psEnc->sCmn.nb_subfr - 1 ]; - psEnc->sCmn.prevSignalType = psEnc->sCmn.indices.signalType; - /* Exit without entropy coding */ if( psEnc->sCmn.prefillFlag ) { /* No payload */ @@ -305,6 +301,10 @@ opus_int silk_encode_frame_FLP( return ret; } + /* Parameters needed for next frame */ + psEnc->sCmn.prevLag = sEncCtrl.pitchL[ psEnc->sCmn.nb_subfr - 1 ]; + psEnc->sCmn.prevSignalType = psEnc->sCmn.indices.signalType; + /****************************************/ /* Finalize payload */ /****************************************/ @@ -316,7 +316,7 @@ opus_int silk_encode_frame_FLP( } /* Low-Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode excitation at lower bitrate */ -static inline void silk_LBRR_encode_FLP( +static OPUS_INLINE void silk_LBRR_encode_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ const silk_float xfw[], /* I Input signal */ diff --git a/code/opus-1.0.2/silk/float/energy_FLP.c b/code/opus-1.1/silk/float/energy_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/energy_FLP.c rename to code/opus-1.1/silk/float/energy_FLP.c index e3eedf97..24b8179f 100644 --- a/code/opus-1.0.2/silk/float/energy_FLP.c +++ b/code/opus-1.1/silk/float/energy_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/find_LPC_FLP.c b/code/opus-1.1/silk/float/find_LPC_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/find_LPC_FLP.c rename to code/opus-1.1/silk/float/find_LPC_FLP.c index 66fa7dd4..61c1ad95 100644 --- a/code/opus-1.0.2/silk/float/find_LPC_FLP.c +++ b/code/opus-1.1/silk/float/find_LPC_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/find_LTP_FLP.c b/code/opus-1.1/silk/float/find_LTP_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/find_LTP_FLP.c rename to code/opus-1.1/silk/float/find_LTP_FLP.c index 0a3c71bb..72299960 100644 --- a/code/opus-1.0.2/silk/float/find_LTP_FLP.c +++ b/code/opus-1.1/silk/float/find_LTP_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/find_pitch_lags_FLP.c b/code/opus-1.1/silk/float/find_pitch_lags_FLP.c similarity index 96% rename from code/opus-1.0.2/silk/float/find_pitch_lags_FLP.c rename to code/opus-1.1/silk/float/find_pitch_lags_FLP.c index 00862a6d..f3b22d25 100644 --- a/code/opus-1.0.2/silk/float/find_pitch_lags_FLP.c +++ b/code/opus-1.1/silk/float/find_pitch_lags_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -37,7 +37,8 @@ void silk_find_pitch_lags_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ silk_float res[], /* O Residual */ - const silk_float x[] /* I Speech signal */ + const silk_float x[], /* I Speech signal */ + int arch /* I Run-time architecture */ ) { opus_int buf_len; @@ -116,7 +117,7 @@ void silk_find_pitch_lags_FLP( /*****************************************/ if( silk_pitch_analysis_core_FLP( res, psEncCtrl->pitchL, &psEnc->sCmn.indices.lagIndex, &psEnc->sCmn.indices.contourIndex, &psEnc->LTPCorr, psEnc->sCmn.prevLag, psEnc->sCmn.pitchEstimationThreshold_Q16 / 65536.0f, - thrhld, psEnc->sCmn.fs_kHz, psEnc->sCmn.pitchEstimationComplexity, psEnc->sCmn.nb_subfr ) == 0 ) + thrhld, psEnc->sCmn.fs_kHz, psEnc->sCmn.pitchEstimationComplexity, psEnc->sCmn.nb_subfr, arch ) == 0 ) { psEnc->sCmn.indices.signalType = TYPE_VOICED; } else { diff --git a/code/opus-1.0.2/silk/float/find_pred_coefs_FLP.c b/code/opus-1.1/silk/float/find_pred_coefs_FLP.c similarity index 97% rename from code/opus-1.0.2/silk/float/find_pred_coefs_FLP.c rename to code/opus-1.1/silk/float/find_pred_coefs_FLP.c index 2156893a..ea2c6c43 100644 --- a/code/opus-1.0.2/silk/float/find_pred_coefs_FLP.c +++ b/code/opus-1.1/silk/float/find_pred_coefs_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -67,7 +67,7 @@ void silk_find_pred_coefs_FLP( /* Quantize LTP gain parameters */ silk_quant_LTP_gains_FLP( psEncCtrl->LTPCoef, psEnc->sCmn.indices.LTPIndex, &psEnc->sCmn.indices.PERIndex, - WLTP, psEnc->sCmn.mu_LTP_Q9, psEnc->sCmn.LTPQuantLowComplexity, psEnc->sCmn.nb_subfr ); + &psEnc->sCmn.sum_log_gain_Q7, WLTP, psEnc->sCmn.mu_LTP_Q9, psEnc->sCmn.LTPQuantLowComplexity, psEnc->sCmn.nb_subfr ); /* Control LTP scaling */ silk_LTP_scale_ctrl_FLP( psEnc, psEncCtrl, condCoding ); @@ -90,6 +90,7 @@ void silk_find_pred_coefs_FLP( } silk_memset( psEncCtrl->LTPCoef, 0, psEnc->sCmn.nb_subfr * LTP_ORDER * sizeof( silk_float ) ); psEncCtrl->LTPredCodGain = 0.0f; + psEnc->sCmn.sum_log_gain_Q7 = 0; } /* Limit on total predictive coding gain */ diff --git a/code/opus-1.0.2/silk/float/inner_product_FLP.c b/code/opus-1.1/silk/float/inner_product_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/inner_product_FLP.c rename to code/opus-1.1/silk/float/inner_product_FLP.c index 60823d6e..029c0129 100644 --- a/code/opus-1.0.2/silk/float/inner_product_FLP.c +++ b/code/opus-1.1/silk/float/inner_product_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/k2a_FLP.c b/code/opus-1.1/silk/float/k2a_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/k2a_FLP.c rename to code/opus-1.1/silk/float/k2a_FLP.c index 6f05d4b9..12af4e76 100644 --- a/code/opus-1.0.2/silk/float/k2a_FLP.c +++ b/code/opus-1.1/silk/float/k2a_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/levinsondurbin_FLP.c b/code/opus-1.1/silk/float/levinsondurbin_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/levinsondurbin_FLP.c rename to code/opus-1.1/silk/float/levinsondurbin_FLP.c index b4cd34e2..f0ba6069 100644 --- a/code/opus-1.0.2/silk/float/levinsondurbin_FLP.c +++ b/code/opus-1.1/silk/float/levinsondurbin_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/main_FLP.h b/code/opus-1.1/silk/float/main_FLP.h similarity index 97% rename from code/opus-1.0.2/silk/float/main_FLP.h rename to code/opus-1.1/silk/float/main_FLP.h index 93455d4d..fb553b61 100644 --- a/code/opus-1.0.2/silk/float/main_FLP.h +++ b/code/opus-1.1/silk/float/main_FLP.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -71,7 +71,8 @@ opus_int silk_encode_frame_FLP( /* Initializes the Silk encoder state */ opus_int silk_init_encoder( - silk_encoder_state_FLP *psEnc /* I/O Encoder state FLP */ + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + int arch /* I Run-tim architecture */ ); /* Control the Silk encoder */ @@ -129,7 +130,8 @@ void silk_find_pitch_lags_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ silk_float res[], /* O Residual */ - const silk_float x[] /* I Speech signal */ + const silk_float x[], /* I Speech signal */ + int arch /* I Run-time architecture */ ); /* Find LPC and LTP coefficients */ @@ -199,6 +201,7 @@ void silk_quant_LTP_gains_FLP( silk_float B[ MAX_NB_SUBFR * LTP_ORDER ], /* I/O (Un-)quantized LTP gains */ opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook index */ opus_int8 *periodicity_index, /* O Periodicity index */ + opus_int32 *sum_log_gain_Q7, /* I/O Cumulative max prediction gain */ const silk_float W[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* I Error weights */ const opus_int mu_Q10, /* I Mu value (R/D tradeoff) */ const opus_int lowComplexity, /* I Flag for low complexity */ diff --git a/code/opus-1.0.2/silk/float/noise_shape_analysis_FLP.c b/code/opus-1.1/silk/float/noise_shape_analysis_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/noise_shape_analysis_FLP.c rename to code/opus-1.1/silk/float/noise_shape_analysis_FLP.c index 33bfd20d..65f6ea58 100644 --- a/code/opus-1.0.2/silk/float/noise_shape_analysis_FLP.c +++ b/code/opus-1.1/silk/float/noise_shape_analysis_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -36,7 +36,7 @@ POSSIBILITY OF SUCH DAMAGE. /* non-warped frequency scale. (So that it can be implemented with a minimum-phase monic filter.) */ /* Note: A monic filter is one with the first coefficient equal to 1.0. In Silk we omit the first */ /* coefficient in an array of coefficients, for monic filters. */ -static inline silk_float warped_gain( +static OPUS_INLINE silk_float warped_gain( const silk_float *coefs, silk_float lambda, opus_int order @@ -54,7 +54,7 @@ static inline silk_float warped_gain( /* Convert warped filter coefficients to monic pseudo-warped coefficients and limit maximum */ /* amplitude of monic warped coefficients by using bandwidth expansion on the true coefficients */ -static inline void warped_true2monic_coefs( +static OPUS_INLINE void warped_true2monic_coefs( silk_float *coefs_syn, silk_float *coefs_ana, silk_float lambda, diff --git a/code/opus-1.0.2/silk/float/pitch_analysis_core_FLP.c b/code/opus-1.1/silk/float/pitch_analysis_core_FLP.c similarity index 87% rename from code/opus-1.0.2/silk/float/pitch_analysis_core_FLP.c rename to code/opus-1.1/silk/float/pitch_analysis_core_FLP.c index fbff90c3..e58f041b 100644 --- a/code/opus-1.0.2/silk/float/pitch_analysis_core_FLP.c +++ b/code/opus-1.1/silk/float/pitch_analysis_core_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -35,9 +35,9 @@ POSSIBILITY OF SUCH DAMAGE. #include "SigProc_FLP.h" #include "SigProc_FIX.h" #include "pitch_est_defines.h" +#include "pitch.h" #define SCRATCH_SIZE 22 -#define eps 1.192092896e-07f /************************************************************/ /* Internally used functions */ @@ -48,7 +48,8 @@ static void silk_P_Ana_calc_corr_st3( opus_int start_lag, /* I start lag */ opus_int sf_length, /* I sub frame length */ opus_int nb_subfr, /* I number of subframes */ - opus_int complexity /* I Complexity setting */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ ); static void silk_P_Ana_calc_energy_st3( @@ -74,7 +75,8 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, const silk_float search_thres2, /* I Final threshold for lag candidates 0 - 1 */ const opus_int Fs_kHz, /* I sample frequency (kHz) */ const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ - const opus_int nb_subfr /* I Number of 5 ms subframes */ + const opus_int nb_subfr, /* I Number of 5 ms subframes */ + int arch /* I Run-time architecture */ ) { opus_int i, k, d, j; @@ -85,6 +87,7 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, opus_int32 filt_state[ 6 ]; silk_float threshold, contour_bias; silk_float C[ PE_MAX_NB_SUBFR][ (PE_MAX_LAG >> 1) + 5 ]; + opus_val32 xcorr[ PE_MAX_LAG_MS * 4 - PE_MIN_LAG_MS * 4 + 1 ]; silk_float CC[ PE_NB_CBKS_STAGE2_EXT ]; const silk_float *target_ptr, *basis_ptr; double cross_corr, normalizer, energy, energy_tmp; @@ -129,8 +132,6 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, max_lag_4kHz = PE_MAX_LAG_MS * 4; max_lag_8kHz = PE_MAX_LAG_MS * 8 - 1; - silk_memset(C, 0, sizeof(silk_float) * nb_subfr * ((PE_MAX_LAG >> 1) + 5)); - /* Resample from input sampled at Fs_kHz to 8 kHz */ if( Fs_kHz == 16 ) { /* Resample to 16 -> 8 khz */ @@ -164,6 +165,7 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, /****************************************************************************** * FIRST STAGE, operating in 4 khz ******************************************************************************/ + silk_memset(C, 0, sizeof(silk_float) * nb_subfr * ((PE_MAX_LAG >> 1) + 5)); target_ptr = &frame_4kHz[ silk_LSHIFT( sf_length_4kHz, 2 ) ]; for( k = 0; k < nb_subfr >> 1; k++ ) { /* Check that we are within range of the array */ @@ -176,27 +178,31 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, silk_assert( basis_ptr >= frame_4kHz ); silk_assert( basis_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); - /* Calculate first vector products before loop */ - cross_corr = silk_inner_product_FLP( target_ptr, basis_ptr, sf_length_8kHz ); - normalizer = silk_energy_FLP( basis_ptr, sf_length_8kHz ) + sf_length_8kHz * 4000.0f; + celt_pitch_xcorr( target_ptr, target_ptr-max_lag_4kHz, xcorr, sf_length_8kHz, max_lag_4kHz - min_lag_4kHz + 1, arch ); - C[ 0 ][ min_lag_4kHz ] += (silk_float)(cross_corr / sqrt(normalizer)); + /* Calculate first vector products before loop */ + cross_corr = xcorr[ max_lag_4kHz - min_lag_4kHz ]; + normalizer = silk_energy_FLP( target_ptr, sf_length_8kHz ) + + silk_energy_FLP( basis_ptr, sf_length_8kHz ) + + sf_length_8kHz * 4000.0f; + + C[ 0 ][ min_lag_4kHz ] += (silk_float)( 2 * cross_corr / normalizer ); /* From now on normalizer is computed recursively */ - for(d = min_lag_4kHz + 1; d <= max_lag_4kHz; d++) { + for( d = min_lag_4kHz + 1; d <= max_lag_4kHz; d++ ) { basis_ptr--; /* Check that we are within range of the array */ silk_assert( basis_ptr >= frame_4kHz ); silk_assert( basis_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); - cross_corr = silk_inner_product_FLP(target_ptr, basis_ptr, sf_length_8kHz); + cross_corr = xcorr[ max_lag_4kHz - d ]; /* Add contribution of new sample and remove contribution from oldest sample */ normalizer += basis_ptr[ 0 ] * (double)basis_ptr[ 0 ] - basis_ptr[ sf_length_8kHz ] * (double)basis_ptr[ sf_length_8kHz ]; - C[ 0 ][ d ] += (silk_float)(cross_corr / sqrt( normalizer )); + C[ 0 ][ d ] += (silk_float)( 2 * cross_corr / normalizer ); } /* Update target pointer */ target_ptr += sf_length_8kHz; @@ -214,13 +220,7 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, /* Escape if correlation is very low already here */ Cmax = C[ 0 ][ min_lag_4kHz ]; - target_ptr = &frame_4kHz[ silk_SMULBB( sf_length_4kHz, nb_subfr ) ]; - energy = 1000.0f; - for( i = 0; i < silk_LSHIFT( sf_length_4kHz, 2 ); i++ ) { - energy += target_ptr[i] * (double)target_ptr[i]; - } - threshold = Cmax * Cmax; - if( energy / 16.0f > threshold ) { + if( Cmax < 0.2f ) { silk_memset( pitch_out, 0, nb_subfr * sizeof( opus_int ) ); *LTPCorr = 0.0f; *lagIndex = 0; @@ -287,14 +287,14 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, target_ptr = &frame_8kHz[ PE_LTP_MEM_LENGTH_MS * 8 ]; } for( k = 0; k < nb_subfr; k++ ) { - energy_tmp = silk_energy_FLP( target_ptr, sf_length_8kHz ); + energy_tmp = silk_energy_FLP( target_ptr, sf_length_8kHz ) + 1.0; for( j = 0; j < length_d_comp; j++ ) { d = d_comp[ j ]; basis_ptr = target_ptr - d; cross_corr = silk_inner_product_FLP( basis_ptr, target_ptr, sf_length_8kHz ); - energy = silk_energy_FLP( basis_ptr, sf_length_8kHz ); if( cross_corr > 0.0f ) { - C[ k ][ d ] = (silk_float)(cross_corr * cross_corr / (energy * energy_tmp + eps)); + energy = silk_energy_FLP( basis_ptr, sf_length_8kHz ); + C[ k ][ d ] = (silk_float)( 2 * cross_corr / ( energy + energy_tmp ) ); } else { C[ k ][ d ] = 0.0f; } @@ -317,7 +317,7 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, } else if( Fs_kHz == 16 ) { prevLag = silk_RSHIFT( prevLag, 1 ); } - prevLag_log2 = silk_log2((silk_float)prevLag); + prevLag_log2 = silk_log2( (silk_float)prevLag ); } else { prevLag_log2 = 0; } @@ -356,23 +356,20 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, CBimax_new = i; } } - CCmax_new = silk_max_float(CCmax_new, 0.0f); /* To avoid taking square root of negative number later */ - CCmax_new_b = CCmax_new; /* Bias towards shorter lags */ - lag_log2 = silk_log2((silk_float)d); - CCmax_new_b -= PE_SHORTLAG_BIAS * nb_subfr * lag_log2; + lag_log2 = silk_log2( (silk_float)d ); + CCmax_new_b = CCmax_new - PE_SHORTLAG_BIAS * nb_subfr * lag_log2; /* Bias towards previous lag */ if( prevLag > 0 ) { delta_lag_log2_sqr = lag_log2 - prevLag_log2; delta_lag_log2_sqr *= delta_lag_log2_sqr; - CCmax_new_b -= PE_PREVLAG_BIAS * nb_subfr * (*LTPCorr) * delta_lag_log2_sqr / (delta_lag_log2_sqr + 0.5f); + CCmax_new_b -= PE_PREVLAG_BIAS * nb_subfr * (*LTPCorr) * delta_lag_log2_sqr / ( delta_lag_log2_sqr + 0.5f ); } - if( CCmax_new_b > CCmax_b && /* Find maximum biased correlation */ - CCmax_new > nb_subfr * search_thres2 * search_thres2 && /* Correlation needs to be high enough to be voiced */ - silk_CB_lags_stage2[ 0 ][ CBimax_new ] <= min_lag_8kHz /* Lag must be in range */ + if( CCmax_new_b > CCmax_b && /* Find maximum biased correlation */ + CCmax_new > nb_subfr * search_thres2 /* Correlation needs to be high enough to be voiced */ ) { CCmax_b = CCmax_new_b; CCmax = CCmax_new; @@ -390,6 +387,10 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, return 1; } + /* Output normalized correlation */ + *LTPCorr = (silk_float)( CCmax / nb_subfr ); + silk_assert( *LTPCorr >= 0.0f ); + if( Fs_kHz > 8 ) { /* Search in original signal */ @@ -406,13 +407,11 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, end_lag = silk_min_int( lag + 2, max_lag ); lag_new = lag; /* to avoid undefined lag */ CBimax = 0; /* to avoid undefined lag */ - silk_assert( CCmax >= 0.0f ); - *LTPCorr = (silk_float)sqrt( CCmax / nb_subfr ); /* Output normalized correlation */ CCmax = -1000.0f; /* Calculate the correlations and energies needed in stage 3 */ - silk_P_Ana_calc_corr_st3( cross_corr_st3, frame, start_lag, sf_length, nb_subfr, complexity ); + silk_P_Ana_calc_corr_st3( cross_corr_st3, frame, start_lag, sf_length, nb_subfr, complexity, arch ); silk_P_Ana_calc_energy_st3( energies_st3, frame, start_lag, sf_length, nb_subfr, complexity ); lag_counter = 0; @@ -430,25 +429,25 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; } + target_ptr = &frame[ PE_LTP_MEM_LENGTH_MS * Fs_kHz ]; + energy_tmp = silk_energy_FLP( target_ptr, nb_subfr * sf_length ) + 1.0; for( d = start_lag; d <= end_lag; d++ ) { for( j = 0; j < nb_cbk_search; j++ ) { cross_corr = 0.0; - energy = eps; + energy = energy_tmp; for( k = 0; k < nb_subfr; k++ ) { - energy += energies_st3[ k ][ j ][ lag_counter ]; cross_corr += cross_corr_st3[ k ][ j ][ lag_counter ]; + energy += energies_st3[ k ][ j ][ lag_counter ]; } if( cross_corr > 0.0 ) { - CCmax_new = (silk_float)(cross_corr * cross_corr / energy); + CCmax_new = (silk_float)( 2 * cross_corr / energy ); /* Reduce depending on flatness of contour */ CCmax_new *= 1.0f - contour_bias * j; } else { CCmax_new = 0.0f; } - if( CCmax_new > CCmax && - ( d + (opus_int)silk_CB_lags_stage3[ 0 ][ j ] ) <= max_lag - ) { + if( CCmax_new > CCmax && ( d + (opus_int)silk_CB_lags_stage3[ 0 ][ j ] ) <= max_lag ) { CCmax = CCmax_new; lag_new = d; CBimax = j; @@ -464,12 +463,10 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, *lagIndex = (opus_int16)( lag_new - min_lag ); *contourIndex = (opus_int8)CBimax; } else { /* Fs_kHz == 8 */ - /* Save Lags and correlation */ - silk_assert( CCmax >= 0.0f ); - *LTPCorr = (silk_float)sqrt( CCmax / nb_subfr ); /* Output normalized correlation */ + /* Save Lags */ for( k = 0; k < nb_subfr; k++ ) { pitch_out[ k ] = lag + matrix_ptr( Lag_CB_ptr, k, CBimax, cbk_size ); - pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], min_lag_8kHz, PE_MAX_LAG_MS * Fs_kHz ); + pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], min_lag_8kHz, PE_MAX_LAG_MS * 8 ); } *lagIndex = (opus_int16)( lag - min_lag_8kHz ); *contourIndex = (opus_int8)CBimax; @@ -479,32 +476,34 @@ opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, return 0; } +/*********************************************************************** + * Calculates the correlations used in stage 3 search. In order to cover + * the whole lag codebook for all the searched offset lags (lag +- 2), + * the following correlations are needed in each sub frame: + * + * sf1: lag range [-8,...,7] total 16 correlations + * sf2: lag range [-4,...,4] total 9 correlations + * sf3: lag range [-3,....4] total 8 correltions + * sf4: lag range [-6,....8] total 15 correlations + * + * In total 48 correlations. The direct implementation computed in worst + * case 4*12*5 = 240 correlations, but more likely around 120. + ***********************************************************************/ static void silk_P_Ana_calc_corr_st3( silk_float cross_corr_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ], /* O 3 DIM correlation array */ const silk_float frame[], /* I vector to correlate */ opus_int start_lag, /* I start lag */ opus_int sf_length, /* I sub frame length */ opus_int nb_subfr, /* I number of subframes */ - opus_int complexity /* I Complexity setting */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ ) - /*********************************************************************** - Calculates the correlations used in stage 3 search. In order to cover - the whole lag codebook for all the searched offset lags (lag +- 2), - the following correlations are needed in each sub frame: - - sf1: lag range [-8,...,7] total 16 correlations - sf2: lag range [-4,...,4] total 9 correlations - sf3: lag range [-3,....4] total 8 correltions - sf4: lag range [-6,....8] total 15 correlations - - In total 48 correlations. The direct implementation computed in worst case - 4*12*5 = 240 correlations, but more likely around 120. - **********************************************************************/ { - const silk_float *target_ptr, *basis_ptr; + const silk_float *target_ptr; opus_int i, j, k, lag_counter, lag_low, lag_high; opus_int nb_cbk_search, delta, idx, cbk_size; silk_float scratch_mem[ SCRATCH_SIZE ]; + opus_val32 xcorr[ SCRATCH_SIZE ]; const opus_int8 *Lag_range_ptr, *Lag_CB_ptr; silk_assert( complexity >= SILK_PE_MIN_COMPLEX ); @@ -530,10 +529,11 @@ static void silk_P_Ana_calc_corr_st3( /* Calculate the correlations for each subframe */ lag_low = matrix_ptr( Lag_range_ptr, k, 0, 2 ); lag_high = matrix_ptr( Lag_range_ptr, k, 1, 2 ); + silk_assert(lag_high-lag_low+1 <= SCRATCH_SIZE); + celt_pitch_xcorr( target_ptr, target_ptr - start_lag - lag_high, xcorr, sf_length, lag_high - lag_low + 1, arch ); for( j = lag_low; j <= lag_high; j++ ) { - basis_ptr = target_ptr - ( start_lag + j ); silk_assert( lag_counter < SCRATCH_SIZE ); - scratch_mem[ lag_counter ] = (silk_float)silk_inner_product_FLP( target_ptr, basis_ptr, sf_length ); + scratch_mem[ lag_counter ] = xcorr[ lag_high - j ]; lag_counter++; } @@ -552,6 +552,10 @@ static void silk_P_Ana_calc_corr_st3( } } +/********************************************************************/ +/* Calculate the energies for first two subframes. The energies are */ +/* calculated recursively. */ +/********************************************************************/ static void silk_P_Ana_calc_energy_st3( silk_float energies_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ], /* O 3 DIM correlation array */ const silk_float frame[], /* I vector to correlate */ @@ -560,10 +564,6 @@ static void silk_P_Ana_calc_energy_st3( opus_int nb_subfr, /* I number of subframes */ opus_int complexity /* I Complexity setting */ ) -/**************************************************************** -Calculate the energies for first two subframes. The energies are -calculated recursively. -****************************************************************/ { const silk_float *target_ptr, *basis_ptr; double energy; diff --git a/code/opus-1.0.2/silk/float/prefilter_FLP.c b/code/opus-1.1/silk/float/prefilter_FLP.c similarity index 98% rename from code/opus-1.0.2/silk/float/prefilter_FLP.c rename to code/opus-1.1/silk/float/prefilter_FLP.c index d6c84398..8bc32fb4 100644 --- a/code/opus-1.0.2/silk/float/prefilter_FLP.c +++ b/code/opus-1.1/silk/float/prefilter_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -35,7 +35,7 @@ POSSIBILITY OF SUCH DAMAGE. /* * Prefilter for finding Quantizer input signal */ -static inline void silk_prefilt_FLP( +static OPUS_INLINE void silk_prefilt_FLP( silk_prefilter_state_FLP *P, /* I/O state */ silk_float st_res[], /* I */ silk_float xw[], /* O */ @@ -153,7 +153,7 @@ void silk_prefilter_FLP( /* * Prefilter for finding Quantizer input signal */ -static inline void silk_prefilt_FLP( +static OPUS_INLINE void silk_prefilt_FLP( silk_prefilter_state_FLP *P, /* I/O state */ silk_float st_res[], /* I */ silk_float xw[], /* O */ diff --git a/code/opus-1.0.2/silk/float/process_gains_FLP.c b/code/opus-1.1/silk/float/process_gains_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/process_gains_FLP.c rename to code/opus-1.1/silk/float/process_gains_FLP.c index d572a4cd..c0da0dae 100644 --- a/code/opus-1.0.2/silk/float/process_gains_FLP.c +++ b/code/opus-1.1/silk/float/process_gains_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/regularize_correlations_FLP.c b/code/opus-1.1/silk/float/regularize_correlations_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/regularize_correlations_FLP.c rename to code/opus-1.1/silk/float/regularize_correlations_FLP.c index f5684637..df461260 100644 --- a/code/opus-1.0.2/silk/float/regularize_correlations_FLP.c +++ b/code/opus-1.1/silk/float/regularize_correlations_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/residual_energy_FLP.c b/code/opus-1.1/silk/float/residual_energy_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/residual_energy_FLP.c rename to code/opus-1.1/silk/float/residual_energy_FLP.c index e65457ab..b2e03a86 100644 --- a/code/opus-1.0.2/silk/float/residual_energy_FLP.c +++ b/code/opus-1.1/silk/float/residual_energy_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/scale_copy_vector_FLP.c b/code/opus-1.1/silk/float/scale_copy_vector_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/scale_copy_vector_FLP.c rename to code/opus-1.1/silk/float/scale_copy_vector_FLP.c index 988795a6..20db32b3 100644 --- a/code/opus-1.0.2/silk/float/scale_copy_vector_FLP.c +++ b/code/opus-1.1/silk/float/scale_copy_vector_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/scale_vector_FLP.c b/code/opus-1.1/silk/float/scale_vector_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/scale_vector_FLP.c rename to code/opus-1.1/silk/float/scale_vector_FLP.c index 387eb4ba..108fdcbe 100644 --- a/code/opus-1.0.2/silk/float/scale_vector_FLP.c +++ b/code/opus-1.1/silk/float/scale_vector_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/schur_FLP.c b/code/opus-1.1/silk/float/schur_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/schur_FLP.c rename to code/opus-1.1/silk/float/schur_FLP.c index 90c3a18b..ee436f83 100644 --- a/code/opus-1.0.2/silk/float/schur_FLP.c +++ b/code/opus-1.1/silk/float/schur_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/solve_LS_FLP.c b/code/opus-1.1/silk/float/solve_LS_FLP.c similarity index 96% rename from code/opus-1.0.2/silk/float/solve_LS_FLP.c rename to code/opus-1.1/silk/float/solve_LS_FLP.c index a4bb0525..7c90d665 100644 --- a/code/opus-1.0.2/silk/float/solve_LS_FLP.c +++ b/code/opus-1.1/silk/float/solve_LS_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -37,7 +37,7 @@ POSSIBILITY OF SUCH DAMAGE. * Matrix D (only the diagonal elements returned in a vector)such that * the symmetric matric A is given by A = L*D*L'. **********************************************************************/ -static inline void silk_LDL_FLP( +static OPUS_INLINE void silk_LDL_FLP( silk_float *A, /* I/O Pointer to Symetric Square Matrix */ opus_int M, /* I Size of Matrix */ silk_float *L, /* I/O Pointer to Square Upper triangular Matrix */ @@ -48,7 +48,7 @@ static inline void silk_LDL_FLP( * Function to solve linear equation Ax = b, when A is a MxM lower * triangular matrix, with ones on the diagonal. **********************************************************************/ -static inline void silk_SolveWithLowerTriangularWdiagOnes_FLP( +static OPUS_INLINE void silk_SolveWithLowerTriangularWdiagOnes_FLP( const silk_float *L, /* I Pointer to Lower Triangular Matrix */ opus_int M, /* I Dim of Matrix equation */ const silk_float *b, /* I b Vector */ @@ -59,7 +59,7 @@ static inline void silk_SolveWithLowerTriangularWdiagOnes_FLP( * Function to solve linear equation (A^T)x = b, when A is a MxM lower * triangular, with ones on the diagonal. (ie then A^T is upper triangular) **********************************************************************/ -static inline void silk_SolveWithUpperTriangularFromLowerWdiagOnes_FLP( +static OPUS_INLINE void silk_SolveWithUpperTriangularFromLowerWdiagOnes_FLP( const silk_float *L, /* I Pointer to Lower Triangular Matrix */ opus_int M, /* I Dim of Matrix equation */ const silk_float *b, /* I b Vector */ @@ -109,7 +109,7 @@ void silk_solve_LDL_FLP( silk_SolveWithUpperTriangularFromLowerWdiagOnes_FLP( &L[ 0 ][ 0 ], M, T, x ); } -static inline void silk_SolveWithUpperTriangularFromLowerWdiagOnes_FLP( +static OPUS_INLINE void silk_SolveWithUpperTriangularFromLowerWdiagOnes_FLP( const silk_float *L, /* I Pointer to Lower Triangular Matrix */ opus_int M, /* I Dim of Matrix equation */ const silk_float *b, /* I b Vector */ @@ -131,7 +131,7 @@ static inline void silk_SolveWithUpperTriangularFromLowerWdiagOnes_FLP( } } -static inline void silk_SolveWithLowerTriangularWdiagOnes_FLP( +static OPUS_INLINE void silk_SolveWithLowerTriangularWdiagOnes_FLP( const silk_float *L, /* I Pointer to Lower Triangular Matrix */ opus_int M, /* I Dim of Matrix equation */ const silk_float *b, /* I b Vector */ @@ -153,7 +153,7 @@ static inline void silk_SolveWithLowerTriangularWdiagOnes_FLP( } } -static inline void silk_LDL_FLP( +static OPUS_INLINE void silk_LDL_FLP( silk_float *A, /* I/O Pointer to Symetric Square Matrix */ opus_int M, /* I Size of Matrix */ silk_float *L, /* I/O Pointer to Square Upper triangular Matrix */ diff --git a/code/opus-1.0.2/silk/float/sort_FLP.c b/code/opus-1.1/silk/float/sort_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/sort_FLP.c rename to code/opus-1.1/silk/float/sort_FLP.c index e290c380..f08d7592 100644 --- a/code/opus-1.0.2/silk/float/sort_FLP.c +++ b/code/opus-1.1/silk/float/sort_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/structs_FLP.h b/code/opus-1.1/silk/float/structs_FLP.h similarity index 99% rename from code/opus-1.0.2/silk/float/structs_FLP.h rename to code/opus-1.1/silk/float/structs_FLP.h index c71e7bc3..bb529e71 100644 --- a/code/opus-1.0.2/silk/float/structs_FLP.h +++ b/code/opus-1.1/silk/float/structs_FLP.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/warped_autocorrelation_FLP.c b/code/opus-1.1/silk/float/warped_autocorrelation_FLP.c similarity index 99% rename from code/opus-1.0.2/silk/float/warped_autocorrelation_FLP.c rename to code/opus-1.1/silk/float/warped_autocorrelation_FLP.c index e9ecc2a3..542414f4 100644 --- a/code/opus-1.0.2/silk/float/warped_autocorrelation_FLP.c +++ b/code/opus-1.1/silk/float/warped_autocorrelation_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/float/wrappers_FLP.c b/code/opus-1.1/silk/float/wrappers_FLP.c similarity index 97% rename from code/opus-1.0.2/silk/float/wrappers_FLP.c rename to code/opus-1.1/silk/float/wrappers_FLP.c index 4259e90e..350599b2 100644 --- a/code/opus-1.0.2/silk/float/wrappers_FLP.c +++ b/code/opus-1.1/silk/float/wrappers_FLP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -175,6 +175,7 @@ void silk_quant_LTP_gains_FLP( silk_float B[ MAX_NB_SUBFR * LTP_ORDER ], /* I/O (Un-)quantized LTP gains */ opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook index */ opus_int8 *periodicity_index, /* O Periodicity index */ + opus_int32 *sum_log_gain_Q7, /* I/O Cumulative max prediction gain */ const silk_float W[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* I Error weights */ const opus_int mu_Q10, /* I Mu value (R/D tradeoff) */ const opus_int lowComplexity, /* I Flag for low complexity */ @@ -192,7 +193,7 @@ void silk_quant_LTP_gains_FLP( W_Q18[ i ] = (opus_int32)silk_float2int( W[ i ] * 262144.0f ); } - silk_quant_LTP_gains( B_Q14, cbk_index, periodicity_index, W_Q18, mu_Q10, lowComplexity, nb_subfr ); + silk_quant_LTP_gains( B_Q14, cbk_index, periodicity_index, sum_log_gain_Q7, W_Q18, mu_Q10, lowComplexity, nb_subfr ); for( i = 0; i < nb_subfr * LTP_ORDER; i++ ) { B[ i ] = (silk_float)B_Q14[ i ] * ( 1.0f / 16384.0f ); diff --git a/code/opus-1.0.2/silk/gain_quant.c b/code/opus-1.1/silk/gain_quant.c similarity index 99% rename from code/opus-1.0.2/silk/gain_quant.c rename to code/opus-1.1/silk/gain_quant.c index b2f73735..64ccd061 100644 --- a/code/opus-1.0.2/silk/gain_quant.c +++ b/code/opus-1.1/silk/gain_quant.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/init_decoder.c b/code/opus-1.1/silk/init_decoder.c similarity index 99% rename from code/opus-1.0.2/silk/init_decoder.c rename to code/opus-1.1/silk/init_decoder.c index 47834890..f887c678 100644 --- a/code/opus-1.0.2/silk/init_decoder.c +++ b/code/opus-1.1/silk/init_decoder.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/init_encoder.c b/code/opus-1.1/silk/init_encoder.c similarity index 89% rename from code/opus-1.0.2/silk/init_encoder.c rename to code/opus-1.1/silk/init_encoder.c index fe3fe968..65995c33 100644 --- a/code/opus-1.0.2/silk/init_encoder.c +++ b/code/opus-1.1/silk/init_encoder.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -34,12 +34,14 @@ POSSIBILITY OF SUCH DAMAGE. #include "main_FLP.h" #endif #include "tuning_parameters.h" +#include "cpu_support.h" /*********************************/ /* Initialize Silk Encoder state */ /*********************************/ opus_int silk_init_encoder( - silk_encoder_state_Fxx *psEnc /* I/O Pointer to Silk FIX encoder state */ + silk_encoder_state_Fxx *psEnc, /* I/O Pointer to Silk FIX encoder state */ + int arch /* I Run-time architecture */ ) { opus_int ret = 0; @@ -47,6 +49,8 @@ opus_int silk_init_encoder( /* Clear the entire encoder state */ silk_memset( psEnc, 0, sizeof( silk_encoder_state_Fxx ) ); + psEnc->sCmn.arch = arch; + psEnc->sCmn.variable_HP_smth1_Q15 = silk_LSHIFT( silk_lin2log( SILK_FIX_CONST( VARIABLE_HP_MIN_CUTOFF_HZ, 16 ) ) - ( 16 << 7 ), 8 ); psEnc->sCmn.variable_HP_smth2_Q15 = psEnc->sCmn.variable_HP_smth1_Q15; diff --git a/code/opus-1.0.2/silk/inner_prod_aligned.c b/code/opus-1.1/silk/inner_prod_aligned.c similarity index 99% rename from code/opus-1.0.2/silk/inner_prod_aligned.c rename to code/opus-1.1/silk/inner_prod_aligned.c index fe20a2b1..257ae9e0 100644 --- a/code/opus-1.0.2/silk/inner_prod_aligned.c +++ b/code/opus-1.1/silk/inner_prod_aligned.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/interpolate.c b/code/opus-1.1/silk/interpolate.c similarity index 99% rename from code/opus-1.0.2/silk/interpolate.c rename to code/opus-1.1/silk/interpolate.c index 226488b9..1bd8ca4d 100644 --- a/code/opus-1.0.2/silk/interpolate.c +++ b/code/opus-1.1/silk/interpolate.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/lin2log.c b/code/opus-1.1/silk/lin2log.c similarity index 99% rename from code/opus-1.0.2/silk/lin2log.c rename to code/opus-1.1/silk/lin2log.c index 212b670d..d4fe5153 100644 --- a/code/opus-1.0.2/silk/lin2log.c +++ b/code/opus-1.1/silk/lin2log.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/log2lin.c b/code/opus-1.1/silk/log2lin.c similarity index 96% rename from code/opus-1.0.2/silk/log2lin.c rename to code/opus-1.1/silk/log2lin.c index 33a19ad1..a692e009 100644 --- a/code/opus-1.0.2/silk/log2lin.c +++ b/code/opus-1.1/silk/log2lin.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -41,7 +41,9 @@ opus_int32 silk_log2lin( if( inLog_Q7 < 0 ) { return 0; - } + } else if ( inLog_Q7 >= 3967 ) { + return silk_int32_MAX; + } out = silk_LSHIFT( 1, silk_RSHIFT( inLog_Q7, 7 ) ); frac_Q7 = inLog_Q7 & 0x7F; diff --git a/code/opus-1.0.2/silk/macros.h b/code/opus-1.1/silk/macros.h similarity index 76% rename from code/opus-1.0.2/silk/macros.h rename to code/opus-1.1/silk/macros.h index 2612fc7a..a84e5a5d 100644 --- a/code/opus-1.0.2/silk/macros.h +++ b/code/opus-1.1/silk/macros.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -32,7 +32,10 @@ POSSIBILITY OF SUCH DAMAGE. #include "config.h" #endif -/* This is an inline header file for general platform. */ +#include "opus_types.h" +#include "opus_defines.h" + +/* This is an OPUS_INLINE header file for general platform. */ /* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */ #define silk_SMULWB(a32, b32) ((((a32) >> 16) * (opus_int32)((opus_int16)(b32))) + ((((a32) & 0x0000FFFF) * (opus_int32)((opus_int16)(b32))) >> 16)) @@ -76,59 +79,36 @@ POSSIBILITY OF SUCH DAMAGE. (( (a) & ((b)^0x80000000) & 0x80000000) ? silk_int32_MIN : (a)-(b)) : \ ((((a)^0x80000000) & (b) & 0x80000000) ? silk_int32_MAX : (a)-(b)) ) -static inline opus_int32 silk_CLZ16(opus_int16 in16) +#include "ecintrin.h" + +static OPUS_INLINE opus_int32 silk_CLZ16(opus_int16 in16) { - opus_int32 out32 = 0; - if( in16 == 0 ) { - return 16; - } - /* test nibbles */ - if( in16 & 0xFF00 ) { - if( in16 & 0xF000 ) { - in16 >>= 12; - } else { - out32 += 4; - in16 >>= 8; - } - } else { - if( in16 & 0xFFF0 ) { - out32 += 8; - in16 >>= 4; - } else { - out32 += 12; - } - } - /* test bits and return */ - if( in16 & 0xC ) { - if( in16 & 0x8 ) - return out32 + 0; - else - return out32 + 1; - } else { - if( in16 & 0xE ) - return out32 + 2; - else - return out32 + 3; - } + return 32 - EC_ILOG(in16<<16|0x8000); } -static inline opus_int32 silk_CLZ32(opus_int32 in32) +static OPUS_INLINE opus_int32 silk_CLZ32(opus_int32 in32) { - /* test highest 16 bits and convert to opus_int16 */ - if( in32 & 0xFFFF0000 ) { - return silk_CLZ16((opus_int16)(in32 >> 16)); - } else { - return silk_CLZ16((opus_int16)in32) + 16; - } + return in32 ? 32 - EC_ILOG(in32) : 32; } /* Row based */ -#define matrix_ptr(Matrix_base_adr, row, column, N) *(Matrix_base_adr + ((row)*(N)+(column))) -#define matrix_adr(Matrix_base_adr, row, column, N) (Matrix_base_adr + ((row)*(N)+(column))) +#define matrix_ptr(Matrix_base_adr, row, column, N) \ + (*((Matrix_base_adr) + ((row)*(N)+(column)))) +#define matrix_adr(Matrix_base_adr, row, column, N) \ + ((Matrix_base_adr) + ((row)*(N)+(column))) /* Column based */ #ifndef matrix_c_ptr -# define matrix_c_ptr(Matrix_base_adr, row, column, M) *(Matrix_base_adr + ((row)+(M)*(column))) +# define matrix_c_ptr(Matrix_base_adr, row, column, M) \ + (*((Matrix_base_adr) + ((row)+(M)*(column)))) +#endif + +#ifdef OPUS_ARM_INLINE_ASM +#include "arm/macros_armv4.h" +#endif + +#ifdef OPUS_ARM_INLINE_EDSP +#include "arm/macros_armv5e.h" #endif #endif /* SILK_MACROS_H */ diff --git a/code/opus-1.0.2/silk/main.h b/code/opus-1.1/silk/main.h similarity index 98% rename from code/opus-1.0.2/silk/main.h rename to code/opus-1.1/silk/main.h index 32675f69..2bdf8978 100644 --- a/code/opus-1.0.2/silk/main.h +++ b/code/opus-1.1/silk/main.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -204,6 +204,7 @@ void silk_quant_LTP_gains( opus_int16 B_Q14[ MAX_NB_SUBFR * LTP_ORDER ], /* I/O (un)quantized LTP gains */ opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook Index */ opus_int8 *periodicity_index, /* O Periodicity Index */ + opus_int32 *sum_gain_dB_Q7, /* I/O Cumulative max prediction gain */ const opus_int32 W_Q18[ MAX_NB_SUBFR*LTP_ORDER*LTP_ORDER ], /* I Error Weights in Q18 */ opus_int mu_Q9, /* I Mu value (R/D tradeoff) */ opus_int lowComplexity, /* I Flag for low complexity */ @@ -214,11 +215,14 @@ void silk_quant_LTP_gains( void silk_VQ_WMat_EC( opus_int8 *ind, /* O index of best codebook vector */ opus_int32 *rate_dist_Q14, /* O best weighted quant error + mu * rate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ const opus_int16 *in_Q14, /* I input vector to be quantized */ const opus_int32 *W_Q18, /* I weighting matrix */ const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ const opus_int mu_Q9, /* I tradeoff betw. weighted error and rate */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ opus_int L /* I number of vectors in codebook */ ); diff --git a/code/opus-1.0.2/silk/pitch_est_defines.h b/code/opus-1.1/silk/pitch_est_defines.h similarity index 99% rename from code/opus-1.0.2/silk/pitch_est_defines.h rename to code/opus-1.1/silk/pitch_est_defines.h index 0b6770eb..e1e4b5d7 100644 --- a/code/opus-1.0.2/silk/pitch_est_defines.h +++ b/code/opus-1.1/silk/pitch_est_defines.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/pitch_est_tables.c b/code/opus-1.1/silk/pitch_est_tables.c similarity index 99% rename from code/opus-1.0.2/silk/pitch_est_tables.c rename to code/opus-1.1/silk/pitch_est_tables.c index 7b139ed2..81a8baca 100644 --- a/code/opus-1.0.2/silk/pitch_est_tables.c +++ b/code/opus-1.1/silk/pitch_est_tables.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/process_NLSFs.c b/code/opus-1.1/silk/process_NLSFs.c similarity index 99% rename from code/opus-1.0.2/silk/process_NLSFs.c rename to code/opus-1.1/silk/process_NLSFs.c index 34ce7914..c27cf030 100644 --- a/code/opus-1.0.2/silk/process_NLSFs.c +++ b/code/opus-1.1/silk/process_NLSFs.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/quant_LTP_gains.c b/code/opus-1.1/silk/quant_LTP_gains.c similarity index 80% rename from code/opus-1.0.2/silk/quant_LTP_gains.c rename to code/opus-1.1/silk/quant_LTP_gains.c index f73c0f50..fd0870da 100644 --- a/code/opus-1.0.2/silk/quant_LTP_gains.c +++ b/code/opus-1.1/silk/quant_LTP_gains.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,11 +30,13 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "tuning_parameters.h" void silk_quant_LTP_gains( opus_int16 B_Q14[ MAX_NB_SUBFR * LTP_ORDER ], /* I/O (un)quantized LTP gains */ opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook Index */ opus_int8 *periodicity_index, /* O Periodicity Index */ + opus_int32 *sum_log_gain_Q7, /* I/O Cumulative max prediction gain */ const opus_int32 W_Q18[ MAX_NB_SUBFR*LTP_ORDER*LTP_ORDER ], /* I Error Weights in Q18 */ opus_int mu_Q9, /* I Mu value (R/D tradeoff) */ opus_int lowComplexity, /* I Flag for low complexity */ @@ -45,18 +47,26 @@ void silk_quant_LTP_gains( opus_int8 temp_idx[ MAX_NB_SUBFR ]; const opus_uint8 *cl_ptr_Q5; const opus_int8 *cbk_ptr_Q7; + const opus_uint8 *cbk_gain_ptr_Q7; const opus_int16 *b_Q14_ptr; const opus_int32 *W_Q18_ptr; opus_int32 rate_dist_Q14_subfr, rate_dist_Q14, min_rate_dist_Q14; + opus_int32 sum_log_gain_tmp_Q7, best_sum_log_gain_Q7, max_gain_Q7, gain_Q7; /***************************************************/ /* iterate over different codebooks with different */ /* rates/distortions, and choose best */ /***************************************************/ min_rate_dist_Q14 = silk_int32_MAX; + best_sum_log_gain_Q7 = 0; for( k = 0; k < 3; k++ ) { + /* Safety margin for pitch gain control, to take into account factors + such as state rescaling/rewhitening. */ + opus_int32 gain_safety = SILK_FIX_CONST( 0.4, 7 ); + cl_ptr_Q5 = silk_LTP_gain_BITS_Q5_ptrs[ k ]; cbk_ptr_Q7 = silk_LTP_vq_ptrs_Q7[ k ]; + cbk_gain_ptr_Q7 = silk_LTP_vq_gain_ptrs_Q7[ k ]; cbk_size = silk_LTP_vq_sizes[ k ]; /* Set up pointer to first subframe */ @@ -64,19 +74,28 @@ void silk_quant_LTP_gains( b_Q14_ptr = B_Q14; rate_dist_Q14 = 0; + sum_log_gain_tmp_Q7 = *sum_log_gain_Q7; for( j = 0; j < nb_subfr; j++ ) { + max_gain_Q7 = silk_log2lin( ( SILK_FIX_CONST( MAX_SUM_LOG_GAIN_DB / 6.0, 7 ) - sum_log_gain_tmp_Q7 ) + + SILK_FIX_CONST( 7, 7 ) ) - gain_safety; + silk_VQ_WMat_EC( &temp_idx[ j ], /* O index of best codebook vector */ &rate_dist_Q14_subfr, /* O best weighted quantization error + mu * rate */ + &gain_Q7, /* O sum of absolute LTP coefficients */ b_Q14_ptr, /* I input vector to be quantized */ W_Q18_ptr, /* I weighting matrix */ cbk_ptr_Q7, /* I codebook */ + cbk_gain_ptr_Q7, /* I codebook effective gains */ cl_ptr_Q5, /* I code length for each codebook vector */ mu_Q9, /* I tradeoff between weighted error and rate */ + max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ cbk_size /* I number of vectors in codebook */ ); rate_dist_Q14 = silk_ADD_POS_SAT32( rate_dist_Q14, rate_dist_Q14_subfr ); + sum_log_gain_tmp_Q7 = silk_max(0, sum_log_gain_tmp_Q7 + + silk_lin2log( gain_safety + gain_Q7 ) - SILK_FIX_CONST( 7, 7 )); b_Q14_ptr += LTP_ORDER; W_Q18_ptr += LTP_ORDER * LTP_ORDER; @@ -89,6 +108,7 @@ void silk_quant_LTP_gains( min_rate_dist_Q14 = rate_dist_Q14; *periodicity_index = (opus_int8)k; silk_memcpy( cbk_index, temp_idx, nb_subfr * sizeof( opus_int8 ) ); + best_sum_log_gain_Q7 = sum_log_gain_tmp_Q7; } /* Break early in low-complexity mode if rate distortion is below threshold */ @@ -103,5 +123,6 @@ void silk_quant_LTP_gains( B_Q14[ j * LTP_ORDER + k ] = silk_LSHIFT( cbk_ptr_Q7[ cbk_index[ j ] * LTP_ORDER + k ], 7 ); } } + *sum_log_gain_Q7 = best_sum_log_gain_Q7; } diff --git a/code/opus-1.0.2/silk/resampler.c b/code/opus-1.1/silk/resampler.c similarity index 99% rename from code/opus-1.0.2/silk/resampler.c rename to code/opus-1.1/silk/resampler.c index 7e58332f..374fbb37 100644 --- a/code/opus-1.0.2/silk/resampler.c +++ b/code/opus-1.1/silk/resampler.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/resampler_down2.c b/code/opus-1.1/silk/resampler_down2.c similarity index 96% rename from code/opus-1.0.2/silk/resampler_down2.c rename to code/opus-1.1/silk/resampler_down2.c index 21d11992..cec36346 100644 --- a/code/opus-1.0.2/silk/resampler_down2.c +++ b/code/opus-1.1/silk/resampler_down2.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -35,8 +35,8 @@ POSSIBILITY OF SUCH DAMAGE. /* Downsample by a factor 2 */ void silk_resampler_down2( opus_int32 *S, /* I/O State vector [ 2 ] */ - opus_int16 *out, /* O Output signal [ len ] */ - const opus_int16 *in, /* I Input signal [ floor(len/2) ] */ + opus_int16 *out, /* O Output signal [ floor(len/2) ] */ + const opus_int16 *in, /* I Input signal [ len ] */ opus_int32 inLen /* I Number of input samples */ ) { diff --git a/code/opus-1.0.2/silk/resampler_down2_3.c b/code/opus-1.1/silk/resampler_down2_3.c similarity index 96% rename from code/opus-1.0.2/silk/resampler_down2_3.c rename to code/opus-1.1/silk/resampler_down2_3.c index fe5b671d..4342614d 100644 --- a/code/opus-1.0.2/silk/resampler_down2_3.c +++ b/code/opus-1.1/silk/resampler_down2_3.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -31,6 +31,7 @@ POSSIBILITY OF SUCH DAMAGE. #include "SigProc_FIX.h" #include "resampler_private.h" +#include "stack_alloc.h" #define ORDER_FIR 4 @@ -43,8 +44,11 @@ void silk_resampler_down2_3( ) { opus_int32 nSamplesIn, counter, res_Q6; - opus_int32 buf[ RESAMPLER_MAX_BATCH_SIZE_IN + ORDER_FIR ]; + VARDECL( opus_int32, buf ); opus_int32 *buf_ptr; + SAVE_STACK; + + ALLOC( buf, RESAMPLER_MAX_BATCH_SIZE_IN + ORDER_FIR, opus_int32 ); /* Copy buffered samples to start of buffer */ silk_memcpy( buf, S, ORDER_FIR * sizeof( opus_int32 ) ); @@ -95,4 +99,5 @@ void silk_resampler_down2_3( /* Copy last part of filtered signal to the state for the next call */ silk_memcpy( S, &buf[ nSamplesIn ], ORDER_FIR * sizeof( opus_int32 ) ); + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/resampler_private.h b/code/opus-1.1/silk/resampler_private.h similarity index 99% rename from code/opus-1.0.2/silk/resampler_private.h rename to code/opus-1.1/silk/resampler_private.h index 45d342c7..422a7d9d 100644 --- a/code/opus-1.0.2/silk/resampler_private.h +++ b/code/opus-1.1/silk/resampler_private.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/resampler_private_AR2.c b/code/opus-1.1/silk/resampler_private_AR2.c similarity index 99% rename from code/opus-1.0.2/silk/resampler_private_AR2.c rename to code/opus-1.1/silk/resampler_private_AR2.c index d069f2d8..5fff2371 100644 --- a/code/opus-1.0.2/silk/resampler_private_AR2.c +++ b/code/opus-1.1/silk/resampler_private_AR2.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/resampler_private_IIR_FIR.c b/code/opus-1.1/silk/resampler_private_IIR_FIR.c similarity index 90% rename from code/opus-1.0.2/silk/resampler_private_IIR_FIR.c rename to code/opus-1.1/silk/resampler_private_IIR_FIR.c index d9e42ca0..6b2b3a2e 100644 --- a/code/opus-1.0.2/silk/resampler_private_IIR_FIR.c +++ b/code/opus-1.1/silk/resampler_private_IIR_FIR.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -31,8 +31,9 @@ POSSIBILITY OF SUCH DAMAGE. #include "SigProc_FIX.h" #include "resampler_private.h" +#include "stack_alloc.h" -static inline opus_int16 *silk_resampler_private_IIR_FIR_INTERPOL( +static OPUS_INLINE opus_int16 *silk_resampler_private_IIR_FIR_INTERPOL( opus_int16 *out, opus_int16 *buf, opus_int32 max_index_Q16, @@ -71,10 +72,13 @@ void silk_resampler_private_IIR_FIR( silk_resampler_state_struct *S = (silk_resampler_state_struct *)SS; opus_int32 nSamplesIn; opus_int32 max_index_Q16, index_increment_Q16; - opus_int16 buf[ RESAMPLER_MAX_BATCH_SIZE_IN + RESAMPLER_ORDER_FIR_12 ]; + VARDECL( opus_int16, buf ); + SAVE_STACK; + + ALLOC( buf, 2 * S->batchSize + RESAMPLER_ORDER_FIR_12, opus_int16 ); /* Copy buffered samples to start of buffer */ - silk_memcpy( buf, S->sFIR, RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); + silk_memcpy( buf, S->sFIR.i16, RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); /* Iterate over blocks of frameSizeIn input samples */ index_increment_Q16 = S->invRatio_Q16; @@ -91,13 +95,13 @@ void silk_resampler_private_IIR_FIR( if( inLen > 0 ) { /* More iterations to do; copy last part of filtered signal to beginning of buffer */ - silk_memcpy( buf, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); + silk_memcpy( buf, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); } else { break; } } /* Copy last part of filtered signal to the state for the next call */ - silk_memcpy( S->sFIR, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); + silk_memcpy( S->sFIR.i16, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + RESTORE_STACK; } - diff --git a/code/opus-1.0.2/silk/resampler_private_down_FIR.c b/code/opus-1.1/silk/resampler_private_down_FIR.c similarity index 96% rename from code/opus-1.0.2/silk/resampler_private_down_FIR.c rename to code/opus-1.1/silk/resampler_private_down_FIR.c index 5d24564c..783e42b3 100644 --- a/code/opus-1.0.2/silk/resampler_private_down_FIR.c +++ b/code/opus-1.1/silk/resampler_private_down_FIR.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -31,8 +31,9 @@ POSSIBILITY OF SUCH DAMAGE. #include "SigProc_FIX.h" #include "resampler_private.h" +#include "stack_alloc.h" -static inline opus_int16 *silk_resampler_private_down_FIR_INTERPOL( +static OPUS_INLINE opus_int16 *silk_resampler_private_down_FIR_INTERPOL( opus_int16 *out, opus_int32 *buf, const opus_int16 *FIR_Coefs, @@ -151,11 +152,14 @@ void silk_resampler_private_down_FIR( silk_resampler_state_struct *S = (silk_resampler_state_struct *)SS; opus_int32 nSamplesIn; opus_int32 max_index_Q16, index_increment_Q16; - opus_int32 buf[ RESAMPLER_MAX_BATCH_SIZE_IN + SILK_RESAMPLER_MAX_FIR_ORDER ]; + VARDECL( opus_int32, buf ); const opus_int16 *FIR_Coefs; + SAVE_STACK; + + ALLOC( buf, S->batchSize + S->FIR_Order, opus_int32 ); /* Copy buffered samples to start of buffer */ - silk_memcpy( buf, S->sFIR, S->FIR_Order * sizeof( opus_int32 ) ); + silk_memcpy( buf, S->sFIR.i32, S->FIR_Order * sizeof( opus_int32 ) ); FIR_Coefs = &S->Coefs[ 2 ]; @@ -185,5 +189,6 @@ void silk_resampler_private_down_FIR( } /* Copy last part of filtered signal to the state for the next call */ - silk_memcpy( S->sFIR, &buf[ nSamplesIn ], S->FIR_Order * sizeof( opus_int32 ) ); + silk_memcpy( S->sFIR.i32, &buf[ nSamplesIn ], S->FIR_Order * sizeof( opus_int32 ) ); + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/resampler_private_up2_HQ.c b/code/opus-1.1/silk/resampler_private_up2_HQ.c similarity index 99% rename from code/opus-1.0.2/silk/resampler_private_up2_HQ.c rename to code/opus-1.1/silk/resampler_private_up2_HQ.c index 9e6dfc9e..c7ec8de3 100644 --- a/code/opus-1.0.2/silk/resampler_private_up2_HQ.c +++ b/code/opus-1.1/silk/resampler_private_up2_HQ.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/resampler_rom.c b/code/opus-1.1/silk/resampler_rom.c similarity index 99% rename from code/opus-1.0.2/silk/resampler_rom.c rename to code/opus-1.1/silk/resampler_rom.c index b50af2e2..2d502706 100644 --- a/code/opus-1.0.2/silk/resampler_rom.c +++ b/code/opus-1.1/silk/resampler_rom.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/resampler_rom.h b/code/opus-1.1/silk/resampler_rom.h similarity index 99% rename from code/opus-1.0.2/silk/resampler_rom.h rename to code/opus-1.1/silk/resampler_rom.h index 473b24a2..490b3388 100644 --- a/code/opus-1.0.2/silk/resampler_rom.h +++ b/code/opus-1.1/silk/resampler_rom.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/resampler_structs.h b/code/opus-1.1/silk/resampler_structs.h similarity index 93% rename from code/opus-1.0.2/silk/resampler_structs.h rename to code/opus-1.1/silk/resampler_structs.h index 4c28bd0a..9e9457d1 100644 --- a/code/opus-1.0.2/silk/resampler_structs.h +++ b/code/opus-1.1/silk/resampler_structs.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -37,7 +37,10 @@ extern "C" { typedef struct _silk_resampler_state_struct{ opus_int32 sIIR[ SILK_RESAMPLER_MAX_IIR_ORDER ]; /* this must be the first element of this struct */ - opus_int32 sFIR[ SILK_RESAMPLER_MAX_FIR_ORDER ]; + union{ + opus_int32 i32[ SILK_RESAMPLER_MAX_FIR_ORDER ]; + opus_int16 i16[ SILK_RESAMPLER_MAX_FIR_ORDER ]; + } sFIR; opus_int16 delayBuf[ 48 ]; opus_int resampler_function; opus_int batchSize; diff --git a/code/opus-1.0.2/silk/shell_coder.c b/code/opus-1.1/silk/shell_coder.c similarity index 98% rename from code/opus-1.0.2/silk/shell_coder.c rename to code/opus-1.1/silk/shell_coder.c index 32d00129..796f57d6 100644 --- a/code/opus-1.0.2/silk/shell_coder.c +++ b/code/opus-1.1/silk/shell_coder.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -33,7 +33,7 @@ POSSIBILITY OF SUCH DAMAGE. /* shell coder; pulse-subframe length is hardcoded */ -static inline void combine_pulses( +static OPUS_INLINE void combine_pulses( opus_int *out, /* O combined pulses vector [len] */ const opus_int *in, /* I input vector [2 * len] */ const opus_int len /* I number of OUTPUT samples */ @@ -45,7 +45,7 @@ static inline void combine_pulses( } } -static inline void encode_split( +static OPUS_INLINE void encode_split( ec_enc *psRangeEnc, /* I/O compressor data structure */ const opus_int p_child1, /* I pulse amplitude of first child subframe */ const opus_int p, /* I pulse amplitude of current subframe */ @@ -57,7 +57,7 @@ static inline void encode_split( } } -static inline void decode_split( +static OPUS_INLINE void decode_split( opus_int *p_child1, /* O pulse amplitude of first child subframe */ opus_int *p_child2, /* O pulse amplitude of second child subframe */ ec_dec *psRangeDec, /* I/O Compressor data structure */ diff --git a/code/opus-1.0.2/silk/sigm_Q15.c b/code/opus-1.1/silk/sigm_Q15.c similarity index 99% rename from code/opus-1.0.2/silk/sigm_Q15.c rename to code/opus-1.1/silk/sigm_Q15.c index cf5af6bc..3c507d25 100644 --- a/code/opus-1.0.2/silk/sigm_Q15.c +++ b/code/opus-1.1/silk/sigm_Q15.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/sort.c b/code/opus-1.1/silk/sort.c similarity index 99% rename from code/opus-1.0.2/silk/sort.c rename to code/opus-1.1/silk/sort.c index a4072ec4..8670dbdd 100644 --- a/code/opus-1.0.2/silk/sort.c +++ b/code/opus-1.1/silk/sort.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/stereo_LR_to_MS.c b/code/opus-1.1/silk/stereo_LR_to_MS.c similarity index 96% rename from code/opus-1.0.2/silk/stereo_LR_to_MS.c rename to code/opus-1.1/silk/stereo_LR_to_MS.c index 6a680e09..42906e6f 100644 --- a/code/opus-1.0.2/silk/stereo_LR_to_MS.c +++ b/code/opus-1.1/silk/stereo_LR_to_MS.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -30,6 +30,7 @@ POSSIBILITY OF SUCH DAMAGE. #endif #include "main.h" +#include "stack_alloc.h" /* Convert Left/Right stereo signal to adaptive Mid/Side representation */ void silk_stereo_LR_to_MS( @@ -49,11 +50,15 @@ void silk_stereo_LR_to_MS( opus_int n, is10msFrame, denom_Q16, delta0_Q13, delta1_Q13; opus_int32 sum, diff, smooth_coef_Q16, pred_Q13[ 2 ], pred0_Q13, pred1_Q13; opus_int32 LP_ratio_Q14, HP_ratio_Q14, frac_Q16, frac_3_Q16, min_mid_rate_bps, width_Q14, w_Q24, deltaw_Q24; - opus_int16 side[ MAX_FRAME_LENGTH + 2 ]; - opus_int16 LP_mid[ MAX_FRAME_LENGTH ], HP_mid[ MAX_FRAME_LENGTH ]; - opus_int16 LP_side[ MAX_FRAME_LENGTH ], HP_side[ MAX_FRAME_LENGTH ]; + VARDECL( opus_int16, side ); + VARDECL( opus_int16, LP_mid ); + VARDECL( opus_int16, HP_mid ); + VARDECL( opus_int16, LP_side ); + VARDECL( opus_int16, HP_side ); opus_int16 *mid = &x1[ -2 ]; + SAVE_STACK; + ALLOC( side, frame_length + 2, opus_int16 ); /* Convert to basic mid/side signals */ for( n = 0; n < frame_length + 2; n++ ) { sum = x1[ n - 2 ] + (opus_int32)x2[ n - 2 ]; @@ -69,6 +74,8 @@ void silk_stereo_LR_to_MS( silk_memcpy( state->sSide, &side[ frame_length ], 2 * sizeof( opus_int16 ) ); /* LP and HP filter mid signal */ + ALLOC( LP_mid, frame_length, opus_int16 ); + ALLOC( HP_mid, frame_length, opus_int16 ); for( n = 0; n < frame_length; n++ ) { sum = silk_RSHIFT_ROUND( silk_ADD_LSHIFT( mid[ n ] + mid[ n + 2 ], mid[ n + 1 ], 1 ), 2 ); LP_mid[ n ] = sum; @@ -76,6 +83,8 @@ void silk_stereo_LR_to_MS( } /* LP and HP filter side signal */ + ALLOC( LP_side, frame_length, opus_int16 ); + ALLOC( HP_side, frame_length, opus_int16 ); for( n = 0; n < frame_length; n++ ) { sum = silk_RSHIFT_ROUND( silk_ADD_LSHIFT( side[ n ] + side[ n + 2 ], side[ n + 1 ], 1 ), 2 ); LP_side[ n ] = sum; @@ -216,4 +225,5 @@ void silk_stereo_LR_to_MS( state->pred_prev_Q13[ 0 ] = (opus_int16)pred_Q13[ 0 ]; state->pred_prev_Q13[ 1 ] = (opus_int16)pred_Q13[ 1 ]; state->width_prev_Q14 = (opus_int16)width_Q14; + RESTORE_STACK; } diff --git a/code/opus-1.0.2/silk/stereo_MS_to_LR.c b/code/opus-1.1/silk/stereo_MS_to_LR.c similarity index 99% rename from code/opus-1.0.2/silk/stereo_MS_to_LR.c rename to code/opus-1.1/silk/stereo_MS_to_LR.c index 23515870..62521a4f 100644 --- a/code/opus-1.0.2/silk/stereo_MS_to_LR.c +++ b/code/opus-1.1/silk/stereo_MS_to_LR.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/stereo_decode_pred.c b/code/opus-1.1/silk/stereo_decode_pred.c similarity index 99% rename from code/opus-1.0.2/silk/stereo_decode_pred.c rename to code/opus-1.1/silk/stereo_decode_pred.c index 026aa7a0..56ba3925 100644 --- a/code/opus-1.0.2/silk/stereo_decode_pred.c +++ b/code/opus-1.1/silk/stereo_decode_pred.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/stereo_encode_pred.c b/code/opus-1.1/silk/stereo_encode_pred.c similarity index 99% rename from code/opus-1.0.2/silk/stereo_encode_pred.c rename to code/opus-1.1/silk/stereo_encode_pred.c index 3cffd367..e6dd1950 100644 --- a/code/opus-1.0.2/silk/stereo_encode_pred.c +++ b/code/opus-1.1/silk/stereo_encode_pred.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/stereo_find_predictor.c b/code/opus-1.1/silk/stereo_find_predictor.c similarity index 99% rename from code/opus-1.0.2/silk/stereo_find_predictor.c rename to code/opus-1.1/silk/stereo_find_predictor.c index aec58dab..e30e90bd 100644 --- a/code/opus-1.0.2/silk/stereo_find_predictor.c +++ b/code/opus-1.1/silk/stereo_find_predictor.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/stereo_quant_pred.c b/code/opus-1.1/silk/stereo_quant_pred.c similarity index 99% rename from code/opus-1.0.2/silk/stereo_quant_pred.c rename to code/opus-1.1/silk/stereo_quant_pred.c index df97c9f6..d4ced6c3 100644 --- a/code/opus-1.0.2/silk/stereo_quant_pred.c +++ b/code/opus-1.1/silk/stereo_quant_pred.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/structs.h b/code/opus-1.1/silk/structs.h similarity index 99% rename from code/opus-1.0.2/silk/structs.h rename to code/opus-1.1/silk/structs.h index 5d37f660..1826b36a 100644 --- a/code/opus-1.0.2/silk/structs.h +++ b/code/opus-1.1/silk/structs.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -171,6 +171,7 @@ typedef struct { opus_int32 pitchEstimationThreshold_Q16; /* Threshold for pitch estimator */ opus_int LTPQuantLowComplexity; /* Flag for low complexity LTP quantization */ opus_int mu_LTP_Q9; /* Rate-distortion tradeoff in LTP quantization */ + opus_int32 sum_log_gain_Q7; /* Cumulative max prediction gain */ opus_int NLSF_MSVQ_Survivors; /* Number of survivors in NLSF MSVQ */ opus_int first_frame_after_reset; /* Flag for deactivating NLSF interpolation, pitch prediction */ opus_int controlled_since_last_payload; /* Flag for ensuring codec_control only runs once per packet */ @@ -191,6 +192,8 @@ typedef struct { SideInfoIndices indices; opus_int8 pulses[ MAX_FRAME_LENGTH ]; + int arch; + /* Input/output buffering */ opus_int16 inputBuf[ MAX_FRAME_LENGTH + 2 ]; /* Buffer containing input signal */ opus_int inputBufIx; diff --git a/code/opus-1.0.2/silk/sum_sqr_shift.c b/code/opus-1.1/silk/sum_sqr_shift.c similarity index 99% rename from code/opus-1.0.2/silk/sum_sqr_shift.c rename to code/opus-1.1/silk/sum_sqr_shift.c index 2eaf77b6..12514c99 100644 --- a/code/opus-1.0.2/silk/sum_sqr_shift.c +++ b/code/opus-1.1/silk/sum_sqr_shift.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/table_LSF_cos.c b/code/opus-1.1/silk/table_LSF_cos.c similarity index 99% rename from code/opus-1.0.2/silk/table_LSF_cos.c rename to code/opus-1.1/silk/table_LSF_cos.c index 710537fb..ec9dc639 100644 --- a/code/opus-1.0.2/silk/table_LSF_cos.c +++ b/code/opus-1.1/silk/table_LSF_cos.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/tables.h b/code/opus-1.1/silk/tables.h similarity index 98% rename from code/opus-1.0.2/silk/tables.h rename to code/opus-1.1/silk/tables.h index 072b7929..a91431e8 100644 --- a/code/opus-1.0.2/silk/tables.h +++ b/code/opus-1.1/silk/tables.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -78,6 +78,8 @@ extern const opus_uint8 * const silk_LTP_gain_iCDF_ptrs[ NB_LTP_CBKS ]; extern const opus_uint8 * const silk_LTP_gain_BITS_Q5_ptrs[ NB_LTP_CBKS ]; /* 3 */ extern const opus_int16 silk_LTP_gain_middle_avg_RD_Q14; extern const opus_int8 * const silk_LTP_vq_ptrs_Q7[ NB_LTP_CBKS ]; /* 168 */ +extern const opus_uint8 * const silk_LTP_vq_gain_ptrs_Q7[NB_LTP_CBKS]; + extern const opus_int8 silk_LTP_vq_sizes[ NB_LTP_CBKS ]; /* 3 */ extern const opus_uint8 silk_LTPscale_iCDF[ 3 ]; /* 4 */ diff --git a/code/opus-1.0.2/silk/tables_LTP.c b/code/opus-1.1/silk/tables_LTP.c similarity index 87% rename from code/opus-1.0.2/silk/tables_LTP.c rename to code/opus-1.1/silk/tables_LTP.c index dd1fb556..0e6a0254 100644 --- a/code/opus-1.0.2/silk/tables_LTP.c +++ b/code/opus-1.1/silk/tables_LTP.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -267,6 +267,30 @@ const opus_int8 * const silk_LTP_vq_ptrs_Q7[NB_LTP_CBKS] = { (opus_int8 *)&silk_LTP_gain_vq_2[0][0] }; +/* Maximum frequency-dependent response of the pitch taps above, + computed as max(abs(freqz(taps))) */ +static const opus_uint8 silk_LTP_gain_vq_0_gain[8] = { + 46, 2, 90, 87, 93, 91, 82, 98 +}; + +static const opus_uint8 silk_LTP_gain_vq_1_gain[16] = { + 109, 120, 118, 12, 113, 115, 117, 119, + 99, 59, 87, 111, 63, 111, 112, 80 +}; + +static const opus_uint8 silk_LTP_gain_vq_2_gain[32] = { + 126, 124, 125, 124, 129, 121, 126, 23, + 132, 127, 127, 127, 126, 127, 122, 133, + 130, 134, 101, 118, 119, 145, 126, 86, + 124, 120, 123, 119, 170, 173, 107, 109 +}; + +const opus_uint8 * const silk_LTP_vq_gain_ptrs_Q7[NB_LTP_CBKS] = { + &silk_LTP_gain_vq_0_gain[0], + &silk_LTP_gain_vq_1_gain[0], + &silk_LTP_gain_vq_2_gain[0] +}; + const opus_int8 silk_LTP_vq_sizes[NB_LTP_CBKS] = { 8, 16, 32 }; diff --git a/code/opus-1.0.2/silk/tables_NLSF_CB_NB_MB.c b/code/opus-1.1/silk/tables_NLSF_CB_NB_MB.c similarity index 99% rename from code/opus-1.0.2/silk/tables_NLSF_CB_NB_MB.c rename to code/opus-1.1/silk/tables_NLSF_CB_NB_MB.c index 75480526..8c59d207 100644 --- a/code/opus-1.0.2/silk/tables_NLSF_CB_NB_MB.c +++ b/code/opus-1.1/silk/tables_NLSF_CB_NB_MB.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/tables_NLSF_CB_WB.c b/code/opus-1.1/silk/tables_NLSF_CB_WB.c similarity index 99% rename from code/opus-1.0.2/silk/tables_NLSF_CB_WB.c rename to code/opus-1.1/silk/tables_NLSF_CB_WB.c index 3d6052e4..50af87eb 100644 --- a/code/opus-1.0.2/silk/tables_NLSF_CB_WB.c +++ b/code/opus-1.1/silk/tables_NLSF_CB_WB.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/tables_gain.c b/code/opus-1.1/silk/tables_gain.c similarity index 99% rename from code/opus-1.0.2/silk/tables_gain.c rename to code/opus-1.1/silk/tables_gain.c index fccef821..37e41d89 100644 --- a/code/opus-1.0.2/silk/tables_gain.c +++ b/code/opus-1.1/silk/tables_gain.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/tables_other.c b/code/opus-1.1/silk/tables_other.c similarity index 99% rename from code/opus-1.0.2/silk/tables_other.c rename to code/opus-1.1/silk/tables_other.c index 3dc68d47..398686bf 100644 --- a/code/opus-1.0.2/silk/tables_other.c +++ b/code/opus-1.1/silk/tables_other.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/tables_pitch_lag.c b/code/opus-1.1/silk/tables_pitch_lag.c similarity index 99% rename from code/opus-1.0.2/silk/tables_pitch_lag.c rename to code/opus-1.1/silk/tables_pitch_lag.c index 819b0ab3..e80cc59a 100644 --- a/code/opus-1.0.2/silk/tables_pitch_lag.c +++ b/code/opus-1.1/silk/tables_pitch_lag.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/tables_pulses_per_block.c b/code/opus-1.1/silk/tables_pulses_per_block.c similarity index 99% rename from code/opus-1.0.2/silk/tables_pulses_per_block.c rename to code/opus-1.1/silk/tables_pulses_per_block.c index 521e6ff6..c7c01c88 100644 --- a/code/opus-1.0.2/silk/tables_pulses_per_block.c +++ b/code/opus-1.1/silk/tables_pulses_per_block.c @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE diff --git a/code/opus-1.0.2/silk/tuning_parameters.h b/code/opus-1.1/silk/tuning_parameters.h similarity index 98% rename from code/opus-1.0.2/silk/tuning_parameters.h rename to code/opus-1.1/silk/tuning_parameters.h index a26de4d2..e1057bba 100644 --- a/code/opus-1.0.2/silk/tuning_parameters.h +++ b/code/opus-1.1/silk/tuning_parameters.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -50,7 +50,7 @@ extern "C" /* Linear prediction */ /*********************/ -/* LPC analysis defines: regularization and bandwidth expansion */ +/* LPC analysis regularization */ #define FIND_LPC_COND_FAC 1e-5f /* LTP analysis defines */ @@ -63,6 +63,9 @@ extern "C" #define MU_LTP_QUANT_MB 0.025f #define MU_LTP_QUANT_WB 0.02f +/* Max cumulative LTP gain */ +#define MAX_SUM_LOG_GAIN_DB 250.0f + /***********************/ /* High pass filtering */ /***********************/ diff --git a/code/opus-1.0.2/silk/typedef.h b/code/opus-1.1/silk/typedef.h similarity index 96% rename from code/opus-1.0.2/silk/typedef.h rename to code/opus-1.1/silk/typedef.h index da981237..97b7e709 100644 --- a/code/opus-1.0.2/silk/typedef.h +++ b/code/opus-1.1/silk/typedef.h @@ -8,11 +8,11 @@ this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. -- Neither the name of Internet Society, IETF or IETF Trust, nor the +- Neither the name of Internet Society, IETF or IETF Trust, nor the names of specific contributors, may be used to endorse or promote products derived from this software without specific prior written permission. -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE @@ -29,6 +29,7 @@ POSSIBILITY OF SUCH DAMAGE. #define SILK_TYPEDEF_H #include "opus_types.h" +#include "opus_defines.h" #ifndef FIXED_POINT # include @@ -63,7 +64,7 @@ POSSIBILITY OF SUCH DAMAGE. #ifdef __GNUC__ __attribute__((noreturn)) #endif -static inline void _silk_fatal(const char *str, const char *file, int line) +static OPUS_INLINE void _silk_fatal(const char *str, const char *file, int line) { fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str); abort(); diff --git a/code/opus-1.1/src/analysis.c b/code/opus-1.1/src/analysis.c new file mode 100644 index 00000000..778a62aa --- /dev/null +++ b/code/opus-1.1/src/analysis.c @@ -0,0 +1,645 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "kiss_fft.h" +#include "celt.h" +#include "modes.h" +#include "arch.h" +#include "quant_bands.h" +#include +#include "analysis.h" +#include "mlp.h" +#include "stack_alloc.h" + +extern const MLP net; + +#ifndef M_PI +#define M_PI 3.141592653 +#endif + +static const float dct_table[128] = { + 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, + 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, + 0.351851f, 0.338330f, 0.311806f, 0.273300f, 0.224292f, 0.166664f, 0.102631f, 0.034654f, + -0.034654f,-0.102631f,-0.166664f,-0.224292f,-0.273300f,-0.311806f,-0.338330f,-0.351851f, + 0.346760f, 0.293969f, 0.196424f, 0.068975f,-0.068975f,-0.196424f,-0.293969f,-0.346760f, + -0.346760f,-0.293969f,-0.196424f,-0.068975f, 0.068975f, 0.196424f, 0.293969f, 0.346760f, + 0.338330f, 0.224292f, 0.034654f,-0.166664f,-0.311806f,-0.351851f,-0.273300f,-0.102631f, + 0.102631f, 0.273300f, 0.351851f, 0.311806f, 0.166664f,-0.034654f,-0.224292f,-0.338330f, + 0.326641f, 0.135299f,-0.135299f,-0.326641f,-0.326641f,-0.135299f, 0.135299f, 0.326641f, + 0.326641f, 0.135299f,-0.135299f,-0.326641f,-0.326641f,-0.135299f, 0.135299f, 0.326641f, + 0.311806f, 0.034654f,-0.273300f,-0.338330f,-0.102631f, 0.224292f, 0.351851f, 0.166664f, + -0.166664f,-0.351851f,-0.224292f, 0.102631f, 0.338330f, 0.273300f,-0.034654f,-0.311806f, + 0.293969f,-0.068975f,-0.346760f,-0.196424f, 0.196424f, 0.346760f, 0.068975f,-0.293969f, + -0.293969f, 0.068975f, 0.346760f, 0.196424f,-0.196424f,-0.346760f,-0.068975f, 0.293969f, + 0.273300f,-0.166664f,-0.338330f, 0.034654f, 0.351851f, 0.102631f,-0.311806f,-0.224292f, + 0.224292f, 0.311806f,-0.102631f,-0.351851f,-0.034654f, 0.338330f, 0.166664f,-0.273300f, +}; + +static const float analysis_window[240] = { + 0.000043f, 0.000171f, 0.000385f, 0.000685f, 0.001071f, 0.001541f, 0.002098f, 0.002739f, + 0.003466f, 0.004278f, 0.005174f, 0.006156f, 0.007222f, 0.008373f, 0.009607f, 0.010926f, + 0.012329f, 0.013815f, 0.015385f, 0.017037f, 0.018772f, 0.020590f, 0.022490f, 0.024472f, + 0.026535f, 0.028679f, 0.030904f, 0.033210f, 0.035595f, 0.038060f, 0.040604f, 0.043227f, + 0.045928f, 0.048707f, 0.051564f, 0.054497f, 0.057506f, 0.060591f, 0.063752f, 0.066987f, + 0.070297f, 0.073680f, 0.077136f, 0.080665f, 0.084265f, 0.087937f, 0.091679f, 0.095492f, + 0.099373f, 0.103323f, 0.107342f, 0.111427f, 0.115579f, 0.119797f, 0.124080f, 0.128428f, + 0.132839f, 0.137313f, 0.141849f, 0.146447f, 0.151105f, 0.155823f, 0.160600f, 0.165435f, + 0.170327f, 0.175276f, 0.180280f, 0.185340f, 0.190453f, 0.195619f, 0.200838f, 0.206107f, + 0.211427f, 0.216797f, 0.222215f, 0.227680f, 0.233193f, 0.238751f, 0.244353f, 0.250000f, + 0.255689f, 0.261421f, 0.267193f, 0.273005f, 0.278856f, 0.284744f, 0.290670f, 0.296632f, + 0.302628f, 0.308658f, 0.314721f, 0.320816f, 0.326941f, 0.333097f, 0.339280f, 0.345492f, + 0.351729f, 0.357992f, 0.364280f, 0.370590f, 0.376923f, 0.383277f, 0.389651f, 0.396044f, + 0.402455f, 0.408882f, 0.415325f, 0.421783f, 0.428254f, 0.434737f, 0.441231f, 0.447736f, + 0.454249f, 0.460770f, 0.467298f, 0.473832f, 0.480370f, 0.486912f, 0.493455f, 0.500000f, + 0.506545f, 0.513088f, 0.519630f, 0.526168f, 0.532702f, 0.539230f, 0.545751f, 0.552264f, + 0.558769f, 0.565263f, 0.571746f, 0.578217f, 0.584675f, 0.591118f, 0.597545f, 0.603956f, + 0.610349f, 0.616723f, 0.623077f, 0.629410f, 0.635720f, 0.642008f, 0.648271f, 0.654508f, + 0.660720f, 0.666903f, 0.673059f, 0.679184f, 0.685279f, 0.691342f, 0.697372f, 0.703368f, + 0.709330f, 0.715256f, 0.721144f, 0.726995f, 0.732807f, 0.738579f, 0.744311f, 0.750000f, + 0.755647f, 0.761249f, 0.766807f, 0.772320f, 0.777785f, 0.783203f, 0.788573f, 0.793893f, + 0.799162f, 0.804381f, 0.809547f, 0.814660f, 0.819720f, 0.824724f, 0.829673f, 0.834565f, + 0.839400f, 0.844177f, 0.848895f, 0.853553f, 0.858151f, 0.862687f, 0.867161f, 0.871572f, + 0.875920f, 0.880203f, 0.884421f, 0.888573f, 0.892658f, 0.896677f, 0.900627f, 0.904508f, + 0.908321f, 0.912063f, 0.915735f, 0.919335f, 0.922864f, 0.926320f, 0.929703f, 0.933013f, + 0.936248f, 0.939409f, 0.942494f, 0.945503f, 0.948436f, 0.951293f, 0.954072f, 0.956773f, + 0.959396f, 0.961940f, 0.964405f, 0.966790f, 0.969096f, 0.971321f, 0.973465f, 0.975528f, + 0.977510f, 0.979410f, 0.981228f, 0.982963f, 0.984615f, 0.986185f, 0.987671f, 0.989074f, + 0.990393f, 0.991627f, 0.992778f, 0.993844f, 0.994826f, 0.995722f, 0.996534f, 0.997261f, + 0.997902f, 0.998459f, 0.998929f, 0.999315f, 0.999615f, 0.999829f, 0.999957f, 1.000000f, +}; + +static const int tbands[NB_TBANDS+1] = { + 2, 4, 6, 8, 10, 12, 14, 16, 20, 24, 28, 32, 40, 48, 56, 68, 80, 96, 120 +}; + +static const int extra_bands[NB_TOT_BANDS+1] = { + 1, 2, 4, 6, 8, 10, 12, 14, 16, 20, 24, 28, 32, 40, 48, 56, 68, 80, 96, 120, 160, 200 +}; + +/*static const float tweight[NB_TBANDS+1] = { + .3, .4, .5, .6, .7, .8, .9, 1., 1., 1., 1., 1., 1., 1., .8, .7, .6, .5 +};*/ + +#define NB_TONAL_SKIP_BANDS 9 + +#define cA 0.43157974f +#define cB 0.67848403f +#define cC 0.08595542f +#define cE ((float)M_PI/2) +static OPUS_INLINE float fast_atan2f(float y, float x) { + float x2, y2; + /* Should avoid underflow on the values we'll get */ + if (ABS16(x)+ABS16(y)<1e-9f) + { + x*=1e12f; + y*=1e12f; + } + x2 = x*x; + y2 = y*y; + if(x2read_pos; + curr_lookahead = tonal->write_pos-tonal->read_pos; + if (curr_lookahead<0) + curr_lookahead += DETECT_SIZE; + + if (len > 480 && pos != tonal->write_pos) + { + pos++; + if (pos==DETECT_SIZE) + pos=0; + } + if (pos == tonal->write_pos) + pos--; + if (pos<0) + pos = DETECT_SIZE-1; + OPUS_COPY(info_out, &tonal->info[pos], 1); + tonal->read_subframe += len/120; + while (tonal->read_subframe>=4) + { + tonal->read_subframe -= 4; + tonal->read_pos++; + } + if (tonal->read_pos>=DETECT_SIZE) + tonal->read_pos-=DETECT_SIZE; + + /* Compensate for the delay in the features themselves. + FIXME: Need a better estimate the 10 I just made up */ + curr_lookahead = IMAX(curr_lookahead-10, 0); + + psum=0; + /* Summing the probability of transition patterns that involve music at + time (DETECT_SIZE-curr_lookahead-1) */ + for (i=0;ipmusic[i]; + for (;ipspeech[i]; + psum = psum*tonal->music_confidence + (1-psum)*tonal->speech_confidence; + /*printf("%f %f %f\n", psum, info_out->music_prob, info_out->tonality);*/ + + info_out->music_prob = psum; +} + +void tonality_analysis(TonalityAnalysisState *tonal, AnalysisInfo *info_out, const CELTMode *celt_mode, const void *x, int len, int offset, int c1, int c2, int C, int lsb_depth, downmix_func downmix) +{ + int i, b; + const kiss_fft_state *kfft; + VARDECL(kiss_fft_cpx, in); + VARDECL(kiss_fft_cpx, out); + int N = 480, N2=240; + float * OPUS_RESTRICT A = tonal->angle; + float * OPUS_RESTRICT dA = tonal->d_angle; + float * OPUS_RESTRICT d2A = tonal->d2_angle; + VARDECL(float, tonality); + VARDECL(float, noisiness); + float band_tonality[NB_TBANDS]; + float logE[NB_TBANDS]; + float BFCC[8]; + float features[25]; + float frame_tonality; + float max_frame_tonality; + /*float tw_sum=0;*/ + float frame_noisiness; + const float pi4 = (float)(M_PI*M_PI*M_PI*M_PI); + float slope=0; + float frame_stationarity; + float relativeE; + float frame_probs[2]; + float alpha, alphaE, alphaE2; + float frame_loudness; + float bandwidth_mask; + int bandwidth=0; + float maxE = 0; + float noise_floor; + int remaining; + AnalysisInfo *info; + SAVE_STACK; + + tonal->last_transition++; + alpha = 1.f/IMIN(20, 1+tonal->count); + alphaE = 1.f/IMIN(50, 1+tonal->count); + alphaE2 = 1.f/IMIN(1000, 1+tonal->count); + + if (tonal->count<4) + tonal->music_prob = .5; + kfft = celt_mode->mdct.kfft[0]; + if (tonal->count==0) + tonal->mem_fill = 240; + downmix(x, &tonal->inmem[tonal->mem_fill], IMIN(len, ANALYSIS_BUF_SIZE-tonal->mem_fill), offset, c1, c2, C); + if (tonal->mem_fill+len < ANALYSIS_BUF_SIZE) + { + tonal->mem_fill += len; + /* Don't have enough to update the analysis */ + RESTORE_STACK; + return; + } + info = &tonal->info[tonal->write_pos++]; + if (tonal->write_pos>=DETECT_SIZE) + tonal->write_pos-=DETECT_SIZE; + + ALLOC(in, 480, kiss_fft_cpx); + ALLOC(out, 480, kiss_fft_cpx); + ALLOC(tonality, 240, float); + ALLOC(noisiness, 240, float); + for (i=0;iinmem[i]); + in[i].i = (kiss_fft_scalar)(w*tonal->inmem[N2+i]); + in[N-i-1].r = (kiss_fft_scalar)(w*tonal->inmem[N-i-1]); + in[N-i-1].i = (kiss_fft_scalar)(w*tonal->inmem[N+N2-i-1]); + } + OPUS_MOVE(tonal->inmem, tonal->inmem+ANALYSIS_BUF_SIZE-240, 240); + remaining = len - (ANALYSIS_BUF_SIZE-tonal->mem_fill); + downmix(x, &tonal->inmem[240], remaining, offset+ANALYSIS_BUF_SIZE-tonal->mem_fill, c1, c2, C); + tonal->mem_fill = 240 + remaining; + opus_fft(kfft, in, out); + + for (i=1;iactivity = 0; + frame_noisiness = 0; + frame_stationarity = 0; + if (!tonal->count) + { + for (b=0;blowE[b] = 1e10; + tonal->highE[b] = -1e10; + } + } + relativeE = 0; + frame_loudness = 0; + for (b=0;bE[tonal->E_count][b] = E; + frame_noisiness += nE/(1e-15f+E); + + frame_loudness += (float)sqrt(E+1e-10f); + logE[b] = (float)log(E+1e-10f); + tonal->lowE[b] = MIN32(logE[b], tonal->lowE[b]+.01f); + tonal->highE[b] = MAX32(logE[b], tonal->highE[b]-.1f); + if (tonal->highE[b] < tonal->lowE[b]+1.f) + { + tonal->highE[b]+=.5f; + tonal->lowE[b]-=.5f; + } + relativeE += (logE[b]-tonal->lowE[b])/(1e-15f+tonal->highE[b]-tonal->lowE[b]); + + L1=L2=0; + for (i=0;iE[i][b]); + L2 += tonal->E[i][b]; + } + + stationarity = MIN16(0.99f,L1/(float)sqrt(1e-15+NB_FRAMES*L2)); + stationarity *= stationarity; + stationarity *= stationarity; + frame_stationarity += stationarity; + /*band_tonality[b] = tE/(1e-15+E)*/; + band_tonality[b] = MAX16(tE/(1e-15f+E), stationarity*tonal->prev_band_tonality[b]); +#if 0 + if (b>=NB_TONAL_SKIP_BANDS) + { + frame_tonality += tweight[b]*band_tonality[b]; + tw_sum += tweight[b]; + } +#else + frame_tonality += band_tonality[b]; + if (b>=NB_TBANDS-NB_TONAL_SKIP_BANDS) + frame_tonality -= band_tonality[b-NB_TBANDS+NB_TONAL_SKIP_BANDS]; +#endif + max_frame_tonality = MAX16(max_frame_tonality, (1.f+.03f*(b-NB_TBANDS))*frame_tonality); + slope += band_tonality[b]*(b-8); + /*printf("%f %f ", band_tonality[b], stationarity);*/ + tonal->prev_band_tonality[b] = band_tonality[b]; + } + + bandwidth_mask = 0; + bandwidth = 0; + maxE = 0; + noise_floor = 5.7e-4f/(1<<(IMAX(0,lsb_depth-8))); +#ifdef FIXED_POINT + noise_floor *= 1<<(15+SIG_SHIFT); +#endif + noise_floor *= noise_floor; + for (b=0;bmeanE[b] = MAX32((1-alphaE2)*tonal->meanE[b], E); + E = MAX32(E, tonal->meanE[b]); + /* Use a simple follower with 13 dB/Bark slope for spreading function */ + bandwidth_mask = MAX32(.05f*bandwidth_mask, E); + /* Consider the band "active" only if all these conditions are met: + 1) less than 10 dB below the simple follower + 2) less than 90 dB below the peak band (maximal masking possible considering + both the ATH and the loudness-dependent slope of the spreading function) + 3) above the PCM quantization noise floor + */ + if (E>.1*bandwidth_mask && E*1e9f > maxE && E > noise_floor*(band_end-band_start)) + bandwidth = b; + } + if (tonal->count<=2) + bandwidth = 20; + frame_loudness = 20*(float)log10(frame_loudness); + tonal->Etracker = MAX32(tonal->Etracker-.03f, frame_loudness); + tonal->lowECount *= (1-alphaE); + if (frame_loudness < tonal->Etracker-30) + tonal->lowECount += alphaE; + + for (i=0;i<8;i++) + { + float sum=0; + for (b=0;b<16;b++) + sum += dct_table[i*16+b]*logE[b]; + BFCC[i] = sum; + } + + frame_stationarity /= NB_TBANDS; + relativeE /= NB_TBANDS; + if (tonal->count<10) + relativeE = .5; + frame_noisiness /= NB_TBANDS; +#if 1 + info->activity = frame_noisiness + (1-frame_noisiness)*relativeE; +#else + info->activity = .5*(1+frame_noisiness-frame_stationarity); +#endif + frame_tonality = (max_frame_tonality/(NB_TBANDS-NB_TONAL_SKIP_BANDS)); + frame_tonality = MAX16(frame_tonality, tonal->prev_tonality*.8f); + tonal->prev_tonality = frame_tonality; + + slope /= 8*8; + info->tonality_slope = slope; + + tonal->E_count = (tonal->E_count+1)%NB_FRAMES; + tonal->count++; + info->tonality = frame_tonality; + + for (i=0;i<4;i++) + features[i] = -0.12299f*(BFCC[i]+tonal->mem[i+24]) + 0.49195f*(tonal->mem[i]+tonal->mem[i+16]) + 0.69693f*tonal->mem[i+8] - 1.4349f*tonal->cmean[i]; + + for (i=0;i<4;i++) + tonal->cmean[i] = (1-alpha)*tonal->cmean[i] + alpha*BFCC[i]; + + for (i=0;i<4;i++) + features[4+i] = 0.63246f*(BFCC[i]-tonal->mem[i+24]) + 0.31623f*(tonal->mem[i]-tonal->mem[i+16]); + for (i=0;i<3;i++) + features[8+i] = 0.53452f*(BFCC[i]+tonal->mem[i+24]) - 0.26726f*(tonal->mem[i]+tonal->mem[i+16]) -0.53452f*tonal->mem[i+8]; + + if (tonal->count > 5) + { + for (i=0;i<9;i++) + tonal->std[i] = (1-alpha)*tonal->std[i] + alpha*features[i]*features[i]; + } + + for (i=0;i<8;i++) + { + tonal->mem[i+24] = tonal->mem[i+16]; + tonal->mem[i+16] = tonal->mem[i+8]; + tonal->mem[i+8] = tonal->mem[i]; + tonal->mem[i] = BFCC[i]; + } + for (i=0;i<9;i++) + features[11+i] = (float)sqrt(tonal->std[i]); + features[20] = info->tonality; + features[21] = info->activity; + features[22] = frame_stationarity; + features[23] = info->tonality_slope; + features[24] = tonal->lowECount; + +#ifndef DISABLE_FLOAT_API + mlp_process(&net, features, frame_probs); + frame_probs[0] = .5f*(frame_probs[0]+1); + /* Curve fitting between the MLP probability and the actual probability */ + frame_probs[0] = .01f + 1.21f*frame_probs[0]*frame_probs[0] - .23f*(float)pow(frame_probs[0], 10); + /* Probability of active audio (as opposed to silence) */ + frame_probs[1] = .5f*frame_probs[1]+.5f; + /* Consider that silence has a 50-50 probability. */ + frame_probs[0] = frame_probs[1]*frame_probs[0] + (1-frame_probs[1])*.5f; + + /*printf("%f %f ", frame_probs[0], frame_probs[1]);*/ + { + /* Probability of state transition */ + float tau; + /* Represents independence of the MLP probabilities, where + beta=1 means fully independent. */ + float beta; + /* Denormalized probability of speech (p0) and music (p1) after update */ + float p0, p1; + /* Probabilities for "all speech" and "all music" */ + float s0, m0; + /* Probability sum for renormalisation */ + float psum; + /* Instantaneous probability of speech and music, with beta pre-applied. */ + float speech0; + float music0; + + /* One transition every 3 minutes of active audio */ + tau = .00005f*frame_probs[1]; + beta = .05f; + if (1) { + /* Adapt beta based on how "unexpected" the new prob is */ + float p, q; + p = MAX16(.05f,MIN16(.95f,frame_probs[0])); + q = MAX16(.05f,MIN16(.95f,tonal->music_prob)); + beta = .01f+.05f*ABS16(p-q)/(p*(1-q)+q*(1-p)); + } + /* p0 and p1 are the probabilities of speech and music at this frame + using only information from previous frame and applying the + state transition model */ + p0 = (1-tonal->music_prob)*(1-tau) + tonal->music_prob *tau; + p1 = tonal->music_prob *(1-tau) + (1-tonal->music_prob)*tau; + /* We apply the current probability with exponent beta to work around + the fact that the probability estimates aren't independent. */ + p0 *= (float)pow(1-frame_probs[0], beta); + p1 *= (float)pow(frame_probs[0], beta); + /* Normalise the probabilities to get the Marokv probability of music. */ + tonal->music_prob = p1/(p0+p1); + info->music_prob = tonal->music_prob; + + /* This chunk of code deals with delayed decision. */ + psum=1e-20f; + /* Instantaneous probability of speech and music, with beta pre-applied. */ + speech0 = (float)pow(1-frame_probs[0], beta); + music0 = (float)pow(frame_probs[0], beta); + if (tonal->count==1) + { + tonal->pspeech[0]=.5; + tonal->pmusic [0]=.5; + } + /* Updated probability of having only speech (s0) or only music (m0), + before considering the new observation. */ + s0 = tonal->pspeech[0] + tonal->pspeech[1]; + m0 = tonal->pmusic [0] + tonal->pmusic [1]; + /* Updates s0 and m0 with instantaneous probability. */ + tonal->pspeech[0] = s0*(1-tau)*speech0; + tonal->pmusic [0] = m0*(1-tau)*music0; + /* Propagate the transition probabilities */ + for (i=1;ipspeech[i] = tonal->pspeech[i+1]*speech0; + tonal->pmusic [i] = tonal->pmusic [i+1]*music0; + } + /* Probability that the latest frame is speech, when all the previous ones were music. */ + tonal->pspeech[DETECT_SIZE-1] = m0*tau*speech0; + /* Probability that the latest frame is music, when all the previous ones were speech. */ + tonal->pmusic [DETECT_SIZE-1] = s0*tau*music0; + + /* Renormalise probabilities to 1 */ + for (i=0;ipspeech[i] + tonal->pmusic[i]; + psum = 1.f/psum; + for (i=0;ipspeech[i] *= psum; + tonal->pmusic [i] *= psum; + } + psum = tonal->pmusic[0]; + for (i=1;ipspeech[i]; + + /* Estimate our confidence in the speech/music decisions */ + if (frame_probs[1]>.75) + { + if (tonal->music_prob>.9) + { + float adapt; + adapt = 1.f/(++tonal->music_confidence_count); + tonal->music_confidence_count = IMIN(tonal->music_confidence_count, 500); + tonal->music_confidence += adapt*MAX16(-.2f,frame_probs[0]-tonal->music_confidence); + } + if (tonal->music_prob<.1) + { + float adapt; + adapt = 1.f/(++tonal->speech_confidence_count); + tonal->speech_confidence_count = IMIN(tonal->speech_confidence_count, 500); + tonal->speech_confidence += adapt*MIN16(.2f,frame_probs[0]-tonal->speech_confidence); + } + } else { + if (tonal->music_confidence_count==0) + tonal->music_confidence = .9f; + if (tonal->speech_confidence_count==0) + tonal->speech_confidence = .1f; + } + } + if (tonal->last_music != (tonal->music_prob>.5f)) + tonal->last_transition=0; + tonal->last_music = tonal->music_prob>.5f; +#else + info->music_prob = 0; +#endif + /*for (i=0;i<25;i++) + printf("%f ", features[i]); + printf("\n");*/ + + info->bandwidth = bandwidth; + /*printf("%d %d\n", info->bandwidth, info->opus_bandwidth);*/ + info->noisiness = frame_noisiness; + info->valid = 1; + if (info_out!=NULL) + OPUS_COPY(info_out, info, 1); + RESTORE_STACK; +} + +void run_analysis(TonalityAnalysisState *analysis, const CELTMode *celt_mode, const void *analysis_pcm, + int analysis_frame_size, int frame_size, int c1, int c2, int C, opus_int32 Fs, + int lsb_depth, downmix_func downmix, AnalysisInfo *analysis_info) +{ + int offset; + int pcm_len; + + if (analysis_pcm != NULL) + { + /* Avoid overflow/wrap-around of the analysis buffer */ + analysis_frame_size = IMIN((DETECT_SIZE-5)*Fs/100, analysis_frame_size); + + pcm_len = analysis_frame_size - analysis->analysis_offset; + offset = analysis->analysis_offset; + do { + tonality_analysis(analysis, NULL, celt_mode, analysis_pcm, IMIN(480, pcm_len), offset, c1, c2, C, lsb_depth, downmix); + offset += 480; + pcm_len -= 480; + } while (pcm_len>0); + analysis->analysis_offset = analysis_frame_size; + + analysis->analysis_offset -= frame_size; + } + + analysis_info->valid = 0; + tonality_get_info(analysis, analysis_info, frame_size); +} diff --git a/code/opus-1.1/src/analysis.h b/code/opus-1.1/src/analysis.h new file mode 100644 index 00000000..be0388fa --- /dev/null +++ b/code/opus-1.1/src/analysis.h @@ -0,0 +1,90 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef ANALYSIS_H +#define ANALYSIS_H + +#include "celt.h" +#include "opus_private.h" + +#define NB_FRAMES 8 +#define NB_TBANDS 18 +#define NB_TOT_BANDS 21 +#define ANALYSIS_BUF_SIZE 720 /* 15 ms at 48 kHz */ + +#define DETECT_SIZE 200 + +typedef struct { + float angle[240]; + float d_angle[240]; + float d2_angle[240]; + opus_val32 inmem[ANALYSIS_BUF_SIZE]; + int mem_fill; /* number of usable samples in the buffer */ + float prev_band_tonality[NB_TBANDS]; + float prev_tonality; + float E[NB_FRAMES][NB_TBANDS]; + float lowE[NB_TBANDS]; + float highE[NB_TBANDS]; + float meanE[NB_TOT_BANDS]; + float mem[32]; + float cmean[8]; + float std[9]; + float music_prob; + float Etracker; + float lowECount; + int E_count; + int last_music; + int last_transition; + int count; + float subframe_mem[3]; + int analysis_offset; + /** Probability of having speech for time i to DETECT_SIZE-1 (and music before). + pspeech[0] is the probability that all frames in the window are speech. */ + float pspeech[DETECT_SIZE]; + /** Probability of having music for time i to DETECT_SIZE-1 (and speech before). + pmusic[0] is the probability that all frames in the window are music. */ + float pmusic[DETECT_SIZE]; + float speech_confidence; + float music_confidence; + int speech_confidence_count; + int music_confidence_count; + int write_pos; + int read_pos; + int read_subframe; + AnalysisInfo info[DETECT_SIZE]; +} TonalityAnalysisState; + +void tonality_analysis(TonalityAnalysisState *tonal, AnalysisInfo *info, + const CELTMode *celt_mode, const void *x, int len, int offset, int c1, int c2, int C, int lsb_depth, downmix_func downmix); + +void tonality_get_info(TonalityAnalysisState *tonal, AnalysisInfo *info_out, int len); + +void run_analysis(TonalityAnalysisState *analysis, const CELTMode *celt_mode, const void *analysis_pcm, + int analysis_frame_size, int frame_size, int c1, int c2, int C, opus_int32 Fs, + int lsb_depth, downmix_func downmix, AnalysisInfo *analysis_info); + +#endif diff --git a/code/opus-1.1/src/mlp.c b/code/opus-1.1/src/mlp.c new file mode 100644 index 00000000..46386026 --- /dev/null +++ b/code/opus-1.1/src/mlp.c @@ -0,0 +1,140 @@ +/* Copyright (c) 2008-2011 Octasic Inc. + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_types.h" +#include "opus_defines.h" + +#include +#include "mlp.h" +#include "arch.h" +#include "tansig_table.h" +#define MAX_NEURONS 100 + +#if 0 +static OPUS_INLINE opus_val16 tansig_approx(opus_val32 _x) /* Q19 */ +{ + int i; + opus_val16 xx; /* Q11 */ + /*double x, y;*/ + opus_val16 dy, yy; /* Q14 */ + /*x = 1.9073e-06*_x;*/ + if (_x>=QCONST32(8,19)) + return QCONST32(1.,14); + if (_x<=-QCONST32(8,19)) + return -QCONST32(1.,14); + xx = EXTRACT16(SHR32(_x, 8)); + /*i = lrint(25*x);*/ + i = SHR32(ADD32(1024,MULT16_16(25, xx)),11); + /*x -= .04*i;*/ + xx -= EXTRACT16(SHR32(MULT16_16(20972,i),8)); + /*x = xx*(1./2048);*/ + /*y = tansig_table[250+i];*/ + yy = tansig_table[250+i]; + /*y = yy*(1./16384);*/ + dy = 16384-MULT16_16_Q14(yy,yy); + yy = yy + MULT16_16_Q14(MULT16_16_Q11(xx,dy),(16384 - MULT16_16_Q11(yy,xx))); + return yy; +} +#else +/*extern const float tansig_table[501];*/ +static OPUS_INLINE float tansig_approx(float x) +{ + int i; + float y, dy; + float sign=1; + /* Tests are reversed to catch NaNs */ + if (!(x<8)) + return 1; + if (!(x>-8)) + return -1; + if (x<0) + { + x=-x; + sign=-1; + } + i = (int)floor(.5f+25*x); + x -= .04f*i; + y = tansig_table[i]; + dy = 1-y*y; + y = y + x*dy*(1 - y*x); + return sign*y; +} +#endif + +#if 0 +void mlp_process(const MLP *m, const opus_val16 *in, opus_val16 *out) +{ + int j; + opus_val16 hidden[MAX_NEURONS]; + const opus_val16 *W = m->weights; + /* Copy to tmp_in */ + for (j=0;jtopo[1];j++) + { + int k; + opus_val32 sum = SHL32(EXTEND32(*W++),8); + for (k=0;ktopo[0];k++) + sum = MAC16_16(sum, in[k],*W++); + hidden[j] = tansig_approx(sum); + } + for (j=0;jtopo[2];j++) + { + int k; + opus_val32 sum = SHL32(EXTEND32(*W++),14); + for (k=0;ktopo[1];k++) + sum = MAC16_16(sum, hidden[k], *W++); + out[j] = tansig_approx(EXTRACT16(PSHR32(sum,17))); + } +} +#else +void mlp_process(const MLP *m, const float *in, float *out) +{ + int j; + float hidden[MAX_NEURONS]; + const float *W = m->weights; + /* Copy to tmp_in */ + for (j=0;jtopo[1];j++) + { + int k; + float sum = *W++; + for (k=0;ktopo[0];k++) + sum = sum + in[k]**W++; + hidden[j] = tansig_approx(sum); + } + for (j=0;jtopo[2];j++) + { + int k; + float sum = *W++; + for (k=0;ktopo[1];k++) + sum = sum + hidden[k]**W++; + out[j] = tansig_approx(sum); + } +} +#endif diff --git a/code/opus-1.1/src/mlp.h b/code/opus-1.1/src/mlp.h new file mode 100644 index 00000000..86c8e061 --- /dev/null +++ b/code/opus-1.1/src/mlp.h @@ -0,0 +1,41 @@ +/* Copyright (c) 2008-2011 Octasic Inc. + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef _MLP_H_ +#define _MLP_H_ + +#include "arch.h" + +typedef struct { + int layers; + const int *topo; + const float *weights; +} MLP; + +void mlp_process(const MLP *m, const float *in, float *out); + +#endif /* _MLP_H_ */ diff --git a/code/opus-1.1/src/mlp_data.c b/code/opus-1.1/src/mlp_data.c new file mode 100644 index 00000000..401c4c02 --- /dev/null +++ b/code/opus-1.1/src/mlp_data.c @@ -0,0 +1,105 @@ +/* The contents of this file was automatically generated by mlp_train.c + It contains multi-layer perceptron (MLP) weights. */ + +#include "mlp.h" + +/* RMS error was 0.138320, seed was 1361535663 */ + +static const float weights[422] = { + +/* hidden layer */ +-0.0941125f, -0.302976f, -0.603555f, -0.19393f, -0.185983f, +-0.601617f, -0.0465317f, -0.114563f, -0.103599f, -0.618938f, +-0.317859f, -0.169949f, -0.0702885f, 0.148065f, 0.409524f, +0.548432f, 0.367649f, -0.494393f, 0.764306f, -1.83957f, +0.170849f, 12.786f, -1.08848f, -1.27284f, -16.2606f, +24.1773f, -5.57454f, -0.17276f, -0.163388f, -0.224421f, +-0.0948944f, -0.0728695f, -0.26557f, -0.100283f, -0.0515459f, +-0.146142f, -0.120674f, -0.180655f, 0.12857f, 0.442138f, +-0.493735f, 0.167767f, 0.206699f, -0.197567f, 0.417999f, +1.50364f, -0.773341f, -10.0401f, 0.401872f, 2.97966f, +15.2165f, -1.88905f, -1.19254f, 0.0285397f, -0.00405139f, +0.0707565f, 0.00825699f, -0.0927269f, -0.010393f, -0.00428882f, +-0.00489743f, -0.0709731f, -0.00255992f, 0.0395619f, 0.226424f, +0.0325231f, 0.162175f, -0.100118f, 0.485789f, 0.12697f, +0.285937f, 0.0155637f, 0.10546f, 3.05558f, 1.15059f, +-1.00904f, -1.83088f, 3.31766f, -3.42516f, -0.119135f, +-0.0405654f, 0.00690068f, 0.0179877f, -0.0382487f, 0.00597941f, +-0.0183611f, 0.00190395f, -0.144322f, -0.0435671f, 0.000990594f, +0.221087f, 0.142405f, 0.484066f, 0.404395f, 0.511955f, +-0.237255f, 0.241742f, 0.35045f, -0.699428f, 10.3993f, +2.6507f, -2.43459f, -4.18838f, 1.05928f, 1.71067f, +0.00667811f, -0.0721335f, -0.0397346f, 0.0362704f, -0.11496f, +-0.0235776f, 0.0082161f, -0.0141741f, -0.0329699f, -0.0354253f, +0.00277404f, -0.290654f, -1.14767f, -0.319157f, -0.686544f, +0.36897f, 0.478899f, 0.182579f, -0.411069f, 0.881104f, +-4.60683f, 1.4697f, 0.335845f, -1.81905f, -30.1699f, +5.55225f, 0.0019508f, -0.123576f, -0.0727332f, -0.0641597f, +-0.0534458f, -0.108166f, -0.0937368f, -0.0697883f, -0.0275475f, +-0.192309f, -0.110074f, 0.285375f, -0.405597f, 0.0926724f, +-0.287881f, -0.851193f, -0.099493f, -0.233764f, -1.2852f, +1.13611f, 3.12168f, -0.0699f, -1.86216f, 2.65292f, +-7.31036f, 2.44776f, -0.00111802f, -0.0632786f, -0.0376296f, +-0.149851f, 0.142963f, 0.184368f, 0.123433f, 0.0756158f, +0.117312f, 0.0933395f, 0.0692163f, 0.0842592f, 0.0704683f, +0.0589963f, 0.0942205f, -0.448862f, 0.0262677f, 0.270352f, +-0.262317f, 0.172586f, 2.00227f, -0.159216f, 0.038422f, +10.2073f, 4.15536f, -2.3407f, -0.0550265f, 0.00964792f, +-0.141336f, 0.0274501f, 0.0343921f, -0.0487428f, 0.0950172f, +-0.00775017f, -0.0372492f, -0.00548121f, -0.0663695f, 0.0960506f, +-0.200008f, -0.0412827f, 0.58728f, 0.0515787f, 0.337254f, +0.855024f, 0.668371f, -0.114904f, -3.62962f, -0.467477f, +-0.215472f, 2.61537f, 0.406117f, -1.36373f, 0.0425394f, +0.12208f, 0.0934502f, 0.123055f, 0.0340935f, -0.142466f, +0.035037f, -0.0490666f, 0.0733208f, 0.0576672f, 0.123984f, +-0.0517194f, -0.253018f, 0.590565f, 0.145849f, 0.315185f, +0.221534f, -0.149081f, 0.216161f, -0.349575f, 24.5664f, +-0.994196f, 0.614289f, -18.7905f, -2.83277f, -0.716801f, +-0.347201f, 0.479515f, -0.246027f, 0.0758683f, 0.137293f, +-0.17781f, 0.118751f, -0.00108329f, -0.237334f, 0.355732f, +-0.12991f, -0.0547627f, -0.318576f, -0.325524f, 0.180494f, +-0.0625604f, 0.141219f, 0.344064f, 0.37658f, -0.591772f, +5.8427f, -0.38075f, 0.221894f, -1.41934f, -1.87943e+06f, +1.34114f, 0.0283355f, -0.0447856f, -0.0211466f, -0.0256927f, +0.0139618f, 0.0207934f, -0.0107666f, 0.0110969f, 0.0586069f, +-0.0253545f, -0.0328433f, 0.11872f, -0.216943f, 0.145748f, +0.119808f, -0.0915211f, -0.120647f, -0.0787719f, -0.143644f, +-0.595116f, -1.152f, -1.25335f, -1.17092f, 4.34023f, +-975268.f, -1.37033f, -0.0401123f, 0.210602f, -0.136656f, +0.135962f, -0.0523293f, 0.0444604f, 0.0143928f, 0.00412666f, +-0.0193003f, 0.218452f, -0.110204f, -2.02563f, 0.918238f, +-2.45362f, 1.19542f, -0.061362f, -1.92243f, 0.308111f, +0.49764f, 0.912356f, 0.209272f, -2.34525f, 2.19326f, +-6.47121f, 1.69771f, -0.725123f, 0.0118929f, 0.0377944f, +0.0554003f, 0.0226452f, -0.0704421f, -0.0300309f, 0.0122978f, +-0.0041782f, -0.0686612f, 0.0313115f, 0.039111f, 0.364111f, +-0.0945548f, 0.0229876f, -0.17414f, 0.329795f, 0.114714f, +0.30022f, 0.106997f, 0.132355f, 5.79932f, 0.908058f, +-0.905324f, -3.3561f, 0.190647f, 0.184211f, -0.673648f, +0.231807f, -0.0586222f, 0.230752f, -0.438277f, 0.245857f, +-0.17215f, 0.0876383f, -0.720512f, 0.162515f, 0.0170571f, +0.101781f, 0.388477f, 1.32931f, 1.08548f, -0.936301f, +-2.36958f, -6.71988f, -3.44376f, 2.13818f, 14.2318f, +4.91459f, -3.09052f, -9.69191f, -0.768234f, 1.79604f, +0.0549653f, 0.163399f, 0.0797025f, 0.0343933f, -0.0555876f, +-0.00505673f, 0.0187258f, 0.0326628f, 0.0231486f, 0.15573f, +0.0476223f, -0.254824f, 1.60155f, -0.801221f, 2.55496f, +0.737629f, -1.36249f, -0.695463f, -2.44301f, -1.73188f, +3.95279f, 1.89068f, 0.486087f, -11.3343f, 3.9416e+06f, + +/* output layer */ +-0.381439f, 0.12115f, -0.906927f, 2.93878f, 1.6388f, +0.882811f, 0.874344f, 1.21726f, -0.874545f, 0.321706f, +0.785055f, 0.946558f, -0.575066f, -3.46553f, 0.884905f, +0.0924047f, -9.90712f, 0.391338f, 0.160103f, -2.04954f, +4.1455f, 0.0684029f, -0.144761f, -0.285282f, 0.379244f, +-1.1584f, -0.0277241f, -9.85f, -4.82386f, 3.71333f, +3.87308f, 3.52558f}; + +static const int topo[3] = {25, 15, 2}; + +const MLP net = { + 3, + topo, + weights +}; diff --git a/code/opus-1.1/src/opus.c b/code/opus-1.1/src/opus.c new file mode 100644 index 00000000..30890b9c --- /dev/null +++ b/code/opus-1.1/src/opus.c @@ -0,0 +1,329 @@ +/* Copyright (c) 2011 Xiph.Org Foundation, Skype Limited + Written by Jean-Marc Valin and Koen Vos */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus.h" +#include "opus_private.h" + +#ifndef DISABLE_FLOAT_API +OPUS_EXPORT void opus_pcm_soft_clip(float *_x, int N, int C, float *declip_mem) +{ + int c; + int i; + float *x; + + if (C<1 || N<1 || !_x || !declip_mem) return; + + /* First thing: saturate everything to +/- 2 which is the highest level our + non-linearity can handle. At the point where the signal reaches +/-2, + the derivative will be zero anyway, so this doesn't introduce any + discontinuity in the derivative. */ + for (i=0;i=0) + break; + x[i*C] = x[i*C]+a*x[i*C]*x[i*C]; + } + + curr=0; + x0 = x[0]; + while(1) + { + int start, end; + float maxval; + int special=0; + int peak_pos; + for (i=curr;i1 || x[i*C]<-1) + break; + } + if (i==N) + { + a=0; + break; + } + peak_pos = i; + start=end=i; + maxval=ABS16(x[i*C]); + /* Look for first zero crossing before clipping */ + while (start>0 && x[i*C]*x[(start-1)*C]>=0) + start--; + /* Look for first zero crossing after clipping */ + while (end=0) + { + /* Look for other peaks until the next zero-crossing. */ + if (ABS16(x[end*C])>maxval) + { + maxval = ABS16(x[end*C]); + peak_pos = end; + } + end++; + } + /* Detect the special case where we clip before the first zero crossing */ + special = (start==0 && x[i*C]*x[0]>=0); + + /* Compute a such that maxval + a*maxval^2 = 1 */ + a=(maxval-1)/(maxval*maxval); + if (x[i*C]>0) + a = -a; + /* Apply soft clipping */ + for (i=start;i=2) + { + /* Add a linear ramp from the first sample to the signal peak. + This avoids a discontinuity at the beginning of the frame. */ + float delta; + float offset = x0-x[0]; + delta = offset / peak_pos; + for (i=curr;i>2; + return 2; + } +} + +static int parse_size(const unsigned char *data, opus_int32 len, opus_int16 *size) +{ + if (len<1) + { + *size = -1; + return -1; + } else if (data[0]<252) + { + *size = data[0]; + return 1; + } else if (len<2) + { + *size = -1; + return -1; + } else { + *size = 4*data[1] + data[0]; + return 2; + } +} + +int opus_packet_parse_impl(const unsigned char *data, opus_int32 len, + int self_delimited, unsigned char *out_toc, + const unsigned char *frames[48], opus_int16 size[48], + int *payload_offset, opus_int32 *packet_offset) +{ + int i, bytes; + int count; + int cbr; + unsigned char ch, toc; + int framesize; + opus_int32 last_size; + opus_int32 pad = 0; + const unsigned char *data0 = data; + + if (size==NULL) + return OPUS_BAD_ARG; + + framesize = opus_packet_get_samples_per_frame(data, 48000); + + cbr = 0; + toc = *data++; + len--; + last_size = len; + switch (toc&0x3) + { + /* One frame */ + case 0: + count=1; + break; + /* Two CBR frames */ + case 1: + count=2; + cbr = 1; + if (!self_delimited) + { + if (len&0x1) + return OPUS_INVALID_PACKET; + last_size = len/2; + /* If last_size doesn't fit in size[0], we'll catch it later */ + size[0] = (opus_int16)last_size; + } + break; + /* Two VBR frames */ + case 2: + count = 2; + bytes = parse_size(data, len, size); + len -= bytes; + if (size[0]<0 || size[0] > len) + return OPUS_INVALID_PACKET; + data += bytes; + last_size = len-size[0]; + break; + /* Multiple CBR/VBR frames (from 0 to 120 ms) */ + default: /*case 3:*/ + if (len<1) + return OPUS_INVALID_PACKET; + /* Number of frames encoded in bits 0 to 5 */ + ch = *data++; + count = ch&0x3F; + if (count <= 0 || framesize*count > 5760) + return OPUS_INVALID_PACKET; + len--; + /* Padding flag is bit 6 */ + if (ch&0x40) + { + int p; + do { + int tmp; + if (len<=0) + return OPUS_INVALID_PACKET; + p = *data++; + len--; + tmp = p==255 ? 254: p; + len -= tmp; + pad += tmp; + } while (p==255); + } + if (len<0) + return OPUS_INVALID_PACKET; + /* VBR flag is bit 7 */ + cbr = !(ch&0x80); + if (!cbr) + { + /* VBR case */ + last_size = len; + for (i=0;i len) + return OPUS_INVALID_PACKET; + data += bytes; + last_size -= bytes+size[i]; + } + if (last_size<0) + return OPUS_INVALID_PACKET; + } else if (!self_delimited) + { + /* CBR case */ + last_size = len/count; + if (last_size*count!=len) + return OPUS_INVALID_PACKET; + for (i=0;i len) + return OPUS_INVALID_PACKET; + data += bytes; + /* For CBR packets, apply the size to all the frames. */ + if (cbr) + { + if (size[count-1]*count > len) + return OPUS_INVALID_PACKET; + for (i=0;i last_size) + return OPUS_INVALID_PACKET; + } else + { + /* Because it's not encoded explicitly, it's possible the size of the + last packet (or all the packets, for the CBR case) is larger than + 1275. Reject them here.*/ + if (last_size > 1275) + return OPUS_INVALID_PACKET; + size[count-1] = (opus_int16)last_size; + } + + if (payload_offset) + *payload_offset = (int)(data-data0); + + for (i=0;i= 2) && !defined(__OPTIMIZE__) +# pragma message "You appear to be compiling without optimization, if so opus will be very slow." #endif #include @@ -46,6 +50,7 @@ #include "structs.h" #include "define.h" #include "mathops.h" +#include "cpu_support.h" struct OpusDecoder { int celt_dec_offset; @@ -65,12 +70,15 @@ struct OpusDecoder { int frame_size; int prev_redundancy; int last_packet_duration; +#ifndef FIXED_POINT + opus_val16 softclip_mem[2]; +#endif opus_uint32 rangeFinal; }; #ifdef FIXED_POINT -static inline opus_int16 SAT16(opus_int32 x) { +static OPUS_INLINE opus_int16 SAT16(opus_int32 x) { return x > 32767 ? 32767 : x < -32768 ? -32768 : (opus_int16)x; } #endif @@ -201,8 +209,14 @@ static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, int i, silk_ret=0, celt_ret=0; ec_dec dec; opus_int32 silk_frame_size; + int pcm_silk_size; VARDECL(opus_int16, pcm_silk); - VARDECL(opus_val16, pcm_transition); + int pcm_transition_silk_size; + VARDECL(opus_val16, pcm_transition_silk); + int pcm_transition_celt_size; + VARDECL(opus_val16, pcm_transition_celt); + opus_val16 *pcm_transition; + int redundant_audio_size; VARDECL(opus_val16, redundant_audio); int audiosize; @@ -245,33 +259,61 @@ static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, ec_dec_init(&dec,(unsigned char*)data,len); } else { audiosize = frame_size; + mode = st->prev_mode; - if (st->prev_mode == 0) + if (mode == 0) { /* If we haven't got any packet yet, all we can do is return zeros */ for (i=0;ichannels;i++) pcm[i] = 0; RESTORE_STACK; return audiosize; - } else { - mode = st->prev_mode; + } + + /* Avoids trying to run the PLC on sizes other than 2.5 (CELT), 5 (CELT), + 10, or 20 (e.g. 12.5 or 30 ms). */ + if (audiosize > F20) + { + do { + int ret = opus_decode_frame(st, NULL, 0, pcm, IMIN(audiosize, F20), 0); + if (ret<0) + { + RESTORE_STACK; + return ret; + } + pcm += ret*st->channels; + audiosize -= ret; + } while (audiosize > 0); + RESTORE_STACK; + return frame_size; + } else if (audiosize < F20) + { + if (audiosize > F10) + audiosize = F10; + else if (mode != MODE_SILK_ONLY && audiosize > F5 && audiosize < F10) + audiosize = F5; } } - /* For CELT/hybrid PLC of more than 20 ms, opus_decode_native() will do - multiple calls */ - if (data==NULL && mode != MODE_SILK_ONLY) - frame_size = IMIN(frame_size, F20); - ALLOC(pcm_transition, F5*st->channels, opus_val16); - + pcm_transition_silk_size = ALLOC_NONE; + pcm_transition_celt_size = ALLOC_NONE; if (data!=NULL && st->prev_mode > 0 && ( (mode == MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY && !st->prev_redundancy) || (mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) ) ) { transition = 1; + /* Decide where to allocate the stack memory for pcm_transition */ if (mode == MODE_CELT_ONLY) - opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); + pcm_transition_celt_size = F5*st->channels; + else + pcm_transition_silk_size = F5*st->channels; + } + ALLOC(pcm_transition_celt, pcm_transition_celt_size, opus_val16); + if (transition && mode == MODE_CELT_ONLY) + { + pcm_transition = pcm_transition_celt; + opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); } if (audiosize > frame_size) { @@ -282,8 +324,9 @@ static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, frame_size = audiosize; } - ALLOC(pcm_silk, IMAX(F10, frame_size)*st->channels, opus_int16); - ALLOC(redundant_audio, F5*st->channels, opus_val16); + /* Don't allocate any memory when in CELT-only mode */ + pcm_silk_size = (mode != MODE_CELT_ONLY) ? IMAX(F10, frame_size)*st->channels : ALLOC_NONE; + ALLOC(pcm_silk, pcm_silk_size, opus_int16); /* SILK processing */ if (mode != MODE_CELT_ONLY) @@ -332,7 +375,7 @@ static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, pcm_ptr[i] = 0; } else { RESTORE_STACK; - return OPUS_INVALID_PACKET; + return OPUS_INTERNAL_ERROR; } } pcm_ptr += silk_frame_size * st->channels; @@ -397,10 +440,22 @@ static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, } if (redundancy) + { transition = 0; + pcm_transition_silk_size=ALLOC_NONE; + } + + ALLOC(pcm_transition_silk, pcm_transition_silk_size, opus_val16); if (transition && mode != MODE_CELT_ONLY) + { + pcm_transition = pcm_transition_silk; opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); + } + + /* Only allocation memory for redundancy if/when needed */ + redundant_audio_size = redundancy ? F5*st->channels : ALLOC_NONE; + ALLOC(redundant_audio, redundant_audio_size, opus_val16); /* 5 ms redundant frame for CELT->SILK*/ if (redundancy && celt_to_silk) @@ -514,197 +569,28 @@ static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, st->prev_mode = mode; st->prev_redundancy = redundancy && !celt_to_silk; + + if (celt_ret>=0) + { + if (OPUS_CHECK_ARRAY(pcm, audiosize*st->channels)) + OPUS_PRINT_INT(audiosize); + } + RESTORE_STACK; return celt_ret < 0 ? celt_ret : audiosize; } -static int parse_size(const unsigned char *data, opus_int32 len, short *size) -{ - if (len<1) - { - *size = -1; - return -1; - } else if (data[0]<252) - { - *size = data[0]; - return 1; - } else if (len<2) - { - *size = -1; - return -1; - } else { - *size = 4*data[1] + data[0]; - return 2; - } -} - -static int opus_packet_parse_impl(const unsigned char *data, opus_int32 len, - int self_delimited, unsigned char *out_toc, - const unsigned char *frames[48], short size[48], int *payload_offset) -{ - int i, bytes; - int count; - int cbr; - unsigned char ch, toc; - int framesize; - opus_int32 last_size; - const unsigned char *data0 = data; - - if (size==NULL) - return OPUS_BAD_ARG; - - framesize = opus_packet_get_samples_per_frame(data, 48000); - - cbr = 0; - toc = *data++; - len--; - last_size = len; - switch (toc&0x3) - { - /* One frame */ - case 0: - count=1; - break; - /* Two CBR frames */ - case 1: - count=2; - cbr = 1; - if (!self_delimited) - { - if (len&0x1) - return OPUS_INVALID_PACKET; - last_size = len/2; - /* If last_size doesn't fit in size[0], we'll catch it later */ - size[0] = (short)last_size; - } - break; - /* Two VBR frames */ - case 2: - count = 2; - bytes = parse_size(data, len, size); - len -= bytes; - if (size[0]<0 || size[0] > len) - return OPUS_INVALID_PACKET; - data += bytes; - last_size = len-size[0]; - break; - /* Multiple CBR/VBR frames (from 0 to 120 ms) */ - default: /*case 3:*/ - if (len<1) - return OPUS_INVALID_PACKET; - /* Number of frames encoded in bits 0 to 5 */ - ch = *data++; - count = ch&0x3F; - if (count <= 0 || framesize*count > 5760) - return OPUS_INVALID_PACKET; - len--; - /* Padding flag is bit 6 */ - if (ch&0x40) - { - int p; - do { - if (len<=0) - return OPUS_INVALID_PACKET; - p = *data++; - len--; - len -= p==255 ? 254: p; - } while (p==255); - } - if (len<0) - return OPUS_INVALID_PACKET; - /* VBR flag is bit 7 */ - cbr = !(ch&0x80); - if (!cbr) - { - /* VBR case */ - last_size = len; - for (i=0;i len) - return OPUS_INVALID_PACKET; - data += bytes; - last_size -= bytes+size[i]; - } - if (last_size<0) - return OPUS_INVALID_PACKET; - } else if (!self_delimited) - { - /* CBR case */ - last_size = len/count; - if (last_size*count!=len) - return OPUS_INVALID_PACKET; - for (i=0;i len) - return OPUS_INVALID_PACKET; - data += bytes; - /* For CBR packets, apply the size to all the frames. */ - if (cbr) - { - if (size[count-1]*count > len) - return OPUS_INVALID_PACKET; - for (i=0;i last_size) - return OPUS_INVALID_PACKET; - } else - { - /* Because it's not encoded explicitly, it's possible the size of the - last packet (or all the packets, for the CBR case) is larger than - 1275. Reject them here.*/ - if (last_size > 1275) - return OPUS_INVALID_PACKET; - size[count-1] = (short)last_size; - } - - if (frames) - { - for (i=0;i1) return OPUS_BAD_ARG; /* For FEC/PLC, frame_size has to be to have a multiple of 2.5 ms */ @@ -715,12 +601,14 @@ int opus_decode_native(OpusDecoder *st, const unsigned char *data, int pcm_count=0; do { int ret; - ret = opus_decode_frame(st, NULL, 0, pcm, frame_size-pcm_count, 0); + ret = opus_decode_frame(st, NULL, 0, pcm+pcm_count*st->channels, frame_size-pcm_count, 0); if (ret<0) return ret; - pcm += st->channels*ret; pcm_count += ret; } while (pcm_count < frame_size); + celt_assert(pcm_count == frame_size); + if (OPUS_CHECK_ARRAY(pcm, pcm_count*st->channels)) + OPUS_PRINT_INT(pcm_count); st->last_packet_duration = pcm_count; return pcm_count; } else if (len<0) @@ -731,7 +619,10 @@ int opus_decode_native(OpusDecoder *st, const unsigned char *data, packet_frame_size = opus_packet_get_samples_per_frame(data, st->Fs); packet_stream_channels = opus_packet_get_nb_channels(data); - count = opus_packet_parse_impl(data, len, self_delimited, &toc, NULL, size, &offset); + count = opus_packet_parse_impl(data, len, self_delimited, &toc, NULL, + size, &offset, packet_offset); + if (count<0) + return count; data += offset; @@ -740,15 +631,19 @@ int opus_decode_native(OpusDecoder *st, const unsigned char *data, int duration_copy; int ret; /* If no FEC can be present, run the PLC (recursive call) */ - if (frame_size <= packet_frame_size || packet_mode == MODE_CELT_ONLY || st->mode == MODE_CELT_ONLY) - return opus_decode_native(st, NULL, 0, pcm, frame_size, 0, 0, NULL); + if (frame_size < packet_frame_size || packet_mode == MODE_CELT_ONLY || st->mode == MODE_CELT_ONLY) + return opus_decode_native(st, NULL, 0, pcm, frame_size, 0, 0, NULL, soft_clip); /* Otherwise, run the PLC on everything except the size for which we might have FEC */ duration_copy = st->last_packet_duration; - ret = opus_decode_native(st, NULL, 0, pcm, frame_size-packet_frame_size, 0, 0, NULL); - if (ret<0) + if (frame_size-packet_frame_size!=0) { - st->last_packet_duration = duration_copy; - return ret; + ret = opus_decode_native(st, NULL, 0, pcm, frame_size-packet_frame_size, 0, 0, NULL, soft_clip); + if (ret<0) + { + st->last_packet_duration = duration_copy; + return ret; + } + celt_assert(ret==frame_size-packet_frame_size); } /* Complete with FEC */ st->mode = packet_mode; @@ -759,14 +654,13 @@ int opus_decode_native(OpusDecoder *st, const unsigned char *data, packet_frame_size, 1); if (ret<0) return ret; - st->last_packet_duration = frame_size; - return frame_size; + else { + if (OPUS_CHECK_ARRAY(pcm, frame_size*st->channels)) + OPUS_PRINT_INT(frame_size); + st->last_packet_duration = frame_size; + return frame_size; + } } - tot_offset = 0; - if (count < 0) - return count; - - tot_offset += offset; if (count*packet_frame_size > frame_size) return OPUS_BUFFER_TOO_SMALL; @@ -781,17 +675,22 @@ int opus_decode_native(OpusDecoder *st, const unsigned char *data, for (i=0;ichannels, frame_size-nb_samples, 0); if (ret<0) return ret; + celt_assert(ret==packet_frame_size); data += size[i]; - tot_offset += size[i]; - pcm += ret*st->channels; nb_samples += ret; } - if (packet_offset != NULL) - *packet_offset = tot_offset; st->last_packet_duration = nb_samples; + if (OPUS_CHECK_ARRAY(pcm, nb_samples*st->channels)) + OPUS_PRINT_INT(nb_samples); +#ifndef FIXED_POINT + if (soft_clip) + opus_pcm_soft_clip(pcm, nb_samples, st->channels, st->softclip_mem); + else + st->softclip_mem[0]=st->softclip_mem[1]=0; +#endif return nb_samples; } @@ -800,7 +699,9 @@ int opus_decode_native(OpusDecoder *st, const unsigned char *data, int opus_decode(OpusDecoder *st, const unsigned char *data, opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) { - return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL); + if(frame_size<=0) + return OPUS_BAD_ARG; + return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL, 0); } #ifndef DISABLE_FLOAT_API @@ -811,9 +712,14 @@ int opus_decode_float(OpusDecoder *st, const unsigned char *data, int ret, i; ALLOC_STACK; + if(frame_size<=0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } ALLOC(out, frame_size*st->channels, opus_int16); - ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL); + ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL, 0); if (ret > 0) { for (i=0;ichannels;i++) @@ -833,7 +739,7 @@ int opus_decode(OpusDecoder *st, const unsigned char *data, int ret, i; ALLOC_STACK; - if(frame_size<0) + if(frame_size<=0) { RESTORE_STACK; return OPUS_BAD_ARG; @@ -841,7 +747,7 @@ int opus_decode(OpusDecoder *st, const unsigned char *data, ALLOC(out, frame_size*st->channels, float); - ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL); + ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL, 1); if (ret > 0) { for (i=0;ichannels;i++) @@ -854,7 +760,9 @@ int opus_decode(OpusDecoder *st, const unsigned char *data, int opus_decode_float(OpusDecoder *st, const unsigned char *data, opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) { - return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL); + if(frame_size<=0) + return OPUS_BAD_ARG; + return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL, 0); } #endif @@ -877,12 +785,20 @@ int opus_decoder_ctl(OpusDecoder *st, int request, ...) case OPUS_GET_BANDWIDTH_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->bandwidth; } break; case OPUS_GET_FINAL_RANGE_REQUEST: { opus_uint32 *value = va_arg(ap, opus_uint32*); + if (!value) + { + goto bad_arg; + } *value = st->rangeFinal; } break; @@ -901,10 +817,9 @@ int opus_decoder_ctl(OpusDecoder *st, int request, ...) case OPUS_GET_SAMPLE_RATE_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); - if (value==NULL) + if (!value) { - ret = OPUS_BAD_ARG; - break; + goto bad_arg; } *value = st->Fs; } @@ -912,10 +827,9 @@ int opus_decoder_ctl(OpusDecoder *st, int request, ...) case OPUS_GET_PITCH_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); - if (value==NULL) + if (!value) { - ret = OPUS_BAD_ARG; - break; + goto bad_arg; } if (st->prev_mode == MODE_CELT_ONLY) celt_decoder_ctl(celt_dec, OPUS_GET_PITCH(value)); @@ -926,10 +840,9 @@ int opus_decoder_ctl(OpusDecoder *st, int request, ...) case OPUS_GET_GAIN_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); - if (value==NULL) + if (!value) { - ret = OPUS_BAD_ARG; - break; + goto bad_arg; } *value = st->decode_gain; } @@ -939,8 +852,7 @@ int opus_decoder_ctl(OpusDecoder *st, int request, ...) opus_int32 value = va_arg(ap, opus_int32); if (value<-32768 || value>32767) { - ret = OPUS_BAD_ARG; - break; + goto bad_arg; } st->decode_gain = value; } @@ -948,6 +860,10 @@ int opus_decoder_ctl(OpusDecoder *st, int request, ...) case OPUS_GET_LAST_PACKET_DURATION_REQUEST: { opus_uint32 *value = va_arg(ap, opus_uint32*); + if (!value) + { + goto bad_arg; + } *value = st->last_packet_duration; } break; @@ -959,6 +875,9 @@ int opus_decoder_ctl(OpusDecoder *st, int request, ...) va_end(ap); return ret; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; } void opus_decoder_destroy(OpusDecoder *st) diff --git a/code/opus-1.0.2/src/opus_encoder.c b/code/opus-1.1/src/opus_encoder.c similarity index 59% rename from code/opus-1.0.2/src/opus_encoder.c rename to code/opus-1.1/src/opus_encoder.c index aae31256..fbd3de63 100644 --- a/code/opus-1.0.2/src/opus_encoder.c +++ b/code/opus-1.1/src/opus_encoder.c @@ -40,7 +40,9 @@ #include "arch.h" #include "opus_private.h" #include "os_support.h" - +#include "cpu_support.h" +#include "analysis.h" +#include "mathops.h" #include "tuning_parameters.h" #ifdef FIXED_POINT #include "fixed/structs_FIX.h" @@ -50,6 +52,12 @@ #define MAX_ENCODER_BUFFER 480 +typedef struct { + opus_val32 XX, XY, YY; + opus_val16 smoothed_width; + opus_val16 max_follower; +} StereoWidthState; + struct OpusEncoder { int celt_enc_offset; int silk_enc_offset; @@ -66,14 +74,18 @@ struct OpusEncoder { opus_int32 Fs; int use_vbr; int vbr_constraint; + int variable_duration; opus_int32 bitrate_bps; opus_int32 user_bitrate_bps; + int lsb_depth; int encoder_buffer; + int lfe; #define OPUS_ENCODER_RESET_START stream_channels int stream_channels; opus_int16 hybrid_stereo_width_Q14; opus_int32 variable_HP_smth2_Q15; + opus_val16 prev_HB_gain; opus_val32 hp_mem[4]; int mode; int prev_mode; @@ -83,9 +95,16 @@ struct OpusEncoder { int silk_bw_switch; /* Sampling rate (at the API level) */ int first; + opus_val16 * energy_masking; + StereoWidthState width_mem; opus_val16 delay_buffer[MAX_ENCODER_BUFFER*2]; - +#ifndef DISABLE_FLOAT_API + TonalityAnalysisState analysis; + int detected_bandwidth; + int analysis_offset; +#endif opus_uint32 rangeFinal; + int arch; }; /* Transition tables for the voice and music. First column is the @@ -94,36 +113,36 @@ struct OpusEncoder { static const opus_int32 mono_voice_bandwidth_thresholds[8] = { 11000, 1000, /* NB<->MB */ 14000, 1000, /* MB<->WB */ - 21000, 2000, /* WB<->SWB */ - 29000, 2000, /* SWB<->FB */ + 17000, 1000, /* WB<->SWB */ + 21000, 2000, /* SWB<->FB */ }; static const opus_int32 mono_music_bandwidth_thresholds[8] = { - 14000, 1000, /* MB not allowed */ - 18000, 2000, /* MB<->WB */ - 24000, 2000, /* WB<->SWB */ - 33000, 2000, /* SWB<->FB */ + 12000, 1000, /* NB<->MB */ + 15000, 1000, /* MB<->WB */ + 18000, 2000, /* WB<->SWB */ + 22000, 2000, /* SWB<->FB */ }; static const opus_int32 stereo_voice_bandwidth_thresholds[8] = { 11000, 1000, /* NB<->MB */ 14000, 1000, /* MB<->WB */ 21000, 2000, /* WB<->SWB */ - 32000, 2000, /* SWB<->FB */ + 28000, 2000, /* SWB<->FB */ }; static const opus_int32 stereo_music_bandwidth_thresholds[8] = { - 14000, 1000, /* MB not allowed */ + 12000, 1000, /* NB<->MB */ 18000, 2000, /* MB<->WB */ - 24000, 2000, /* WB<->SWB */ - 48000, 2000, /* SWB<->FB */ + 21000, 2000, /* WB<->SWB */ + 30000, 2000, /* SWB<->FB */ }; /* Threshold bit-rates for switching between mono and stereo */ -static const opus_int32 stereo_voice_threshold = 26000; -static const opus_int32 stereo_music_threshold = 36000; +static const opus_int32 stereo_voice_threshold = 30000; +static const opus_int32 stereo_music_threshold = 30000; /* Threshold bit-rate for switching between SILK/hybrid and CELT-only */ static const opus_int32 mode_thresholds[2][2] = { /* voice */ /* music */ - { 48000, 24000}, /* mono */ - { 48000, 24000}, /* stereo */ + { 64000, 16000}, /* mono */ + { 36000, 16000}, /* stereo */ }; int opus_encoder_get_size(int channels) @@ -167,7 +186,9 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat st->Fs = Fs; - ret = silk_InitEncoder( silk_enc, &st->silk_mode ); + st->arch = opus_select_arch(); + + ret = silk_InitEncoder( silk_enc, st->arch, &st->silk_mode ); if(ret)return OPUS_INTERNAL_ERROR; /* default SILK parameters */ @@ -180,18 +201,19 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat st->silk_mode.payloadSize_ms = 20; st->silk_mode.bitRate = 25000; st->silk_mode.packetLossPercentage = 0; - st->silk_mode.complexity = 10; + st->silk_mode.complexity = 9; st->silk_mode.useInBandFEC = 0; st->silk_mode.useDTX = 0; st->silk_mode.useCBR = 0; + st->silk_mode.reducedDependency = 0; /* Create CELT encoder */ /* Initialize CELT encoder */ - err = celt_encoder_init(celt_enc, Fs, channels); + err = celt_encoder_init(celt_enc, Fs, channels, st->arch); if(err!=OPUS_OK)return OPUS_INTERNAL_ERROR; celt_encoder_ctl(celt_enc, CELT_SET_SIGNALLING(0)); - celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(10)); + celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(st->silk_mode.complexity)); st->use_vbr = 1; /* Makes constrained VBR the default (safer for real-time use) */ @@ -206,12 +228,15 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat st->user_forced_mode = OPUS_AUTO; st->voice_ratio = -1; st->encoder_buffer = st->Fs/100; + st->lsb_depth = 24; + st->variable_duration = OPUS_FRAMESIZE_ARG; /* Delay compensation of 4 ms (2.5 ms for SILK's extra look-ahead + 1.5 ms for SILK resamplers and stereo prediction) */ st->delay_compensation = st->Fs/250; st->hybrid_stereo_width_Q14 = 1 << 14; + st->prev_HB_gain = Q15ONE; st->variable_HP_smth2_Q15 = silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 ); st->first = 1; st->mode = MODE_HYBRID; @@ -220,44 +245,6 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat return OPUS_OK; } -static int pad_frame(unsigned char *data, opus_int32 len, opus_int32 new_len) -{ - if (len == new_len) - return 0; - if (len > new_len) - return 1; - - if ((data[0]&0x3)==0) - { - int i; - int padding, nb_255s; - - padding = new_len - len; - if (padding >= 2) - { - nb_255s = (padding-2)/255; - - for (i=len-1;i>=1;i--) - data[i+nb_255s+2] = data[i]; - data[0] |= 0x3; - data[1] = 0x41; - for (i=0;i=1;i--) - data[i+1] = data[i]; - data[0] |= 0x3; - data[1] = 1; - } - return 0; - } else { - return 1; - } -} - static unsigned char gen_toc(int mode, int framerate, int bandwidth, int channels) { int period; @@ -322,7 +309,7 @@ static void silk_biquad_float( S[ 0 ] = S[1] - vout*A[0] + B[1]*inval; - S[ 1 ] = - vout*A[1] + B[2]*inval; + S[ 1 ] = - vout*A[1] + B[2]*inval + VERY_SMALL; /* Scale back to Q0 and saturate */ out[ k*stride ] = vout; @@ -365,6 +352,56 @@ static void hp_cutoff(const opus_val16 *in, opus_int32 cutoff_Hz, opus_val16 *ou #endif } +#ifdef FIXED_POINT +static void dc_reject(const opus_val16 *in, opus_int32 cutoff_Hz, opus_val16 *out, opus_val32 *hp_mem, int len, int channels, opus_int32 Fs) +{ + int c, i; + int shift; + + /* Approximates -round(log2(4.*cutoff_Hz/Fs)) */ + shift=celt_ilog2(Fs/(cutoff_Hz*3)); + for (c=0;cuser_bitrate_bps; } -#ifdef FIXED_POINT -#define opus_encode_native opus_encode -opus_int32 opus_encode(OpusEncoder *st, const opus_val16 *pcm, int frame_size, - unsigned char *data, opus_int32 out_data_bytes) -#else -#define opus_encode_native opus_encode_float -opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_size, - unsigned char *data, opus_int32 out_data_bytes) +#ifndef DISABLE_FLOAT_API +/* Don't use more than 60 ms for the frame size analysis */ +#define MAX_DYNAMIC_FRAMESIZE 24 +/* Estimates how much the bitrate will be boosted based on the sub-frame energy */ +static float transient_boost(const float *E, const float *E_1, int LM, int maxM) +{ + int i; + int M; + float sumE=0, sumE_1=0; + float metric; + + M = IMIN(maxM, (1<10 ? 1 : 0;*/ + /*return MAX16(0,1-exp(-.25*(metric-2.)));*/ + return MIN16(1,(float)sqrt(MAX16(0,.05f*(metric-2)))); +} + +/* Viterbi decoding trying to find the best frame size combination using look-ahead + + State numbering: + 0: unused + 1: 2.5 ms + 2: 5 ms (#1) + 3: 5 ms (#2) + 4: 10 ms (#1) + 5: 10 ms (#2) + 6: 10 ms (#3) + 7: 10 ms (#4) + 8: 20 ms (#1) + 9: 20 ms (#2) + 10: 20 ms (#3) + 11: 20 ms (#4) + 12: 20 ms (#5) + 13: 20 ms (#6) + 14: 20 ms (#7) + 15: 20 ms (#8) +*/ +static int transient_viterbi(const float *E, const float *E_1, int N, int frame_cost, int rate) +{ + int i; + float cost[MAX_DYNAMIC_FRAMESIZE][16]; + int states[MAX_DYNAMIC_FRAMESIZE][16]; + float best_cost; + int best_state; + float factor; + /* Take into account that we damp VBR in the 32 kb/s to 64 kb/s range. */ + if (rate<80) + factor=0; + else if (rate>160) + factor=1; + else + factor = (rate-80.f)/80.f; + /* Makes variable framesize less aggressive at lower bitrates, but I can't + find any valid theoretical justification for this (other than it seems + to help) */ + for (i=0;i<16;i++) + { + /* Impossible state */ + states[0][i] = -1; + cost[0][i] = 1e10; + } + for (i=0;i<4;i++) + { + cost[0][1<=0;i--) + { + /*printf("%d ", best_state);*/ + best_state = states[i][best_state]; + } + /*printf("%d\n", best_state);*/ + return best_state; +} + +int optimize_framesize(const opus_val16 *x, int len, int C, opus_int32 Fs, + int bitrate, opus_val16 tonality, float *mem, int buffering, + downmix_func downmix) +{ + int N; + int i; + float e[MAX_DYNAMIC_FRAMESIZE+4]; + float e_1[MAX_DYNAMIC_FRAMESIZE+3]; + opus_val32 memx; + int bestLM=0; + int subframe; + int pos; + VARDECL(opus_val32, sub); + + subframe = Fs/400; + ALLOC(sub, subframe, opus_val32); + e[0]=mem[0]; + e_1[0]=1.f/(EPSILON+mem[0]); + if (buffering) + { + /* Consider the CELT delay when not in restricted-lowdelay */ + /* We assume the buffering is between 2.5 and 5 ms */ + int offset = 2*subframe - buffering; + celt_assert(offset>=0 && offset <= subframe); + x += C*offset; + len -= offset; + e[1]=mem[1]; + e_1[1]=1.f/(EPSILON+mem[1]); + e[2]=mem[2]; + e_1[2]=1.f/(EPSILON+mem[2]); + pos = 3; + } else { + pos=1; + } + N=IMIN(len/subframe, MAX_DYNAMIC_FRAMESIZE); + /* Just silencing a warning, it's really initialized later */ + memx = 0; + for (i=0;i-1) + { + for (j=0;j-1) + { + for (j=0;j= OPUS_FRAMESIZE_2_5_MS && variable_duration <= OPUS_FRAMESIZE_60_MS) + new_size = IMIN(3*Fs/50, (Fs/400)<<(variable_duration-OPUS_FRAMESIZE_2_5_MS)); + else + return -1; + if (new_size>frame_size) + return -1; + if (400*new_size!=Fs && 200*new_size!=Fs && 100*new_size!=Fs && + 50*new_size!=Fs && 25*new_size!=Fs && 50*new_size!=3*Fs) + return -1; + return new_size; +} + +opus_int32 compute_frame_size(const void *analysis_pcm, int frame_size, + int variable_duration, int C, opus_int32 Fs, int bitrate_bps, + int delay_compensation, downmix_func downmix +#ifndef DISABLE_FLOAT_API + , float *subframe_mem +#endif + ) +{ +#ifndef DISABLE_FLOAT_API + if (variable_duration == OPUS_FRAMESIZE_VARIABLE && frame_size >= Fs/200) + { + int LM = 3; + LM = optimize_framesize(analysis_pcm, frame_size, C, Fs, bitrate_bps, + 0, subframe_mem, delay_compensation, downmix); + while ((Fs/400<frame_size) + LM--; + frame_size = (Fs/400<XX += MULT16_32_Q15(short_alpha, xx-mem->XX); + mem->XY += MULT16_32_Q15(short_alpha, xy-mem->XY); + mem->YY += MULT16_32_Q15(short_alpha, yy-mem->YY); + mem->XX = MAX32(0, mem->XX); + mem->XY = MAX32(0, mem->XY); + mem->YY = MAX32(0, mem->YY); + if (MAX32(mem->XX, mem->YY)>QCONST16(8e-4f, 18)) + { + sqrt_xx = celt_sqrt(mem->XX); + sqrt_yy = celt_sqrt(mem->YY); + qrrt_xx = celt_sqrt(sqrt_xx); + qrrt_yy = celt_sqrt(sqrt_yy); + /* Inter-channel correlation */ + mem->XY = MIN32(mem->XY, sqrt_xx*sqrt_yy); + corr = SHR32(frac_div32(mem->XY,EPSILON+MULT16_16(sqrt_xx,sqrt_yy)),16); + /* Approximate loudness difference */ + ldiff = Q15ONE*ABS16(qrrt_xx-qrrt_yy)/(EPSILON+qrrt_xx+qrrt_yy); + width = MULT16_16_Q15(celt_sqrt(QCONST32(1.f,30)-MULT16_16(corr,corr)), ldiff); + /* Smoothing over one second */ + mem->smoothed_width += (width-mem->smoothed_width)/frame_rate; + /* Peak follower */ + mem->max_follower = MAX16(mem->max_follower-QCONST16(.02f,15)/frame_rate, mem->smoothed_width); + } else { + width = 0; + corr=Q15ONE; + ldiff=0; + } + /*printf("%f %f %f %f %f ", corr/(float)Q15ONE, ldiff/(float)Q15ONE, width/(float)Q15ONE, mem->smoothed_width/(float)Q15ONE, mem->max_follower/(float)Q15ONE);*/ + return EXTRACT16(MIN32(Q15ONE,20*mem->max_follower)); +} + +opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_size, + unsigned char *data, opus_int32 out_data_bytes, int lsb_depth, + const void *analysis_pcm, opus_int32 analysis_size, int c1, int c2, int analysis_channels, downmix_func downmix) { void *silk_enc; CELTEncoder *celt_enc; @@ -471,7 +949,14 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s int frame_rate; opus_int32 max_rate; /* Max bitrate we're allowed to use */ int curr_bandwidth; + opus_val16 HB_gain; opus_int32 max_data_bytes; /* Max number of bytes we're allowed to use */ + int total_buffer; + opus_val16 stereo_width; + const CELTMode *celt_mode; + AnalysisInfo analysis_info; + int analysis_read_pos_bak=-1; + int analysis_read_subframe_bak=-1; VARDECL(opus_val16, tmp_prefill); ALLOC_STACK; @@ -479,25 +964,70 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s max_data_bytes = IMIN(1276, out_data_bytes); st->rangeFinal = 0; - if (400*frame_size != st->Fs && 200*frame_size != st->Fs && 100*frame_size != st->Fs && + if ((!st->variable_duration && 400*frame_size != st->Fs && 200*frame_size != st->Fs && 100*frame_size != st->Fs && 50*frame_size != st->Fs && 25*frame_size != st->Fs && 50*frame_size != 3*st->Fs) - { - RESTORE_STACK; - return OPUS_BAD_ARG; - } - if (max_data_bytes<=0) + || (400*frame_size < st->Fs) + || max_data_bytes<=0 + ) { RESTORE_STACK; return OPUS_BAD_ARG; } silk_enc = (char*)st+st->silk_enc_offset; celt_enc = (CELTEncoder*)((char*)st+st->celt_enc_offset); - if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) delay_compensation = 0; else delay_compensation = st->delay_compensation; + lsb_depth = IMIN(lsb_depth, st->lsb_depth); + + analysis_info.valid = 0; + celt_encoder_ctl(celt_enc, CELT_GET_MODE(&celt_mode)); +#ifndef DISABLE_FLOAT_API +#ifdef FIXED_POINT + if (st->silk_mode.complexity >= 10 && st->Fs==48000) +#else + if (st->silk_mode.complexity >= 7 && st->Fs==48000) +#endif + { + analysis_read_pos_bak = st->analysis.read_pos; + analysis_read_subframe_bak = st->analysis.read_subframe; + run_analysis(&st->analysis, celt_mode, analysis_pcm, analysis_size, frame_size, + c1, c2, analysis_channels, st->Fs, + lsb_depth, downmix, &analysis_info); + } +#endif + + st->voice_ratio = -1; + +#ifndef DISABLE_FLOAT_API + st->detected_bandwidth = 0; + if (analysis_info.valid) + { + int analysis_bandwidth; + if (st->signal_type == OPUS_AUTO) + st->voice_ratio = (int)floor(.5+100*(1-analysis_info.music_prob)); + + analysis_bandwidth = analysis_info.bandwidth; + if (analysis_bandwidth<=12) + st->detected_bandwidth = OPUS_BANDWIDTH_NARROWBAND; + else if (analysis_bandwidth<=14) + st->detected_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + else if (analysis_bandwidth<=16) + st->detected_bandwidth = OPUS_BANDWIDTH_WIDEBAND; + else if (analysis_bandwidth<=18) + st->detected_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; + else + st->detected_bandwidth = OPUS_BANDWIDTH_FULLBAND; + } +#endif + + if (st->channels==2 && st->force_channels!=1) + stereo_width = compute_stereo_width(pcm, frame_size, st->Fs, &st->width_mem); + else + stereo_width = 0; + total_buffer = delay_compensation; st->bitrate_bps = user_bitrate_to_bitrate(st, frame_size, max_data_bytes); frame_rate = st->Fs/frame_size; @@ -533,15 +1063,19 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s max_rate = frame_rate*max_data_bytes*8; /* Equivalent 20-ms rate for mode/channel/bandwidth decisions */ - equiv_rate = st->bitrate_bps - 60*(st->Fs/frame_size - 50); + equiv_rate = st->bitrate_bps - (40*st->channels+20)*(st->Fs/frame_size - 50); if (st->signal_type == OPUS_SIGNAL_VOICE) voice_est = 127; else if (st->signal_type == OPUS_SIGNAL_MUSIC) voice_est = 0; else if (st->voice_ratio >= 0) + { voice_est = st->voice_ratio*327>>8; - else if (st->application == OPUS_APPLICATION_VOIP) + /* For AUDIO, never be more than 90% confident of having speech */ + if (st->application == OPUS_APPLICATION_AUDIO) + voice_est = IMIN(voice_est, 115); + } else if (st->application == OPUS_APPLICATION_VOIP) voice_est = 115; else voice_est = 48; @@ -561,15 +1095,16 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s opus_int32 stereo_threshold; stereo_threshold = stereo_music_threshold + ((voice_est*voice_est*(stereo_voice_threshold-stereo_music_threshold))>>14); if (st->stream_channels == 2) - stereo_threshold -= 4000; + stereo_threshold -= 1000; else - stereo_threshold += 4000; + stereo_threshold += 1000; st->stream_channels = (equiv_rate > stereo_threshold) ? 2 : 1; } else { st->stream_channels = st->channels; } #endif } + equiv_rate = st->bitrate_bps - (40*st->stream_channels+20)*(st->Fs/frame_size - 50); /* Mode selection depending on application and signal type */ if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) @@ -592,15 +1127,21 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s st->mode = MODE_SILK_ONLY; } #else - int chan; opus_int32 mode_voice, mode_music; opus_int32 threshold; - chan = (st->channels==2) && st->force_channels!=1; - mode_voice = mode_thresholds[chan][0]; - mode_music = mode_thresholds[chan][1]; + /* Interpolate based on stereo width */ + mode_voice = (opus_int32)(MULT16_32_Q15(Q15ONE-stereo_width,mode_thresholds[0][0]) + + MULT16_32_Q15(stereo_width,mode_thresholds[1][0])); + mode_music = (opus_int32)(MULT16_32_Q15(Q15ONE-stereo_width,mode_thresholds[1][1]) + + MULT16_32_Q15(stereo_width,mode_thresholds[1][1])); + /* Interpolate based on speech/music probability */ threshold = mode_music + ((voice_est*voice_est*(mode_voice-mode_music))>>14); + /* Bias towards SILK for VoIP because of some useful features */ + if (st->application == OPUS_APPLICATION_VOIP) + threshold += 8000; + /*printf("%f %d\n", stereo_width/(float)Q15ONE, threshold);*/ /* Hysteresis */ if (st->prev_mode == MODE_CELT_ONLY) threshold -= 4000; @@ -623,6 +1164,11 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s /* Override the chosen mode to make sure we meet the requested frame size */ if (st->mode != MODE_CELT_ONLY && frame_size < st->Fs/100) st->mode = MODE_CELT_ONLY; + if (st->lfe) + st->mode = MODE_CELT_ONLY; + /* If max_data_bytes represents less than 8 kb/s, switch to CELT-only mode */ + if (max_data_bytes < (frame_rate > 50 ? 12000 : 8000)*frame_size / (st->Fs * 8)) + st->mode = MODE_CELT_ONLY; if (st->stream_channels == 1 && st->prev_channels ==2 && st->silk_mode.toMono==0 && st->mode != MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY) @@ -658,6 +1204,7 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s redundancy = 1; celt_to_silk = 1; st->silk_bw_switch = 0; + prefill=1; } if (redundancy) @@ -672,7 +1219,7 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s if (st->mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) { silk_EncControlStruct dummy; - silk_InitEncoder( silk_enc, &dummy); + silk_InitEncoder( silk_enc, st->arch, &dummy); prefill=1; } @@ -750,14 +1297,37 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s st->bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; if (st->Fs <= 8000 && st->bandwidth > OPUS_BANDWIDTH_NARROWBAND) st->bandwidth = OPUS_BANDWIDTH_NARROWBAND; +#ifndef DISABLE_FLOAT_API + /* Use detected bandwidth to reduce the encoded bandwidth. */ + if (st->detected_bandwidth && st->user_bandwidth == OPUS_AUTO) + { + int min_detected_bandwidth; + /* Makes bandwidth detection more conservative just in case the detector + gets it wrong when we could have coded a high bandwidth transparently. + When operating in SILK/hybrid mode, we don't go below wideband to avoid + more complicated switches that require redundancy. */ + if (equiv_rate <= 18000*st->stream_channels && st->mode == MODE_CELT_ONLY) + min_detected_bandwidth = OPUS_BANDWIDTH_NARROWBAND; + else if (equiv_rate <= 24000*st->stream_channels && st->mode == MODE_CELT_ONLY) + min_detected_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + else if (equiv_rate <= 30000*st->stream_channels) + min_detected_bandwidth = OPUS_BANDWIDTH_WIDEBAND; + else if (equiv_rate <= 44000*st->stream_channels) + min_detected_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; + else + min_detected_bandwidth = OPUS_BANDWIDTH_FULLBAND; - /* If max_data_bytes represents less than 8 kb/s, switch to CELT-only mode */ - if (max_data_bytes < (frame_rate > 50 ? 12000 : 8000)*frame_size / (st->Fs * 8)) - st->mode = MODE_CELT_ONLY; + st->detected_bandwidth = IMAX(st->detected_bandwidth, min_detected_bandwidth); + st->bandwidth = IMIN(st->bandwidth, st->detected_bandwidth); + } +#endif + celt_encoder_ctl(celt_enc, OPUS_SET_LSB_DEPTH(lsb_depth)); /* CELT mode doesn't support mediumband, use wideband instead */ if (st->mode == MODE_CELT_ONLY && st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) st->bandwidth = OPUS_BANDWIDTH_WIDEBAND; + if (st->lfe) + st->bandwidth = OPUS_BANDWIDTH_NARROWBAND; /* Can't support higher than wideband for >20 ms frames */ if (frame_size > st->Fs/50 && (st->mode == MODE_CELT_ONLY || st->bandwidth > OPUS_BANDWIDTH_WIDEBAND)) @@ -765,16 +1335,25 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s VARDECL(unsigned char, tmp_data); int nb_frames; int bak_mode, bak_bandwidth, bak_channels, bak_to_mono; - OpusRepacketizer rp; + VARDECL(OpusRepacketizer, rp); opus_int32 bytes_per_frame; + opus_int32 repacketize_len; +#ifndef DISABLE_FLOAT_API + if (analysis_read_pos_bak!= -1) + { + st->analysis.read_pos = analysis_read_pos_bak; + st->analysis.read_subframe = analysis_read_subframe_bak; + } +#endif nb_frames = frame_size > st->Fs/25 ? 3 : 2; bytes_per_frame = IMIN(1276,(out_data_bytes-3)/nb_frames); ALLOC(tmp_data, nb_frames*bytes_per_frame, unsigned char); - opus_repacketizer_init(&rp); + ALLOC(rp, 1, OpusRepacketizer); + opus_repacketizer_init(rp); bak_mode = st->user_forced_mode; bak_bandwidth = st->user_bandwidth; @@ -796,20 +1375,26 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s /* When switching from SILK/Hybrid to CELT, only ask for a switch at the last frame */ if (to_celt && i==nb_frames-1) st->user_forced_mode = MODE_CELT_ONLY; - tmp_len = opus_encode_native(st, pcm+i*(st->channels*st->Fs/50), st->Fs/50, tmp_data+i*bytes_per_frame, bytes_per_frame); + tmp_len = opus_encode_native(st, pcm+i*(st->channels*st->Fs/50), st->Fs/50, + tmp_data+i*bytes_per_frame, bytes_per_frame, lsb_depth, + NULL, 0, c1, c2, analysis_channels, downmix); if (tmp_len<0) { RESTORE_STACK; return OPUS_INTERNAL_ERROR; } - ret = opus_repacketizer_cat(&rp, tmp_data+i*bytes_per_frame, tmp_len); + ret = opus_repacketizer_cat(rp, tmp_data+i*bytes_per_frame, tmp_len); if (ret<0) { RESTORE_STACK; return OPUS_INTERNAL_ERROR; } } - ret = opus_repacketizer_out(&rp, data, out_data_bytes); + if (st->use_vbr) + repacketize_len = out_data_bytes; + else + repacketize_len = IMIN(3*st->bitrate_bps/(3*8*50/nb_frames), out_data_bytes); + ret = opus_repacketizer_out_range_impl(rp, 0, nb_frames, data, repacketize_len, 0, !st->use_vbr); if (ret<0) { RESTORE_STACK; @@ -822,7 +1407,6 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s RESTORE_STACK; return ret; } - curr_bandwidth = st->bandwidth; /* Chooses the appropriate mode for speech @@ -839,9 +1423,9 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s ec_enc_init(&enc, data, max_data_bytes-1); - ALLOC(pcm_buf, (delay_compensation+frame_size)*st->channels, opus_val16); - for (i=0;ichannels;i++) - pcm_buf[i] = st->delay_buffer[(st->encoder_buffer-delay_compensation)*st->channels+i]; + ALLOC(pcm_buf, (total_buffer+frame_size)*st->channels, opus_val16); + for (i=0;ichannels;i++) + pcm_buf[i] = st->delay_buffer[(st->encoder_buffer-total_buffer)*st->channels+i]; if (st->mode == MODE_CELT_ONLY) hp_freq_smth1 = silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 ); @@ -856,46 +1440,96 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s if (st->application == OPUS_APPLICATION_VOIP) { - hp_cutoff(pcm, cutoff_Hz, &pcm_buf[delay_compensation*st->channels], st->hp_mem, frame_size, st->channels, st->Fs); + hp_cutoff(pcm, cutoff_Hz, &pcm_buf[total_buffer*st->channels], st->hp_mem, frame_size, st->channels, st->Fs); } else { - for (i=0;ichannels;i++) - pcm_buf[delay_compensation*st->channels + i] = pcm[i]; + dc_reject(pcm, 3, &pcm_buf[total_buffer*st->channels], st->hp_mem, frame_size, st->channels, st->Fs); } + + /* SILK processing */ + HB_gain = Q15ONE; if (st->mode != MODE_CELT_ONLY) { + opus_int32 total_bitRate, celt_rate; #ifdef FIXED_POINT const opus_int16 *pcm_silk; #else VARDECL(opus_int16, pcm_silk); ALLOC(pcm_silk, st->channels*frame_size, opus_int16); #endif - st->silk_mode.bitRate = 8*bytes_target*frame_rate; + + /* Distribute bits between SILK and CELT */ + total_bitRate = 8 * bytes_target * frame_rate; if( st->mode == MODE_HYBRID ) { - st->silk_mode.bitRate /= st->stream_channels; + int HB_gain_ref; + /* Base rate for SILK */ + st->silk_mode.bitRate = st->stream_channels * ( 5000 + 1000 * ( st->Fs == 100 * frame_size ) ); if( curr_bandwidth == OPUS_BANDWIDTH_SUPERWIDEBAND ) { - if( st->Fs == 100 * frame_size ) { - /* 24 kHz, 10 ms */ - st->silk_mode.bitRate = ( ( st->silk_mode.bitRate + 2000 + st->use_vbr * 1000 ) * 2 ) / 3; - } else { - /* 24 kHz, 20 ms */ - st->silk_mode.bitRate = ( ( st->silk_mode.bitRate + 1000 + st->use_vbr * 1000 ) * 2 ) / 3; - } - } else { - if( st->Fs == 100 * frame_size ) { - /* 48 kHz, 10 ms */ - st->silk_mode.bitRate = ( st->silk_mode.bitRate + 8000 + st->use_vbr * 3000 ) / 2; - } else { - /* 48 kHz, 20 ms */ - st->silk_mode.bitRate = ( st->silk_mode.bitRate + 9000 + st->use_vbr * 1000 ) / 2; - } + /* SILK gets 2/3 of the remaining bits */ + st->silk_mode.bitRate += ( total_bitRate - st->silk_mode.bitRate ) * 2 / 3; + } else { /* FULLBAND */ + /* SILK gets 3/5 of the remaining bits */ + st->silk_mode.bitRate += ( total_bitRate - st->silk_mode.bitRate ) * 3 / 5; } - st->silk_mode.bitRate *= st->stream_channels; - /* don't let SILK use more than 80% */ - if( st->silk_mode.bitRate > ( st->bitrate_bps - 8*st->Fs/frame_size ) * 4/5 ) { - st->silk_mode.bitRate = ( st->bitrate_bps - 8*st->Fs/frame_size ) * 4/5; + /* Don't let SILK use more than 80% */ + if( st->silk_mode.bitRate > total_bitRate * 4/5 ) { + st->silk_mode.bitRate = total_bitRate * 4/5; } + if (!st->energy_masking) + { + /* Increasingly attenuate high band when it gets allocated fewer bits */ + celt_rate = total_bitRate - st->silk_mode.bitRate; + HB_gain_ref = (curr_bandwidth == OPUS_BANDWIDTH_SUPERWIDEBAND) ? 3000 : 3600; + HB_gain = SHL32((opus_val32)celt_rate, 9) / SHR32((opus_val32)celt_rate + st->stream_channels * HB_gain_ref, 6); + HB_gain = HB_gain < Q15ONE*6/7 ? HB_gain + Q15ONE/7 : Q15ONE; + } + } else { + /* SILK gets all bits */ + st->silk_mode.bitRate = total_bitRate; + } + + /* Surround masking for SILK */ + if (st->energy_masking && st->use_vbr && !st->lfe) + { + opus_val32 mask_sum=0; + opus_val16 masking_depth; + opus_int32 rate_offset; + int c; + int end = 17; + opus_int16 srate = 16000; + if (st->bandwidth == OPUS_BANDWIDTH_NARROWBAND) + { + end = 13; + srate = 8000; + } else if (st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) + { + end = 15; + srate = 12000; + } + for (c=0;cchannels;c++) + { + for(i=0;ienergy_masking[21*c+i], + QCONST16(.5f, DB_SHIFT)), -QCONST16(2.0f, DB_SHIFT)); + if (mask > 0) + mask = HALF16(mask); + mask_sum += mask; + } + } + /* Conservative rate reduction, we cut the masking in half */ + masking_depth = mask_sum / end*st->channels; + masking_depth += QCONST16(.2f, DB_SHIFT); + rate_offset = (opus_int32)PSHR32(MULT16_16(srate, masking_depth), DB_SHIFT); + rate_offset = MAX32(rate_offset, -2*st->silk_mode.bitRate/3); + /* Split the rate change between the SILK and CELT part for hybrid. */ + if (st->bandwidth==OPUS_BANDWIDTH_SUPERWIDEBAND || st->bandwidth==OPUS_BANDWIDTH_FULLBAND) + st->silk_mode.bitRate += 3*rate_offset/5; + else + st->silk_mode.bitRate += rate_offset; + bytes_target += rate_offset * frame_size / (8 * st->Fs); } st->silk_mode.payloadSize_ms = 1000 * frame_size / st->Fs; @@ -955,6 +1589,18 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s if (prefill) { opus_int32 zero=0; + int prefill_offset; + /* Use a smooth onset for the SILK prefill to avoid the encoder trying to encode + a discontinuity. The exact location is what we need to avoid leaving any "gap" + in the audio when mixing with the redundant CELT frame. Here we can afford to + overwrite st->delay_buffer because the only thing that uses it before it gets + rewritten is tmp_prefill[] and even then only the part after the ramp really + gets used (rather than sent to the encoder and discarded) */ + prefill_offset = st->channels*(st->encoder_buffer-st->delay_compensation-st->Fs/400); + gain_fade(st->delay_buffer+prefill_offset, st->delay_buffer+prefill_offset, + 0, Q15ONE, celt_mode->overlap, st->Fs/400, st->channels, celt_mode->window, st->Fs); + for(i=0;idelay_buffer[i]=0; #ifdef FIXED_POINT pcm_silk = st->delay_buffer; #else @@ -965,10 +1611,10 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s } #ifdef FIXED_POINT - pcm_silk = pcm_buf+delay_compensation*st->channels; + pcm_silk = pcm_buf+total_buffer*st->channels; #else for (i=0;ichannels;i++) - pcm_silk[i] = FLOAT2INT16(pcm_buf[delay_compensation*st->channels + i]); + pcm_silk[i] = FLOAT2INT16(pcm_buf[total_buffer*st->channels + i]); #endif ret = silk_Encode( silk_enc, &st->silk_mode, pcm_silk, frame_size, &enc, &nBytes, 0 ); if( ret ) { @@ -1033,9 +1679,12 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(OPUS_BITRATE_MAX)); if (st->mode != MODE_SILK_ONLY) { + opus_val32 celt_pred=2; celt_encoder_ctl(celt_enc, OPUS_SET_VBR(0)); - /* Allow prediction unless we decide to disable it later */ - celt_encoder_ctl(celt_enc, CELT_SET_PREDICTION(2)); + /* We may still decide to disable prediction later */ + if (st->silk_mode.reducedDependency) + celt_pred = 0; + celt_encoder_ctl(celt_enc, CELT_SET_PREDICTION(celt_pred)); if (st->mode == MODE_HYBRID) { @@ -1053,9 +1702,18 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s } else { if (st->use_vbr) { + opus_int32 bonus=0; +#ifndef DISABLE_FLOAT_API + if (st->variable_duration==OPUS_FRAMESIZE_VARIABLE && frame_size != st->Fs/50) + { + bonus = (60*st->stream_channels+40)*(st->Fs/frame_size-50); + if (analysis_info.valid) + bonus = (opus_int32)(bonus*(1.f+.5f*analysis_info.tonality)); + } +#endif celt_encoder_ctl(celt_enc, OPUS_SET_VBR(1)); celt_encoder_ctl(celt_enc, OPUS_SET_VBR_CONSTRAINT(st->vbr_constraint)); - celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(st->bitrate_bps)); + celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(st->bitrate_bps+bonus)); nb_compr_bytes = max_data_bytes-1-redundancy_bytes; } else { nb_compr_bytes = bytes_target; @@ -1070,24 +1728,27 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s if (st->mode != MODE_SILK_ONLY && st->mode != st->prev_mode && st->prev_mode > 0) { for (i=0;ichannels*st->Fs/400;i++) - tmp_prefill[i] = st->delay_buffer[(st->encoder_buffer-st->delay_compensation-st->Fs/400)*st->channels + i]; + tmp_prefill[i] = st->delay_buffer[(st->encoder_buffer-total_buffer-st->Fs/400)*st->channels + i]; } - for (i=0;ichannels*(st->encoder_buffer-(frame_size+delay_compensation));i++) + for (i=0;ichannels*(st->encoder_buffer-(frame_size+total_buffer));i++) st->delay_buffer[i] = st->delay_buffer[i+st->channels*frame_size]; for (;iencoder_buffer*st->channels;i++) - st->delay_buffer[i] = pcm_buf[(frame_size+delay_compensation-st->encoder_buffer)*st->channels+i]; - + st->delay_buffer[i] = pcm_buf[(frame_size+total_buffer-st->encoder_buffer)*st->channels+i]; + /* gain_fade() and stereo_fade() need to be after the buffer copying + because we don't want any of this to affect the SILK part */ + if( st->prev_HB_gain < Q15ONE || HB_gain < Q15ONE ) { + gain_fade(pcm_buf, pcm_buf, + st->prev_HB_gain, HB_gain, celt_mode->overlap, frame_size, st->channels, celt_mode->window, st->Fs); + } + st->prev_HB_gain = HB_gain; if (st->mode != MODE_HYBRID || st->stream_channels==1) - st->silk_mode.stereoWidth_Q14 = 1<<14; - if( st->channels == 2 ) { + st->silk_mode.stereoWidth_Q14 = IMIN((1<<14),2*IMAX(0,equiv_rate-30000)); + if( !st->energy_masking && st->channels == 2 ) { /* Apply stereo width reduction (at low bitrates) */ if( st->hybrid_stereo_width_Q14 < (1 << 14) || st->silk_mode.stereoWidth_Q14 < (1 << 14) ) { opus_val16 g1, g2; - const CELTMode *celt_mode; - - celt_encoder_ctl(celt_enc, CELT_GET_MODE(&celt_mode)); g1 = st->hybrid_stereo_width_Q14; g2 = (opus_val16)(st->silk_mode.stereoWidth_Q14); #ifdef FIXED_POINT @@ -1113,7 +1774,7 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s int max_redundancy; ec_enc_bit_logp(&enc, celt_to_silk, 1); if (st->mode == MODE_HYBRID) - max_redundancy = (max_data_bytes-1)-nb_compr_bytes-1; + max_redundancy = (max_data_bytes-1)-nb_compr_bytes; else max_redundancy = (max_data_bytes-1)-((ec_tell(&enc)+7)>>3); /* Target the same bit-rate for redundancy as for the rest, @@ -1144,6 +1805,10 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s ec_enc_shrink(&enc, nb_compr_bytes); } +#ifndef DISABLE_FLOAT_API + if (redundancy || st->mode != MODE_SILK_ONLY) + celt_encoder_ctl(celt_enc, CELT_SET_ANALYSIS(&analysis_info)); +#endif /* 5 ms redundant frame for CELT->SILK */ if (redundancy && celt_to_silk) @@ -1252,9 +1917,10 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s } /* Count ToC and redundancy */ ret += 1+redundancy_bytes; - if (!st->use_vbr && ret >= 3) + if (!st->use_vbr) { - if (pad_frame(data, ret, max_data_bytes)) + if (opus_packet_pad(data, ret, max_data_bytes) != OPUS_OK) + { RESTORE_STACK; return OPUS_INTERNAL_ERROR; @@ -1268,45 +1934,93 @@ opus_int32 opus_encode_float(OpusEncoder *st, const opus_val16 *pcm, int frame_s #ifdef FIXED_POINT #ifndef DISABLE_FLOAT_API -opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int frame_size, +opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int analysis_frame_size, unsigned char *data, opus_int32 max_data_bytes) { int i, ret; + int frame_size; + int delay_compensation; VARDECL(opus_int16, in); ALLOC_STACK; - if(frame_size<0) - { - RESTORE_STACK; - return OPUS_BAD_ARG; - } + if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) + delay_compensation = 0; + else + delay_compensation = st->delay_compensation; + frame_size = compute_frame_size(pcm, analysis_frame_size, + st->variable_duration, st->channels, st->Fs, st->bitrate_bps, + delay_compensation, downmix_float, st->analysis.subframe_mem); ALLOC(in, frame_size*st->channels, opus_int16); for (i=0;ichannels;i++) in[i] = FLOAT2INT16(pcm[i]); - ret = opus_encode(st, in, frame_size, data, max_data_bytes); + ret = opus_encode_native(st, in, frame_size, data, max_data_bytes, 16, pcm, analysis_frame_size, 0, -2, st->channels, downmix_float); RESTORE_STACK; return ret; } #endif +opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int analysis_frame_size, + unsigned char *data, opus_int32 out_data_bytes) +{ + int frame_size; + int delay_compensation; + if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) + delay_compensation = 0; + else + delay_compensation = st->delay_compensation; + frame_size = compute_frame_size(pcm, analysis_frame_size, + st->variable_duration, st->channels, st->Fs, st->bitrate_bps, + delay_compensation, downmix_int +#ifndef DISABLE_FLOAT_API + , st->analysis.subframe_mem +#endif + ); + return opus_encode_native(st, pcm, frame_size, data, out_data_bytes, 16, pcm, analysis_frame_size, 0, -2, st->channels, downmix_int); +} + #else -opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int frame_size, +opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int analysis_frame_size, unsigned char *data, opus_int32 max_data_bytes) { int i, ret; + int frame_size; + int delay_compensation; VARDECL(float, in); ALLOC_STACK; + if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) + delay_compensation = 0; + else + delay_compensation = st->delay_compensation; + frame_size = compute_frame_size(pcm, analysis_frame_size, + st->variable_duration, st->channels, st->Fs, st->bitrate_bps, + delay_compensation, downmix_int, st->analysis.subframe_mem); + ALLOC(in, frame_size*st->channels, float); for (i=0;ichannels;i++) in[i] = (1.0f/32768)*pcm[i]; - ret = opus_encode_float(st, in, frame_size, data, max_data_bytes); + ret = opus_encode_native(st, in, frame_size, data, max_data_bytes, 16, pcm, analysis_frame_size, 0, -2, st->channels, downmix_int); RESTORE_STACK; return ret; } +opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int analysis_frame_size, + unsigned char *data, opus_int32 out_data_bytes) +{ + int frame_size; + int delay_compensation; + if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) + delay_compensation = 0; + else + delay_compensation = st->delay_compensation; + frame_size = compute_frame_size(pcm, analysis_frame_size, + st->variable_duration, st->channels, st->Fs, st->bitrate_bps, + delay_compensation, downmix_float, st->analysis.subframe_mem); + return opus_encode_native(st, pcm, frame_size, data, out_data_bytes, 24, + pcm, analysis_frame_size, 0, -2, st->channels, downmix_float); +} #endif @@ -1339,6 +2053,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_APPLICATION_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->application; } break; @@ -1360,6 +2078,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_BITRATE_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = user_bitrate_to_bitrate(st, st->prev_framesize, 1276); } break; @@ -1367,21 +2089,29 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if((value<1 || value>st->channels) && value != OPUS_AUTO) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->force_channels = value; } break; case OPUS_GET_FORCE_CHANNELS_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->force_channels; } break; case OPUS_SET_MAX_BANDWIDTH_REQUEST: { opus_int32 value = va_arg(ap, opus_int32); - if (value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND) - return OPUS_BAD_ARG; + if (value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND) + { + goto bad_arg; + } st->max_bandwidth = value; if (st->max_bandwidth == OPUS_BANDWIDTH_NARROWBAND) { st->silk_mode.maxInternalSampleRate = 8000; @@ -1395,6 +2125,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_MAX_BANDWIDTH_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->max_bandwidth; } break; @@ -1402,7 +2136,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if ((value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND) && value != OPUS_AUTO) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->user_bandwidth = value; if (st->user_bandwidth == OPUS_BANDWIDTH_NARROWBAND) { st->silk_mode.maxInternalSampleRate = 8000; @@ -1416,6 +2152,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_BANDWIDTH_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->bandwidth; } break; @@ -1423,13 +2163,19 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if(value<0 || value>1) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->silk_mode.useDTX = value; } break; case OPUS_GET_DTX_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->silk_mode.useDTX; } break; @@ -1437,7 +2183,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if(value<0 || value>10) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->silk_mode.complexity = value; celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(value)); } @@ -1445,6 +2193,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_COMPLEXITY_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->silk_mode.complexity; } break; @@ -1452,13 +2204,19 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if(value<0 || value>1) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->silk_mode.useInBandFEC = value; } break; case OPUS_GET_INBAND_FEC_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->silk_mode.useInBandFEC; } break; @@ -1466,7 +2224,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if (value < 0 || value > 100) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->silk_mode.packetLossPercentage = value; celt_encoder_ctl(celt_enc, OPUS_SET_PACKET_LOSS_PERC(value)); } @@ -1474,6 +2234,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_PACKET_LOSS_PERC_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->silk_mode.packetLossPercentage; } break; @@ -1481,7 +2245,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if(value<0 || value>1) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->use_vbr = value; st->silk_mode.useCBR = 1-value; } @@ -1489,20 +2255,30 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_VBR_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->use_vbr; } break; case OPUS_SET_VOICE_RATIO_REQUEST: { opus_int32 value = va_arg(ap, opus_int32); - if (value>100 || value<-1) - goto bad_arg; + if (value<-1 || value>100) + { + goto bad_arg; + } st->voice_ratio = value; } break; case OPUS_GET_VOICE_RATIO_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->voice_ratio; } break; @@ -1510,13 +2286,19 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if(value<0 || value>1) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->vbr_constraint = value; } break; case OPUS_GET_VBR_CONSTRAINT_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->vbr_constraint; } break; @@ -1524,19 +2306,29 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if(value!=OPUS_AUTO && value!=OPUS_SIGNAL_VOICE && value!=OPUS_SIGNAL_MUSIC) - return OPUS_BAD_ARG; + { + goto bad_arg; + } st->signal_type = value; } break; case OPUS_GET_SIGNAL_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->signal_type; } break; case OPUS_GET_LOOKAHEAD_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } *value = st->Fs/400; if (st->application != OPUS_APPLICATION_RESTRICTED_LOWDELAY) *value += st->delay_compensation; @@ -1545,10 +2337,9 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_SAMPLE_RATE_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); - if (value==NULL) + if (!value) { - ret = OPUS_BAD_ARG; - break; + goto bad_arg; } *value = st->Fs; } @@ -1556,19 +2347,71 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) case OPUS_GET_FINAL_RANGE_REQUEST: { opus_uint32 *value = va_arg(ap, opus_uint32*); + if (!value) + { + goto bad_arg; + } *value = st->rangeFinal; } break; case OPUS_SET_LSB_DEPTH_REQUEST: { opus_int32 value = va_arg(ap, opus_int32); - ret = celt_encoder_ctl(celt_enc, OPUS_SET_LSB_DEPTH(value)); + if (value<8 || value>24) + { + goto bad_arg; + } + st->lsb_depth=value; } break; case OPUS_GET_LSB_DEPTH_REQUEST: { opus_int32 *value = va_arg(ap, opus_int32*); - celt_encoder_ctl(celt_enc, OPUS_GET_LSB_DEPTH(value)); + if (!value) + { + goto bad_arg; + } + *value = st->lsb_depth; + } + break; + case OPUS_SET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value != OPUS_FRAMESIZE_ARG && value != OPUS_FRAMESIZE_2_5_MS && + value != OPUS_FRAMESIZE_5_MS && value != OPUS_FRAMESIZE_10_MS && + value != OPUS_FRAMESIZE_20_MS && value != OPUS_FRAMESIZE_40_MS && + value != OPUS_FRAMESIZE_60_MS && value != OPUS_FRAMESIZE_VARIABLE) + { + goto bad_arg; + } + st->variable_duration = value; + celt_encoder_ctl(celt_enc, OPUS_SET_EXPERT_FRAME_DURATION(value)); + } + break; + case OPUS_GET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->variable_duration; + } + break; + case OPUS_SET_PREDICTION_DISABLED_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value > 1 || value < 0) + goto bad_arg; + st->silk_mode.reducedDependency = value; + } + break; + case OPUS_GET_PREDICTION_DISABLED_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + goto bad_arg; + *value = st->silk_mode.reducedDependency; } break; case OPUS_RESET_STATE: @@ -1582,9 +2425,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) ((char*)&st->OPUS_ENCODER_RESET_START - (char*)st)); celt_encoder_ctl(celt_enc, OPUS_RESET_STATE); - silk_InitEncoder( silk_enc, &dummy ); + silk_InitEncoder( silk_enc, st->arch, &dummy ); st->stream_channels = st->channels; st->hybrid_stereo_width_Q14 = 1 << 14; + st->prev_HB_gain = Q15ONE; st->first = 1; st->mode = MODE_HYBRID; st->bandwidth = OPUS_BANDWIDTH_FULLBAND; @@ -1595,10 +2439,37 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...) { opus_int32 value = va_arg(ap, opus_int32); if ((value < MODE_SILK_ONLY || value > MODE_CELT_ONLY) && value != OPUS_AUTO) + { goto bad_arg; + } st->user_forced_mode = value; } break; + case OPUS_SET_LFE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->lfe = value; + ret = celt_encoder_ctl(celt_enc, OPUS_SET_LFE(value)); + } + break; + case OPUS_SET_ENERGY_MASK_REQUEST: + { + opus_val16 *value = va_arg(ap, opus_val16*); + st->energy_masking = value; + ret = celt_encoder_ctl(celt_enc, OPUS_SET_ENERGY_MASK(value)); + } + break; + + case CELT_GET_MODE_REQUEST: + { + const CELTMode ** value = va_arg(ap, const CELTMode**); + if (!value) + { + goto bad_arg; + } + ret = celt_encoder_ctl(celt_enc, CELT_GET_MODE(value)); + } + break; default: /* fprintf(stderr, "unknown opus_encoder_ctl() request: %d", request);*/ ret = OPUS_UNIMPLEMENTED; diff --git a/code/opus-1.1/src/opus_multistream.c b/code/opus-1.1/src/opus_multistream.c new file mode 100644 index 00000000..09c3639b --- /dev/null +++ b/code/opus-1.1/src/opus_multistream.c @@ -0,0 +1,92 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_multistream.h" +#include "opus.h" +#include "opus_private.h" +#include "stack_alloc.h" +#include +#include "float_cast.h" +#include "os_support.h" + + +int validate_layout(const ChannelLayout *layout) +{ + int i, max_channel; + + max_channel = layout->nb_streams+layout->nb_coupled_streams; + if (max_channel>255) + return 0; + for (i=0;inb_channels;i++) + { + if (layout->mapping[i] >= max_channel && layout->mapping[i] != 255) + return 0; + } + return 1; +} + + +int get_left_channel(const ChannelLayout *layout, int stream_id, int prev) +{ + int i; + i = (prev<0) ? 0 : prev+1; + for (;inb_channels;i++) + { + if (layout->mapping[i]==stream_id*2) + return i; + } + return -1; +} + +int get_right_channel(const ChannelLayout *layout, int stream_id, int prev) +{ + int i; + i = (prev<0) ? 0 : prev+1; + for (;inb_channels;i++) + { + if (layout->mapping[i]==stream_id*2+1) + return i; + } + return -1; +} + +int get_mono_channel(const ChannelLayout *layout, int stream_id, int prev) +{ + int i; + i = (prev<0) ? 0 : prev+1; + for (;inb_channels;i++) + { + if (layout->mapping[i]==stream_id+layout->nb_coupled_streams) + return i; + } + return -1; +} + diff --git a/code/opus-1.1/src/opus_multistream_decoder.c b/code/opus-1.1/src/opus_multistream_decoder.c new file mode 100644 index 00000000..a05fa1e7 --- /dev/null +++ b/code/opus-1.1/src/opus_multistream_decoder.c @@ -0,0 +1,537 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_multistream.h" +#include "opus.h" +#include "opus_private.h" +#include "stack_alloc.h" +#include +#include "float_cast.h" +#include "os_support.h" + +struct OpusMSDecoder { + ChannelLayout layout; + /* Decoder states go here */ +}; + + + + +/* DECODER */ + +opus_int32 opus_multistream_decoder_get_size(int nb_streams, int nb_coupled_streams) +{ + int coupled_size; + int mono_size; + + if(nb_streams<1||nb_coupled_streams>nb_streams||nb_coupled_streams<0)return 0; + coupled_size = opus_decoder_get_size(2); + mono_size = opus_decoder_get_size(1); + return align(sizeof(OpusMSDecoder)) + + nb_coupled_streams * align(coupled_size) + + (nb_streams-nb_coupled_streams) * align(mono_size); +} + +int opus_multistream_decoder_init( + OpusMSDecoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping +) +{ + int coupled_size; + int mono_size; + int i, ret; + char *ptr; + + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) + return OPUS_BAD_ARG; + + st->layout.nb_channels = channels; + st->layout.nb_streams = streams; + st->layout.nb_coupled_streams = coupled_streams; + + for (i=0;ilayout.nb_channels;i++) + st->layout.mapping[i] = mapping[i]; + if (!validate_layout(&st->layout)) + return OPUS_BAD_ARG; + + ptr = (char*)st + align(sizeof(OpusMSDecoder)); + coupled_size = opus_decoder_get_size(2); + mono_size = opus_decoder_get_size(1); + + for (i=0;ilayout.nb_coupled_streams;i++) + { + ret=opus_decoder_init((OpusDecoder*)ptr, Fs, 2); + if(ret!=OPUS_OK)return ret; + ptr += align(coupled_size); + } + for (;ilayout.nb_streams;i++) + { + ret=opus_decoder_init((OpusDecoder*)ptr, Fs, 1); + if(ret!=OPUS_OK)return ret; + ptr += align(mono_size); + } + return OPUS_OK; +} + + +OpusMSDecoder *opus_multistream_decoder_create( + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int *error +) +{ + int ret; + OpusMSDecoder *st; + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + st = (OpusMSDecoder *)opus_alloc(opus_multistream_decoder_get_size(streams, coupled_streams)); + if (st==NULL) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + ret = opus_multistream_decoder_init(st, Fs, channels, streams, coupled_streams, mapping); + if (error) + *error = ret; + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + return st; +} + +typedef void (*opus_copy_channel_out_func)( + void *dst, + int dst_stride, + int dst_channel, + const opus_val16 *src, + int src_stride, + int frame_size +); + +static int opus_multistream_packet_validate(const unsigned char *data, + opus_int32 len, int nb_streams, opus_int32 Fs) +{ + int s; + int count; + unsigned char toc; + opus_int16 size[48]; + int samples=0; + opus_int32 packet_offset; + + for (s=0;slayout.nb_streams-1) + { + RESTORE_STACK; + return OPUS_INVALID_PACKET; + } + if (!do_plc) + { + int ret = opus_multistream_packet_validate(data, len, st->layout.nb_streams, Fs); + if (ret < 0) + { + RESTORE_STACK; + return ret; + } else if (ret > frame_size) + { + RESTORE_STACK; + return OPUS_BUFFER_TOO_SMALL; + } + } + for (s=0;slayout.nb_streams;s++) + { + OpusDecoder *dec; + int packet_offset, ret; + + dec = (OpusDecoder*)ptr; + ptr += (s < st->layout.nb_coupled_streams) ? align(coupled_size) : align(mono_size); + + if (!do_plc && len<=0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + packet_offset = 0; + ret = opus_decode_native(dec, data, len, buf, frame_size, decode_fec, s!=st->layout.nb_streams-1, &packet_offset, soft_clip); + data += packet_offset; + len -= packet_offset; + if (ret <= 0) + { + RESTORE_STACK; + return ret; + } + frame_size = ret; + if (s < st->layout.nb_coupled_streams) + { + int chan, prev; + prev = -1; + /* Copy "left" audio to the channel(s) where it belongs */ + while ( (chan = get_left_channel(&st->layout, s, prev)) != -1) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, chan, + buf, 2, frame_size); + prev = chan; + } + prev = -1; + /* Copy "right" audio to the channel(s) where it belongs */ + while ( (chan = get_right_channel(&st->layout, s, prev)) != -1) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, chan, + buf+1, 2, frame_size); + prev = chan; + } + } else { + int chan, prev; + prev = -1; + /* Copy audio to the channel(s) where it belongs */ + while ( (chan = get_mono_channel(&st->layout, s, prev)) != -1) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, chan, + buf, 1, frame_size); + prev = chan; + } + } + } + /* Handle muted channels */ + for (c=0;clayout.nb_channels;c++) + { + if (st->layout.mapping[c] == 255) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, c, + NULL, 0, frame_size); + } + } + RESTORE_STACK; + return frame_size; +} + +#if !defined(DISABLE_FLOAT_API) +static void opus_copy_channel_out_float( + void *dst, + int dst_stride, + int dst_channel, + const opus_val16 *src, + int src_stride, + int frame_size +) +{ + float *float_dst; + opus_int32 i; + float_dst = (float*)dst; + if (src != NULL) + { + for (i=0;ilayout.nb_streams;s++) + { + OpusDecoder *dec; + dec = (OpusDecoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_decoder_ctl(dec, request, &tmp); + if (ret != OPUS_OK) break; + *value ^= tmp; + } + } + break; + case OPUS_RESET_STATE: + { + int s; + for (s=0;slayout.nb_streams;s++) + { + OpusDecoder *dec; + + dec = (OpusDecoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_decoder_ctl(dec, OPUS_RESET_STATE); + if (ret != OPUS_OK) + break; + } + } + break; + case OPUS_MULTISTREAM_GET_DECODER_STATE_REQUEST: + { + int s; + opus_int32 stream_id; + OpusDecoder **value; + stream_id = va_arg(ap, opus_int32); + if (stream_id<0 || stream_id >= st->layout.nb_streams) + ret = OPUS_BAD_ARG; + value = va_arg(ap, OpusDecoder**); + if (!value) + { + goto bad_arg; + } + for (s=0;slayout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + *value = (OpusDecoder*)ptr; + } + break; + case OPUS_SET_GAIN_REQUEST: + { + int s; + /* This works for int32 params */ + opus_int32 value = va_arg(ap, opus_int32); + for (s=0;slayout.nb_streams;s++) + { + OpusDecoder *dec; + + dec = (OpusDecoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_decoder_ctl(dec, request, value); + if (ret != OPUS_OK) + break; + } + } + break; + default: + ret = OPUS_UNIMPLEMENTED; + break; + } + + va_end(ap); + return ret; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +} + + +void opus_multistream_decoder_destroy(OpusMSDecoder *st) +{ + opus_free(st); +} diff --git a/code/opus-1.1/src/opus_multistream_encoder.c b/code/opus-1.1/src/opus_multistream_encoder.c new file mode 100644 index 00000000..49e27913 --- /dev/null +++ b/code/opus-1.1/src/opus_multistream_encoder.c @@ -0,0 +1,1174 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_multistream.h" +#include "opus.h" +#include "opus_private.h" +#include "stack_alloc.h" +#include +#include "float_cast.h" +#include "os_support.h" +#include "mathops.h" +#include "mdct.h" +#include "modes.h" +#include "bands.h" +#include "quant_bands.h" + +typedef struct { + int nb_streams; + int nb_coupled_streams; + unsigned char mapping[8]; +} VorbisLayout; + +/* Index is nb_channel-1*/ +static const VorbisLayout vorbis_mappings[8] = { + {1, 0, {0}}, /* 1: mono */ + {1, 1, {0, 1}}, /* 2: stereo */ + {2, 1, {0, 2, 1}}, /* 3: 1-d surround */ + {2, 2, {0, 1, 2, 3}}, /* 4: quadraphonic surround */ + {3, 2, {0, 4, 1, 2, 3}}, /* 5: 5-channel surround */ + {4, 2, {0, 4, 1, 2, 3, 5}}, /* 6: 5.1 surround */ + {4, 3, {0, 4, 1, 2, 3, 5, 6}}, /* 7: 6.1 surround */ + {5, 3, {0, 6, 1, 2, 3, 4, 5, 7}}, /* 8: 7.1 surround */ +}; + +typedef void (*opus_copy_channel_in_func)( + opus_val16 *dst, + int dst_stride, + const void *src, + int src_stride, + int src_channel, + int frame_size +); + +struct OpusMSEncoder { + ChannelLayout layout; + int lfe_stream; + int application; + int variable_duration; + int surround; + opus_int32 bitrate_bps; + float subframe_mem[3]; + /* Encoder states go here */ + /* then opus_val32 window_mem[channels*120]; */ + /* then opus_val32 preemph_mem[channels]; */ +}; + +static opus_val32 *ms_get_preemph_mem(OpusMSEncoder *st) +{ + int s; + char *ptr; + int coupled_size, mono_size; + + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + for (s=0;slayout.nb_streams;s++) + { + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + return (opus_val32*)(ptr+st->layout.nb_channels*120*sizeof(opus_val32)); +} + +static opus_val32 *ms_get_window_mem(OpusMSEncoder *st) +{ + int s; + char *ptr; + int coupled_size, mono_size; + + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + for (s=0;slayout.nb_streams;s++) + { + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + return (opus_val32*)ptr; +} + +static int validate_encoder_layout(const ChannelLayout *layout) +{ + int s; + for (s=0;snb_streams;s++) + { + if (s < layout->nb_coupled_streams) + { + if (get_left_channel(layout, s, -1)==-1) + return 0; + if (get_right_channel(layout, s, -1)==-1) + return 0; + } else { + if (get_mono_channel(layout, s, -1)==-1) + return 0; + } + } + return 1; +} + +static void channel_pos(int channels, int pos[8]) +{ + /* Position in the mix: 0 don't mix, 1: left, 2: center, 3:right */ + if (channels==4) + { + pos[0]=1; + pos[1]=3; + pos[2]=1; + pos[3]=3; + } else if (channels==3||channels==5||channels==6) + { + pos[0]=1; + pos[1]=2; + pos[2]=3; + pos[3]=1; + pos[4]=3; + pos[5]=0; + } else if (channels==7) + { + pos[0]=1; + pos[1]=2; + pos[2]=3; + pos[3]=1; + pos[4]=3; + pos[5]=2; + pos[6]=0; + } else if (channels==8) + { + pos[0]=1; + pos[1]=2; + pos[2]=3; + pos[3]=1; + pos[4]=3; + pos[5]=1; + pos[6]=3; + pos[7]=0; + } +} + +#if 1 +/* Computes a rough approximation of log2(2^a + 2^b) */ +static opus_val16 logSum(opus_val16 a, opus_val16 b) +{ + opus_val16 max; + opus_val32 diff; + opus_val16 frac; + static const opus_val16 diff_table[17] = { + QCONST16(0.5000000f, DB_SHIFT), QCONST16(0.2924813f, DB_SHIFT), QCONST16(0.1609640f, DB_SHIFT), QCONST16(0.0849625f, DB_SHIFT), + QCONST16(0.0437314f, DB_SHIFT), QCONST16(0.0221971f, DB_SHIFT), QCONST16(0.0111839f, DB_SHIFT), QCONST16(0.0056136f, DB_SHIFT), + QCONST16(0.0028123f, DB_SHIFT) + }; + int low; + if (a>b) + { + max = a; + diff = SUB32(EXTEND32(a),EXTEND32(b)); + } else { + max = b; + diff = SUB32(EXTEND32(b),EXTEND32(a)); + } + if (diff >= QCONST16(8.f, DB_SHIFT)) + return max; +#ifdef FIXED_POINT + low = SHR32(diff, DB_SHIFT-1); + frac = SHL16(diff - SHL16(low, DB_SHIFT-1), 16-DB_SHIFT); +#else + low = (int)floor(2*diff); + frac = 2*diff - low; +#endif + return max + diff_table[low] + MULT16_16_Q15(frac, SUB16(diff_table[low+1], diff_table[low])); +} +#else +opus_val16 logSum(opus_val16 a, opus_val16 b) +{ + return log2(pow(4, a)+ pow(4, b))/2; +} +#endif + +void surround_analysis(const CELTMode *celt_mode, const void *pcm, opus_val16 *bandLogE, opus_val32 *mem, opus_val32 *preemph_mem, + int len, int overlap, int channels, int rate, opus_copy_channel_in_func copy_channel_in +) +{ + int c; + int i; + int LM; + int pos[8] = {0}; + int upsample; + int frame_size; + opus_val16 channel_offset; + opus_val32 bandE[21]; + opus_val16 maskLogE[3][21]; + VARDECL(opus_val32, in); + VARDECL(opus_val16, x); + VARDECL(opus_val32, freq); + SAVE_STACK; + + upsample = resampling_factor(rate); + frame_size = len*upsample; + + for (LM=0;LMmaxLM;LM++) + if (celt_mode->shortMdctSize<preemph, preemph_mem+c, 0); + clt_mdct_forward(&celt_mode->mdct, in, freq, celt_mode->window, overlap, celt_mode->maxLM-LM, 1); + if (upsample != 1) + { + int bound = len; + for (i=0;i=0;i--) + bandLogE[21*c+i] = MAX16(bandLogE[21*c+i], bandLogE[21*c+i+1]-QCONST16(2.f, DB_SHIFT)); + if (pos[c]==1) + { + for (i=0;i<21;i++) + maskLogE[0][i] = logSum(maskLogE[0][i], bandLogE[21*c+i]); + } else if (pos[c]==3) + { + for (i=0;i<21;i++) + maskLogE[2][i] = logSum(maskLogE[2][i], bandLogE[21*c+i]); + } else if (pos[c]==2) + { + for (i=0;i<21;i++) + { + maskLogE[0][i] = logSum(maskLogE[0][i], bandLogE[21*c+i]-QCONST16(.5f, DB_SHIFT)); + maskLogE[2][i] = logSum(maskLogE[2][i], bandLogE[21*c+i]-QCONST16(.5f, DB_SHIFT)); + } + } +#if 0 + for (i=0;i<21;i++) + printf("%f ", bandLogE[21*c+i]); + float sum=0; + for (i=0;i<21;i++) + sum += bandLogE[21*c+i]; + printf("%f ", sum/21); +#endif + OPUS_COPY(mem+c*overlap, in+frame_size, overlap); + } + for (i=0;i<21;i++) + maskLogE[1][i] = MIN32(maskLogE[0][i],maskLogE[2][i]); + channel_offset = HALF16(celt_log2(QCONST32(2.f,14)/(channels-1))); + for (c=0;c<3;c++) + for (i=0;i<21;i++) + maskLogE[c][i] += channel_offset; +#if 0 + for (c=0;c<3;c++) + { + for (i=0;i<21;i++) + printf("%f ", maskLogE[c][i]); + } +#endif + for (c=0;cnb_streams||nb_coupled_streams<0)return 0; + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + return align(sizeof(OpusMSEncoder)) + + nb_coupled_streams * align(coupled_size) + + (nb_streams-nb_coupled_streams) * align(mono_size); +} + +opus_int32 opus_multistream_surround_encoder_get_size(int channels, int mapping_family) +{ + int nb_streams; + int nb_coupled_streams; + opus_int32 size; + + if (mapping_family==0) + { + if (channels==1) + { + nb_streams=1; + nb_coupled_streams=0; + } else if (channels==2) + { + nb_streams=1; + nb_coupled_streams=1; + } else + return 0; + } else if (mapping_family==1 && channels<=8 && channels>=1) + { + nb_streams=vorbis_mappings[channels-1].nb_streams; + nb_coupled_streams=vorbis_mappings[channels-1].nb_coupled_streams; + } else if (mapping_family==255) + { + nb_streams=channels; + nb_coupled_streams=0; + } else + return 0; + size = opus_multistream_encoder_get_size(nb_streams, nb_coupled_streams); + if (channels>2) + { + size += channels*(120*sizeof(opus_val32) + sizeof(opus_val32)); + } + return size; +} + + +static int opus_multistream_encoder_init_impl( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application, + int surround +) +{ + int coupled_size; + int mono_size; + int i, ret; + char *ptr; + + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) + return OPUS_BAD_ARG; + + st->layout.nb_channels = channels; + st->layout.nb_streams = streams; + st->layout.nb_coupled_streams = coupled_streams; + st->subframe_mem[0]=st->subframe_mem[1]=st->subframe_mem[2]=0; + if (!surround) + st->lfe_stream = -1; + st->bitrate_bps = OPUS_AUTO; + st->application = application; + st->variable_duration = OPUS_FRAMESIZE_ARG; + for (i=0;ilayout.nb_channels;i++) + st->layout.mapping[i] = mapping[i]; + if (!validate_layout(&st->layout) || !validate_encoder_layout(&st->layout)) + return OPUS_BAD_ARG; + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + + for (i=0;ilayout.nb_coupled_streams;i++) + { + ret = opus_encoder_init((OpusEncoder*)ptr, Fs, 2, application); + if(ret!=OPUS_OK)return ret; + if (i==st->lfe_stream) + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_SET_LFE(1)); + ptr += align(coupled_size); + } + for (;ilayout.nb_streams;i++) + { + ret = opus_encoder_init((OpusEncoder*)ptr, Fs, 1, application); + if (i==st->lfe_stream) + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_SET_LFE(1)); + if(ret!=OPUS_OK)return ret; + ptr += align(mono_size); + } + if (surround) + { + OPUS_CLEAR(ms_get_preemph_mem(st), channels); + OPUS_CLEAR(ms_get_window_mem(st), channels*120); + } + st->surround = surround; + return OPUS_OK; +} + +int opus_multistream_encoder_init( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application +) +{ + return opus_multistream_encoder_init_impl(st, Fs, channels, streams, coupled_streams, mapping, application, 0); +} + +int opus_multistream_surround_encoder_init( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application +) +{ + if ((channels>255) || (channels<1)) + return OPUS_BAD_ARG; + st->lfe_stream = -1; + if (mapping_family==0) + { + if (channels==1) + { + *streams=1; + *coupled_streams=0; + mapping[0]=0; + } else if (channels==2) + { + *streams=1; + *coupled_streams=1; + mapping[0]=0; + mapping[1]=1; + } else + return OPUS_UNIMPLEMENTED; + } else if (mapping_family==1 && channels<=8 && channels>=1) + { + int i; + *streams=vorbis_mappings[channels-1].nb_streams; + *coupled_streams=vorbis_mappings[channels-1].nb_coupled_streams; + for (i=0;i=6) + st->lfe_stream = *streams-1; + } else if (mapping_family==255) + { + int i; + *streams=channels; + *coupled_streams=0; + for(i=0;i2&&mapping_family==1); +} + +OpusMSEncoder *opus_multistream_encoder_create( + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application, + int *error +) +{ + int ret; + OpusMSEncoder *st; + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (coupled_streams+streams>255) || (streams<1) || (coupled_streams<0)) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + st = (OpusMSEncoder *)opus_alloc(opus_multistream_encoder_get_size(streams, coupled_streams)); + if (st==NULL) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + ret = opus_multistream_encoder_init(st, Fs, channels, streams, coupled_streams, mapping, application); + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} + +OpusMSEncoder *opus_multistream_surround_encoder_create( + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application, + int *error +) +{ + int ret; + OpusMSEncoder *st; + if ((channels>255) || (channels<1)) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + st = (OpusMSEncoder *)opus_alloc(opus_multistream_surround_encoder_get_size(channels, mapping_family)); + if (st==NULL) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + ret = opus_multistream_surround_encoder_init(st, Fs, channels, mapping_family, streams, coupled_streams, mapping, application); + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} + +static void surround_rate_allocation( + OpusMSEncoder *st, + opus_int32 *rate, + int frame_size + ) +{ + int i; + opus_int32 channel_rate; + opus_int32 Fs; + char *ptr; + int stream_offset; + int lfe_offset; + int coupled_ratio; /* Q8 */ + int lfe_ratio; /* Q8 */ + + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_SAMPLE_RATE(&Fs)); + + if (st->bitrate_bps > st->layout.nb_channels*40000) + stream_offset = 20000; + else + stream_offset = st->bitrate_bps/st->layout.nb_channels/2; + stream_offset += 60*(Fs/frame_size-50); + /* We start by giving each stream (coupled or uncoupled) the same bitrate. + This models the main saving of coupled channels over uncoupled. */ + /* The LFE stream is an exception to the above and gets fewer bits. */ + lfe_offset = 3500 + 60*(Fs/frame_size-50); + /* Coupled streams get twice the mono rate after the first 20 kb/s. */ + coupled_ratio = 512; + /* Should depend on the bitrate, for now we assume LFE gets 1/8 the bits of mono */ + lfe_ratio = 32; + + /* Compute bitrate allocation between streams */ + if (st->bitrate_bps==OPUS_AUTO) + { + channel_rate = Fs+60*Fs/frame_size; + } else if (st->bitrate_bps==OPUS_BITRATE_MAX) + { + channel_rate = 300000; + } else { + int nb_lfe; + int nb_uncoupled; + int nb_coupled; + int total; + nb_lfe = (st->lfe_stream!=-1); + nb_coupled = st->layout.nb_coupled_streams; + nb_uncoupled = st->layout.nb_streams-nb_coupled-nb_lfe; + total = (nb_uncoupled<<8) /* mono */ + + coupled_ratio*nb_coupled /* stereo */ + + nb_lfe*lfe_ratio; + channel_rate = 256*(st->bitrate_bps-lfe_offset*nb_lfe-stream_offset*(nb_coupled+nb_uncoupled))/total; + } +#ifndef FIXED_POINT + if (st->variable_duration==OPUS_FRAMESIZE_VARIABLE && frame_size != Fs/50) + { + opus_int32 bonus; + bonus = 60*(Fs/frame_size-50); + channel_rate += bonus; + } +#endif + + for (i=0;ilayout.nb_streams;i++) + { + if (ilayout.nb_coupled_streams) + rate[i] = stream_offset+(channel_rate*coupled_ratio>>8); + else if (i!=st->lfe_stream) + rate[i] = stream_offset+channel_rate; + else + rate[i] = lfe_offset+(channel_rate*lfe_ratio>>8); + } +} + +/* Max size in case the encoder decides to return three frames */ +#define MS_FRAME_TMP (3*1275+7) +static int opus_multistream_encode_native +( + OpusMSEncoder *st, + opus_copy_channel_in_func copy_channel_in, + const void *pcm, + int analysis_frame_size, + unsigned char *data, + opus_int32 max_data_bytes, + int lsb_depth, + downmix_func downmix +) +{ + opus_int32 Fs; + int coupled_size; + int mono_size; + int s; + char *ptr; + int tot_size; + VARDECL(opus_val16, buf); + VARDECL(opus_val16, bandSMR); + unsigned char tmp_data[MS_FRAME_TMP]; + OpusRepacketizer rp; + opus_int32 vbr; + const CELTMode *celt_mode; + opus_int32 bitrates[256]; + opus_val16 bandLogE[42]; + opus_val32 *mem = NULL; + opus_val32 *preemph_mem=NULL; + int frame_size; + ALLOC_STACK; + + if (st->surround) + { + preemph_mem = ms_get_preemph_mem(st); + mem = ms_get_window_mem(st); + } + + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_SAMPLE_RATE(&Fs)); + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_VBR(&vbr)); + opus_encoder_ctl((OpusEncoder*)ptr, CELT_GET_MODE(&celt_mode)); + + { + opus_int32 delay_compensation; + int channels; + + channels = st->layout.nb_streams + st->layout.nb_coupled_streams; + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_LOOKAHEAD(&delay_compensation)); + delay_compensation -= Fs/400; + frame_size = compute_frame_size(pcm, analysis_frame_size, + st->variable_duration, channels, Fs, st->bitrate_bps, + delay_compensation, downmix +#ifndef DISABLE_FLOAT_API + , st->subframe_mem +#endif + ); + } + + if (400*frame_size < Fs) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + /* Validate frame_size before using it to allocate stack space. + This mirrors the checks in opus_encode[_float](). */ + if (400*frame_size != Fs && 200*frame_size != Fs && + 100*frame_size != Fs && 50*frame_size != Fs && + 25*frame_size != Fs && 50*frame_size != 3*Fs) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + ALLOC(buf, 2*frame_size, opus_val16); + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + + ALLOC(bandSMR, 21*st->layout.nb_channels, opus_val16); + if (st->surround) + { + surround_analysis(celt_mode, pcm, bandSMR, mem, preemph_mem, frame_size, 120, st->layout.nb_channels, Fs, copy_channel_in); + } + + if (max_data_bytes < 4*st->layout.nb_streams-1) + { + RESTORE_STACK; + return OPUS_BUFFER_TOO_SMALL; + } + + /* Compute bitrate allocation between streams (this could be a lot better) */ + surround_rate_allocation(st, bitrates, frame_size); + + if (!vbr) + max_data_bytes = IMIN(max_data_bytes, 3*st->bitrate_bps/(3*8*Fs/frame_size)); + + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(bitrates[s])); + if (st->surround) + { + opus_int32 equiv_rate; + equiv_rate = st->bitrate_bps; + if (frame_size*50 < Fs) + equiv_rate -= 60*(Fs/frame_size - 50)*st->layout.nb_channels; + if (equiv_rate > 10000*st->layout.nb_channels) + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + else if (equiv_rate > 7000*st->layout.nb_channels) + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_SUPERWIDEBAND)); + else if (equiv_rate > 5000*st->layout.nb_channels) + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_WIDEBAND)); + else + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + if (s < st->layout.nb_coupled_streams) + { + /* To preserve the spatial image, force stereo CELT on coupled streams */ + opus_encoder_ctl(enc, OPUS_SET_FORCE_MODE(MODE_CELT_ONLY)); + opus_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(2)); + } + } + } + + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + /* Counting ToC */ + tot_size = 0; + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + int len; + int curr_max; + int c1, c2; + + opus_repacketizer_init(&rp); + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + { + int i; + int left, right; + left = get_left_channel(&st->layout, s, -1); + right = get_right_channel(&st->layout, s, -1); + (*copy_channel_in)(buf, 2, + pcm, st->layout.nb_channels, left, frame_size); + (*copy_channel_in)(buf+1, 2, + pcm, st->layout.nb_channels, right, frame_size); + ptr += align(coupled_size); + if (st->surround) + { + for (i=0;i<21;i++) + { + bandLogE[i] = bandSMR[21*left+i]; + bandLogE[21+i] = bandSMR[21*right+i]; + } + } + c1 = left; + c2 = right; + } else { + int i; + int chan = get_mono_channel(&st->layout, s, -1); + (*copy_channel_in)(buf, 1, + pcm, st->layout.nb_channels, chan, frame_size); + ptr += align(mono_size); + if (st->surround) + { + for (i=0;i<21;i++) + bandLogE[i] = bandSMR[21*chan+i]; + } + c1 = chan; + c2 = -1; + } + if (st->surround) + opus_encoder_ctl(enc, OPUS_SET_ENERGY_MASK(bandLogE)); + /* number of bytes left (+Toc) */ + curr_max = max_data_bytes - tot_size; + /* Reserve three bytes for the last stream and four for the others */ + curr_max -= IMAX(0,4*(st->layout.nb_streams-s-1)-1); + curr_max = IMIN(curr_max,MS_FRAME_TMP); + if (!vbr && s == st->layout.nb_streams-1) + opus_encoder_ctl(enc, OPUS_SET_BITRATE(curr_max*(8*Fs/frame_size))); + len = opus_encode_native(enc, buf, frame_size, tmp_data, curr_max, lsb_depth, + pcm, analysis_frame_size, c1, c2, st->layout.nb_channels, downmix); + if (len<0) + { + RESTORE_STACK; + return len; + } + /* We need to use the repacketizer to add the self-delimiting lengths + while taking into account the fact that the encoder can now return + more than one frame at a time (e.g. 60 ms CELT-only) */ + opus_repacketizer_cat(&rp, tmp_data, len); + len = opus_repacketizer_out_range_impl(&rp, 0, opus_repacketizer_get_nb_frames(&rp), + data, max_data_bytes-tot_size, s != st->layout.nb_streams-1, !vbr && s == st->layout.nb_streams-1); + data += len; + tot_size += len; + } + /*printf("\n");*/ + RESTORE_STACK; + return tot_size; +} + +#if !defined(DISABLE_FLOAT_API) +static void opus_copy_channel_in_float( + opus_val16 *dst, + int dst_stride, + const void *src, + int src_stride, + int src_channel, + int frame_size +) +{ + const float *float_src; + opus_int32 i; + float_src = (const float *)src; + for (i=0;ibitrate_bps = value; + } + break; + case OPUS_GET_BITRATE_REQUEST: + { + int s; + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = 0; + for (s=0;slayout.nb_streams;s++) + { + opus_int32 rate; + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + opus_encoder_ctl(enc, request, &rate); + *value += rate; + } + } + break; + case OPUS_GET_LSB_DEPTH_REQUEST: + case OPUS_GET_VBR_REQUEST: + case OPUS_GET_APPLICATION_REQUEST: + case OPUS_GET_BANDWIDTH_REQUEST: + case OPUS_GET_COMPLEXITY_REQUEST: + case OPUS_GET_PACKET_LOSS_PERC_REQUEST: + case OPUS_GET_DTX_REQUEST: + case OPUS_GET_VOICE_RATIO_REQUEST: + case OPUS_GET_VBR_CONSTRAINT_REQUEST: + case OPUS_GET_SIGNAL_REQUEST: + case OPUS_GET_LOOKAHEAD_REQUEST: + case OPUS_GET_SAMPLE_RATE_REQUEST: + case OPUS_GET_INBAND_FEC_REQUEST: + case OPUS_GET_FORCE_CHANNELS_REQUEST: + case OPUS_GET_PREDICTION_DISABLED_REQUEST: + { + OpusEncoder *enc; + /* For int32* GET params, just query the first stream */ + opus_int32 *value = va_arg(ap, opus_int32*); + enc = (OpusEncoder*)ptr; + ret = opus_encoder_ctl(enc, request, value); + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + int s; + opus_uint32 *value = va_arg(ap, opus_uint32*); + opus_uint32 tmp; + if (!value) + { + goto bad_arg; + } + *value=0; + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_encoder_ctl(enc, request, &tmp); + if (ret != OPUS_OK) break; + *value ^= tmp; + } + } + break; + case OPUS_SET_LSB_DEPTH_REQUEST: + case OPUS_SET_COMPLEXITY_REQUEST: + case OPUS_SET_VBR_REQUEST: + case OPUS_SET_VBR_CONSTRAINT_REQUEST: + case OPUS_SET_MAX_BANDWIDTH_REQUEST: + case OPUS_SET_BANDWIDTH_REQUEST: + case OPUS_SET_SIGNAL_REQUEST: + case OPUS_SET_APPLICATION_REQUEST: + case OPUS_SET_INBAND_FEC_REQUEST: + case OPUS_SET_PACKET_LOSS_PERC_REQUEST: + case OPUS_SET_DTX_REQUEST: + case OPUS_SET_FORCE_MODE_REQUEST: + case OPUS_SET_FORCE_CHANNELS_REQUEST: + case OPUS_SET_PREDICTION_DISABLED_REQUEST: + { + int s; + /* This works for int32 params */ + opus_int32 value = va_arg(ap, opus_int32); + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_encoder_ctl(enc, request, value); + if (ret != OPUS_OK) + break; + } + } + break; + case OPUS_MULTISTREAM_GET_ENCODER_STATE_REQUEST: + { + int s; + opus_int32 stream_id; + OpusEncoder **value; + stream_id = va_arg(ap, opus_int32); + if (stream_id<0 || stream_id >= st->layout.nb_streams) + ret = OPUS_BAD_ARG; + value = va_arg(ap, OpusEncoder**); + if (!value) + { + goto bad_arg; + } + for (s=0;slayout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + *value = (OpusEncoder*)ptr; + } + break; + case OPUS_SET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->variable_duration = value; + } + break; + case OPUS_GET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->variable_duration; + } + break; + case OPUS_RESET_STATE: + { + int s; + st->subframe_mem[0] = st->subframe_mem[1] = st->subframe_mem[2] = 0; + if (st->surround) + { + OPUS_CLEAR(ms_get_preemph_mem(st), st->layout.nb_channels); + OPUS_CLEAR(ms_get_window_mem(st), st->layout.nb_channels*120); + } + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_encoder_ctl(enc, OPUS_RESET_STATE); + if (ret != OPUS_OK) + break; + } + } + break; + default: + ret = OPUS_UNIMPLEMENTED; + break; + } + + va_end(ap); + return ret; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +} + +void opus_multistream_encoder_destroy(OpusMSEncoder *st) +{ + opus_free(st); +} diff --git a/code/opus-1.0.2/src/opus_private.h b/code/opus-1.1/src/opus_private.h similarity index 59% rename from code/opus-1.0.2/src/opus_private.h rename to code/opus-1.1/src/opus_private.h index 52482bc1..83225f2b 100644 --- a/code/opus-1.0.2/src/opus_private.h +++ b/code/opus-1.1/src/opus_private.h @@ -31,15 +31,29 @@ #include "arch.h" #include "opus.h" +#include "celt.h" struct OpusRepacketizer { unsigned char toc; int nb_frames; const unsigned char *frames[48]; - short len[48]; + opus_int16 len[48]; int framesize; }; +typedef struct ChannelLayout { + int nb_channels; + int nb_streams; + int nb_coupled_streams; + unsigned char mapping[256]; +} ChannelLayout; + +int validate_layout(const ChannelLayout *layout); +int get_left_channel(const ChannelLayout *layout, int stream_id, int prev); +int get_right_channel(const ChannelLayout *layout, int stream_id, int prev); +int get_mono_channel(const ChannelLayout *layout, int stream_id, int prev); + + #define MODE_SILK_ONLY 1000 #define MODE_HYBRID 1001 @@ -68,18 +82,48 @@ struct OpusRepacketizer { #define OPUS_SET_FORCE_MODE_REQUEST 11002 #define OPUS_SET_FORCE_MODE(x) OPUS_SET_FORCE_MODE_REQUEST, __opus_check_int(x) +typedef void (*downmix_func)(const void *, opus_val32 *, int, int, int, int, int); +void downmix_float(const void *_x, opus_val32 *sub, int subframe, int offset, int c1, int c2, int C); +void downmix_int(const void *_x, opus_val32 *sub, int subframe, int offset, int c1, int c2, int C); + +int optimize_framesize(const opus_val16 *x, int len, int C, opus_int32 Fs, + int bitrate, opus_val16 tonality, float *mem, int buffering, + downmix_func downmix); int encode_size(int size, unsigned char *data); +opus_int32 frame_size_select(opus_int32 frame_size, int variable_duration, opus_int32 Fs); + +opus_int32 compute_frame_size(const void *analysis_pcm, int frame_size, + int variable_duration, int C, opus_int32 Fs, int bitrate_bps, + int delay_compensation, downmix_func downmix +#ifndef DISABLE_FLOAT_API + , float *subframe_mem +#endif + ); + +opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_size, + unsigned char *data, opus_int32 out_data_bytes, int lsb_depth, + const void *analysis_pcm, opus_int32 analysis_size, int c1, int c2, int analysis_channels, downmix_func downmix); + int opus_decode_native(OpusDecoder *st, const unsigned char *data, opus_int32 len, - opus_val16 *pcm, int frame_size, int decode_fec, int self_delimited, int *packet_offset); + opus_val16 *pcm, int frame_size, int decode_fec, int self_delimited, + opus_int32 *packet_offset, int soft_clip); /* Make sure everything's aligned to sizeof(void *) bytes */ -static inline int align(int i) +static OPUS_INLINE int align(int i) { - return (i+sizeof(void *)-1)&-((int)sizeof(void *)); + return (i+(int)sizeof(void *)-1)&-(int)sizeof(void *); } -opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int end, unsigned char *data, opus_int32 maxlen, int self_delimited); +int opus_packet_parse_impl(const unsigned char *data, opus_int32 len, + int self_delimited, unsigned char *out_toc, + const unsigned char *frames[48], opus_int16 size[48], + int *payload_offset, opus_int32 *packet_offset); + +opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int end, + unsigned char *data, opus_int32 maxlen, int self_delimited, int pad); + +int pad_frame(unsigned char *data, opus_int32 len, opus_int32 new_len); #endif /* OPUS_PRIVATE_H */ diff --git a/code/opus-1.0.2/src/repacketizer.c b/code/opus-1.1/src/repacketizer.c similarity index 50% rename from code/opus-1.0.2/src/repacketizer.c rename to code/opus-1.1/src/repacketizer.c index 26315b62..a62675ce 100644 --- a/code/opus-1.0.2/src/repacketizer.c +++ b/code/opus-1.1/src/repacketizer.c @@ -58,7 +58,7 @@ void opus_repacketizer_destroy(OpusRepacketizer *rp) opus_free(rp); } -int opus_repacketizer_cat(OpusRepacketizer *rp, const unsigned char *data, opus_int32 len) +static int opus_repacketizer_cat_impl(OpusRepacketizer *rp, const unsigned char *data, opus_int32 len, int self_delimited) { unsigned char tmp_toc; int curr_nb_frames,ret; @@ -82,24 +82,31 @@ int opus_repacketizer_cat(OpusRepacketizer *rp, const unsigned char *data, opus_ return OPUS_INVALID_PACKET; } - ret=opus_packet_parse(data, len, &tmp_toc, &rp->frames[rp->nb_frames], &rp->len[rp->nb_frames], NULL); + ret=opus_packet_parse_impl(data, len, self_delimited, &tmp_toc, &rp->frames[rp->nb_frames], &rp->len[rp->nb_frames], NULL, NULL); if(ret<1)return ret; rp->nb_frames += curr_nb_frames; return OPUS_OK; } +int opus_repacketizer_cat(OpusRepacketizer *rp, const unsigned char *data, opus_int32 len) +{ + return opus_repacketizer_cat_impl(rp, data, len, 0); +} + int opus_repacketizer_get_nb_frames(OpusRepacketizer *rp) { return rp->nb_frames; } -opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int end, unsigned char *data, opus_int32 maxlen, int self_delimited) +opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int end, + unsigned char *data, opus_int32 maxlen, int self_delimited, int pad) { int i, count; opus_int32 tot_size; - short *len; + opus_int16 *len; const unsigned char **frames; + unsigned char * ptr; if (begin<0 || begin>=end || end>rp->nb_frames) { @@ -115,18 +122,15 @@ opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int else tot_size = 0; - switch (count) - { - case 1: + ptr = data; + if (count==1) { /* Code 0 */ tot_size += len[0]+1; if (tot_size > maxlen) return OPUS_BUFFER_TOO_SMALL; - *data++ = rp->toc&0xFC; - } - break; - case 2: + *ptr++ = rp->toc&0xFC; + } else if (count==2) { if (len[1] == len[0]) { @@ -134,22 +138,28 @@ opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int tot_size += 2*len[0]+1; if (tot_size > maxlen) return OPUS_BUFFER_TOO_SMALL; - *data++ = (rp->toc&0xFC) | 0x1; + *ptr++ = (rp->toc&0xFC) | 0x1; } else { /* Code 2 */ tot_size += len[0]+len[1]+2+(len[0]>=252); if (tot_size > maxlen) return OPUS_BUFFER_TOO_SMALL; - *data++ = (rp->toc&0xFC) | 0x2; - data += encode_size(len[0], data); + *ptr++ = (rp->toc&0xFC) | 0x2; + ptr += encode_size(len[0], ptr); } } - break; - default: + if (count > 2 || (pad && tot_size < maxlen)) { /* Code 3 */ int vbr; + int pad_amount=0; + /* Restart the process for the padding case */ + ptr = data; + if (self_delimited) + tot_size = 1 + (len[count-1]>=252); + else + tot_size = 0; vbr = 0; for (i=1;i maxlen) return OPUS_BUFFER_TOO_SMALL; - *data++ = (rp->toc&0xFC) | 0x3; - *data++ = count | 0x80; - for (i=0;itoc&0xFC) | 0x3; + *ptr++ = count | 0x80; } else { tot_size += count*len[0]+2; if (tot_size > maxlen) return OPUS_BUFFER_TOO_SMALL; - *data++ = (rp->toc&0xFC) | 0x3; - *data++ = count; + *ptr++ = (rp->toc&0xFC) | 0x3; + *ptr++ = count; + } + pad_amount = pad ? (maxlen-tot_size) : 0; + if (pad_amount != 0) + { + int nb_255s; + data[1] |= 0x40; + nb_255s = (pad_amount-1)/255; + for (i=0;inb_frames, data, maxlen, 0); + return opus_repacketizer_out_range_impl(rp, 0, rp->nb_frames, data, maxlen, 0, 0); } +int opus_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len) +{ + OpusRepacketizer rp; + opus_int32 ret; + if (len < 1) + return OPUS_BAD_ARG; + if (len==new_len) + return OPUS_OK; + else if (len > new_len) + return OPUS_BAD_ARG; + opus_repacketizer_init(&rp); + /* Moving payload to the end of the packet so we can do in-place padding */ + OPUS_MOVE(data+new_len-len, data, len); + opus_repacketizer_cat(&rp, data+new_len-len, len); + ret = opus_repacketizer_out_range_impl(&rp, 0, rp.nb_frames, data, new_len, 0, 1); + if (ret > 0) + return OPUS_OK; + else + return ret; +} + +opus_int32 opus_packet_unpad(unsigned char *data, opus_int32 len) +{ + OpusRepacketizer rp; + opus_int32 ret; + if (len < 1) + return OPUS_BAD_ARG; + opus_repacketizer_init(&rp); + ret = opus_repacketizer_cat(&rp, data, len); + if (ret < 0) + return ret; + ret = opus_repacketizer_out_range_impl(&rp, 0, rp.nb_frames, data, len, 0, 0); + celt_assert(ret > 0 && ret <= len); + return ret; +} + +int opus_multistream_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len, int nb_streams) +{ + int s; + int count; + unsigned char toc; + opus_int16 size[48]; + opus_int32 packet_offset; + opus_int32 amount; + + if (len < 1) + return OPUS_BAD_ARG; + if (len==new_len) + return OPUS_OK; + else if (len > new_len) + return OPUS_BAD_ARG; + amount = new_len - len; + /* Seek to last stream */ + for (s=0;s -#include - -static unsigned op_parse_uint16le(const unsigned char *_data){ - return _data[0]|_data[1]<<8; -} - -static int op_parse_int16le(const unsigned char *_data){ - int ret; - ret=_data[0]|_data[1]<<8; - return (ret^0x8000)-0x8000; -} - -static opus_uint32 op_parse_uint32le(const unsigned char *_data){ - return _data[0]|_data[1]<<8|_data[2]<<16|_data[3]<<24; -} - -int opus_head_parse(OpusHead *_head,const unsigned char *_data,size_t _len){ - OpusHead head; - if(_len<8)return OP_ENOTFORMAT; - if(memcmp(_data,"OpusHead",8)!=0)return OP_ENOTFORMAT; - if(_len<9)return OP_EBADHEADER; - head.version=_data[8]; - if(head.version>15)return OP_EVERSION; - if(_len<19)return OP_EBADHEADER; - head.channel_count=_data[9]; - head.pre_skip=op_parse_uint16le(_data+10); - head.input_sample_rate=op_parse_uint32le(_data+12); - head.output_gain=op_parse_int16le(_data+16); - head.mapping_family=_data[18]; - if(head.mapping_family==0){ - if(head.channel_count<1||head.channel_count>2)return OP_EBADHEADER; - if(head.version<=1&&_len>19)return OP_EBADHEADER; - head.stream_count=1; - head.coupled_count=head.channel_count-1; - if(_head!=NULL){ - _head->mapping[0]=0; - _head->mapping[1]=1; - } - } - else if(head.mapping_family==1){ - size_t size; - int ci; - if(head.channel_count<1||head.channel_count>8)return OP_EBADHEADER; - size=21+head.channel_count; - if(_lensize)return OP_EBADHEADER; - head.stream_count=_data[19]; - if(head.stream_count<1)return OP_EBADHEADER; - head.coupled_count=_data[20]; - if(head.coupled_count>head.stream_count)return OP_EBADHEADER; - for(ci=0;ci=head.stream_count+head.coupled_count - &&_data[21+ci]!=255){ - return OP_EBADHEADER; - } - } - if(_head!=NULL)memcpy(_head->mapping,_data+21,head.channel_count); - } - /*General purpose players should not attempt to play back content with - channel mapping family 255.*/ - else if(head.mapping_family==255)return OP_EIMPL; - /*No other channel mapping families are currently defined.*/ - else return OP_EBADHEADER; - if(_head!=NULL)memcpy(_head,&head,head.mapping-(unsigned char *)&head); - return 0; -} - -void opus_tags_init(OpusTags *_tags){ - memset(_tags,0,sizeof(*_tags)); -} - -void opus_tags_clear(OpusTags *_tags){ - int i; - for(i=_tags->comments;i-->0;)_ogg_free(_tags->user_comments[i]); - _ogg_free(_tags->user_comments); - _ogg_free(_tags->comment_lengths); - _ogg_free(_tags->vendor); -} - -/*The actual implementation of opus_tags_parse(). - Unlike the public API, this function requires _tags to already be - initialized, modifies its contents before success is guaranteed, and assumes - the caller will clear it on error.*/ -int opus_tags_parse_impl(OpusTags *_tags, - const unsigned char *_data,size_t _len){ - opus_uint32 count; - size_t size; - size_t len; - int ncomments; - int i; - len=_len; - if(len<8)return OP_ENOTFORMAT; - if(memcmp(_data,"OpusTags",8)!=0)return OP_ENOTFORMAT; - if(len<16)return OP_EBADHEADER; - _data+=8; - len-=8; - count=op_parse_uint32le(_data); - _data+=4; - len-=4; - if(count>len)return OP_EBADHEADER; - if(_tags!=NULL){ - char *vendor; - size=count+1; - if(sizevendor=vendor; - } - _data+=count; - len-=count; - if(len<4)return OP_EBADHEADER; - count=op_parse_uint32le(_data); - _data+=4; - len-=4; - /*Check to make sure there's minimally sufficient data left in the packet.*/ - if(count>len>>2)return OP_EBADHEADER; - /*Check for overflow (the API limits this to an int).*/ - if(count>(opus_uint32)INT_MAX-1)return OP_EFAULT; - if(_tags!=NULL){ - size=sizeof(*_tags->comment_lengths)*(count+1); - if(size/sizeof(*_tags->comment_lengths)!=count+1)return OP_EFAULT; - _tags->comment_lengths=(int *)_ogg_malloc(size); - size=sizeof(*_tags->user_comments)*(count+1); - if(size/sizeof(*_tags->user_comments)!=count+1)return OP_EFAULT; - _tags->user_comments=(char **)_ogg_malloc(size); - if(_tags->comment_lengths==NULL||_tags->user_comments==NULL){ - return OP_EFAULT; - } - } - ncomments=(int)count; - for(i=0;ilen>>2)return OP_EBADHEADER; - count=op_parse_uint32le(_data); - _data+=4; - len-=4; - if(count>len)return OP_EBADHEADER; - /*Check for overflow (the API limits this to an int).*/ - if(count>(opus_uint32)INT_MAX)return OP_EFAULT; - if(_tags!=NULL){ - _tags->comment_lengths[i]=(int)count; - size=count+1; - if(sizeuser_comments[i]=(char *)_ogg_malloc(size); - if(_tags->user_comments[i]==NULL)return OP_EFAULT; - _tags->comments=i+1; - memcpy(_tags->user_comments[i],_data,count); - _tags->user_comments[i][count]='\0'; - } - _data+=count; - len-=count; - } - if(_tags!=NULL){ - _tags->user_comments[ncomments]=NULL; - _tags->comment_lengths[ncomments]=0; - } - return 0; -} - -int opus_tags_parse(OpusTags *_tags,const unsigned char *_data,size_t _len){ - if(_tags!=NULL){ - OpusTags tags; - int ret; - opus_tags_init(&tags); - ret=opus_tags_parse_impl(&tags,_data,_len); - if(ret<0)opus_tags_clear(&tags); - else *_tags=*&tags; - return ret; - } - else return opus_tags_parse_impl(NULL,_data,_len); -} - -/*Add room for a new comment.*/ -static int op_tags_add_prepare(OpusTags *_tags){ - char **user_comments; - int *comment_lengths; - int ncomments; - ncomments=_tags->comments; - user_comments=_ogg_realloc(_tags->user_comments, - sizeof(*_tags->user_comments)*(ncomments+2)); - if(OP_UNLIKELY(user_comments==NULL))return OP_EFAULT; - _tags->user_comments=user_comments; - comment_lengths=_ogg_realloc(_tags->comment_lengths, - sizeof(*_tags->comment_lengths)*(ncomments+2)); - if(OP_UNLIKELY(comment_lengths==NULL))return OP_EFAULT; - _tags->comment_lengths=comment_lengths; - comment_lengths[ncomments]=comment_lengths[ncomments+1]=0; - /*Our caller will always set user_comments[ncomments].*/ - user_comments[ncomments+1]=NULL; - return 0; -} - -int opus_tags_add(OpusTags *_tags,const char *_tag,const char *_value){ - char *comment; - int tag_len; - int value_len; - int ncomments; - int ret; - ret=op_tags_add_prepare(_tags); - if(OP_UNLIKELY(ret<0))return ret; - tag_len=strlen(_tag); - value_len=strlen(_value); - ncomments=_tags->comments; - /*+2 for '=' and '\0'.*/ - _tags->user_comments[ncomments]=comment= - (char *)_ogg_malloc(sizeof(*comment)*(tag_len+value_len+2)); - if(OP_UNLIKELY(comment==NULL))return OP_EFAULT; - _tags->comment_lengths[ncomments]=tag_len+value_len+1; - memcpy(comment,_tag,sizeof(*comment)*tag_len); - comment[tag_len]='='; - memcpy(comment+tag_len+1,_value,sizeof(*comment)*(value_len+1)); - return 0; -} - -int opus_tags_add_comment(OpusTags *_tags,const char *_comment){ - char *comment; - int ncomments; - int comment_len; - int ret; - ret=op_tags_add_prepare(_tags); - if(OP_UNLIKELY(ret<0))return ret; - comment_len=strlen(_comment); - ncomments=_tags->comments; - _tags->user_comments[ncomments]=comment=(char *) - _ogg_malloc(sizeof(*_tags->user_comments[ncomments])*(comment_len+1)); - if(OP_UNLIKELY(comment==NULL))return OP_EFAULT; - _tags->comment_lengths[ncomments]=comment_len; - memcpy(comment,_comment,sizeof(*comment)*(comment_len+1)); - return 0; -} - -/*Is _a a "tag=value" comment whose tag matches _b? - 0 if it is, a non-zero value otherwise.*/ -static int op_tagcompare(const char *_a,const char *_b,int _n){ - return op_strncasecmp(_a,_b,_n)||_a[_n]!='='; -} - -const char *opus_tags_query(const OpusTags *_tags,const char *_tag,int _count){ - char **user_comments; - int tag_len; - int found; - int ncomments; - int ci; - tag_len=strlen(_tag); - ncomments=_tags->comments; - user_comments=_tags->user_comments; - found=0; - for(ci=0;cicomments; - user_comments=_tags->user_comments; - found=0; - for(ci=0;ci -#include -#include -#include - -typedef struct OpusMemStream OpusMemStream; - -#define OP_MEM_SIZE_MAX (~(size_t)0>>1) -#define OP_MEM_DIFF_MAX ((ptrdiff_t)OP_MEM_SIZE_MAX) - -/*The context information needed to read from a block of memory as if it were a - file.*/ -struct OpusMemStream{ - /*The block of memory to read from.*/ - const unsigned char *data; - /*The total size of the block. - This must be at most OP_MEM_SIZE_MAX to prevent signed overflow while - seeking.*/ - ptrdiff_t size; - /*The current file position. - This is allowed to be set arbitrarily greater than size (i.e., past the end - of the block, though we will not read data past the end of the block), but - is not allowed to be negative (i.e., before the beginning of the block).*/ - ptrdiff_t pos; -}; - -static int op_fread(void *_stream,unsigned char *_ptr,int _buf_size){ - FILE *stream; - size_t ret; - /*Check for empty read.*/ - if(_buf_size<=0)return 0; - stream=(FILE *)_stream; - ret=fread(_ptr,1,_buf_size,stream); - OP_ASSERT(ret<=(size_t)_buf_size); - /*If ret==0 and !feof(stream), there was a read error.*/ - return ret>0||feof(stream)?(int)ret:OP_EREAD; -} - -static int op_fseek(void *_stream,opus_int64 _offset,int _whence){ -#if defined(__MINGW32__) - return fseeko64((FILE *)_stream,_offset,_whence); -#elif defined(_MSC_VER) - return _fseeki64((FILE *)_stream,_offset,_whence); -#else - return fseeko((FILE *)_stream,(off_t)_offset,_whence); -#endif -} - -static opus_int64 op_ftell(void *_stream){ -#if defined(__MINGW32__) - return ftello64((FILE *)_stream); -#elif defined(_MSC_VER) - return _ftelli64((FILE *)_stream); -#else - return ftello((FILE *)_stream); -#endif -} - -static const OpusFileCallbacks OP_FILE_CALLBACKS={ - op_fread, - op_fseek, - op_ftell, - (op_close_func)fclose -}; - -void *op_fopen(OpusFileCallbacks *_cb,const char *_path,const char *_mode){ - FILE *fp; - fp=fopen(_path,_mode); - if(fp!=NULL)*_cb=*&OP_FILE_CALLBACKS; - return fp; -} - -void *op_fdopen(OpusFileCallbacks *_cb,int _fd,const char *_mode){ - FILE *fp; - fp=fdopen(_fd,_mode); - if(fp!=NULL)*_cb=*&OP_FILE_CALLBACKS; - return fp; -} - -void *op_freopen(OpusFileCallbacks *_cb,const char *_path,const char *_mode, - void *_stream){ - FILE *fp; - fp=freopen(_path,_mode,(FILE *)_stream); - if(fp!=NULL)*_cb=*&OP_FILE_CALLBACKS; - return fp; -} - -static int op_mem_read(void *_stream,unsigned char *_ptr,int _buf_size){ - OpusMemStream *stream; - ptrdiff_t size; - ptrdiff_t pos; - stream=(OpusMemStream *)_stream; - /*Check for empty read.*/ - if(_buf_size<=0)return 0; - size=stream->size; - pos=stream->pos; - /*Check for EOF.*/ - if(pos>=size)return 0; - /*Check for a short read.*/ - _buf_size=(int)OP_MAX(size-pos,_buf_size); - memcpy(_ptr,stream->data+pos,_buf_size); - pos+=_buf_size; - stream->pos=pos; - return _buf_size; -} - -static int op_mem_seek(void *_stream,opus_int64 _offset,int _whence){ - OpusMemStream *stream; - ptrdiff_t pos; - stream=(OpusMemStream *)_stream; - pos=stream->pos; - switch(_whence){ - case SEEK_SET:{ - /*Check for overflow:*/ - if(_offset<0||_offset>OP_MEM_DIFF_MAX)return -1; - pos=(ptrdiff_t)_offset; - }break; - case SEEK_CUR:{ - /*Check for overflow:*/ - if(_offset<-pos||_offset>OP_MEM_DIFF_MAX-pos)return -1; - pos=(ptrdiff_t)(pos+_offset); - }break; - case SEEK_END:{ - ptrdiff_t size; - size=stream->size; - OP_ASSERT(size>=0); - /*Check for overflow:*/ - if(_offset>size||_offsetpos=pos; - return 0; -} - -static opus_int64 op_mem_tell(void *_stream){ - OpusMemStream *stream; - stream=(OpusMemStream *)_stream; - return (ogg_int64_t)stream->pos; -} - -static int op_mem_close(void *_stream){ - _ogg_free(_stream); - return 0; -} - -static const OpusFileCallbacks OP_MEM_CALLBACKS={ - op_mem_read, - op_mem_seek, - op_mem_tell, - op_mem_close -}; - -void *op_mem_stream_create(OpusFileCallbacks *_cb, - const unsigned char *_data,size_t _size){ - OpusMemStream *stream; - if(_size>OP_MEM_SIZE_MAX)return NULL; - stream=(OpusMemStream *)_ogg_malloc(sizeof(*stream)); - if(stream!=NULL){ - *_cb=*&OP_MEM_CALLBACKS; - stream->data=_data; - stream->size=_size; - stream->pos=0; - } - return stream; -} diff --git a/code/opusfile-0.2/include/opusfile.h b/code/opusfile-0.5/include/opusfile.h similarity index 76% rename from code/opusfile-0.2/include/opusfile.h rename to code/opusfile-0.5/include/opusfile.h index bc7c7384..850cd6b9 100644 --- a/code/opusfile-0.2/include/opusfile.h +++ b/code/opusfile-0.5/include/opusfile.h @@ -16,7 +16,6 @@ ********************************************************************/ #if !defined(_opusfile_h) # define _opusfile_h (1) -# include /**\mainpage \section Introduction @@ -50,17 +49,68 @@ Several additional sections are not tied to the main API. - \ref stream_callbacks - \ref header_info - - \ref error_codes*/ + - \ref error_codes + \section Overview + + The libopusfile API always decodes files to 48 kHz. + The original sample rate is not preserved by the lossy compression, though + it is stored in the header to allow you to resample to it after decoding + (the libopusfile API does not currently provide a resampler, + but the + the + Speex resampler is a good choice if you need one). + In general, if you are playing back the audio, you should leave it at + 48 kHz, provided your audio hardware supports it. + When decoding to a file, it may be worth resampling back to the original + sample rate, so as not to surprise users who might not expect the sample + rate to change after encoding to Opus and decoding. + + Opus files can contain anywhere from 1 to 255 channels of audio. + The channel mappings for up to 8 channels are the same as the + Vorbis + mappings. + A special stereo API can convert everything to 2 channels, making it simple + to support multichannel files in an application which only has stereo + output. + Although the libopusfile ABI provides support for the theoretical + maximum number of channels, the current implementation does not support + files with more than 8 channels, as they do not have well-defined channel + mappings. + + Like all Ogg files, Opus files may be "chained". + That is, multiple Opus files may be combined into a single, longer file just + by concatenating the original files. + This is commonly done in internet radio streaming, as it allows the title + and artist to be updated each time the song changes, since each link in the + chain includes its own set of metadata. + + libopusfile fully supports chained files. + It will decode the first Opus stream found in each link of a chained file + (ignoring any other streams that might be concurrently multiplexed with it, + such as a video stream). + + The channel count can also change between links. + If your application is not prepared to deal with this, it can use the stereo + API to ensure the audio from all links will always get decoded into a + common format. + Since libopusfile always decodes to 48 kHz, you do not have to + worry about the sample rate changing between links (as was possible with + Vorbis). + This makes application support for chained files with libopusfile + very easy.*/ # if defined(__cplusplus) extern "C" { # endif +# include # include # include # include +/**@cond PRIVATE*/ + /*Enable special features for gcc and gcc-compatible compilers.*/ # if !defined(OP_GNUC_PREREQ) # if defined(__GNUC__)&&defined(__GNUC_MINOR__) @@ -75,9 +125,12 @@ extern "C" { # pragma GCC visibility push(default) # endif -typedef struct OpusHead OpusHead; -typedef struct OpusTags OpusTags; -typedef struct OggOpusFile OggOpusFile; +typedef struct OpusHead OpusHead; +typedef struct OpusTags OpusTags; +typedef struct OpusPictureTag OpusPictureTag; +typedef struct OpusServerInfo OpusServerInfo; +typedef struct OpusFileCallbacks OpusFileCallbacks; +typedef struct OggOpusFile OggOpusFile; /*Warning attributes for libopusfile functions.*/ # if OP_GNUC_PREREQ(3,4) @@ -91,6 +144,8 @@ typedef struct OggOpusFile OggOpusFile; # define OP_ARG_NONNULL(_x) # endif +/**@endcond*/ + /**\defgroup error_codes Error Codes*/ /*@{*/ /**\name List of possible error codes @@ -182,8 +237,9 @@ struct OpusHead{ opus_uint32 input_sample_rate; /**The gain to apply to the decoded output, in dB, as a Q8 value in the range -32768...32767. - The decoder will automatically scale the output by - pow(10,output_gain/(20.0*256)).*/ + The libopusfile API will automatically apply this gain to the + decoded output before returning it, scaling it by + pow(10,output_gain/(20.0*256)).*/ int output_gain; /**The channel mapping family, in the range 0...255. Channel mapping family 0 covers mono or stereo in a single stream. @@ -253,6 +309,87 @@ struct OpusTags{ char *vendor; }; +/**\name Picture tag image formats*/ +/*@{*/ + +/**The MIME type was not recognized, or the image data did not match the + declared MIME type.*/ +#define OP_PIC_FORMAT_UNKNOWN (-1) +/**The MIME type indicates the image data is really a URL.*/ +#define OP_PIC_FORMAT_URL (0) +/**The image is a JPEG.*/ +#define OP_PIC_FORMAT_JPEG (1) +/**The image is a PNG.*/ +#define OP_PIC_FORMAT_PNG (2) +/**The image is a GIF.*/ +#define OP_PIC_FORMAT_GIF (3) + +/*@}*/ + +/**The contents of a METADATA_BLOCK_PICTURE tag.*/ +struct OpusPictureTag{ + /**The picture type according to the ID3v2 APIC frame: +
    +
  1. Other
    2. +
  2. 32x32 pixels 'file icon' (PNG only)
    3. +
  3. Other file icon
    4. +
  4. Cover (front)
    5. +
  5. Cover (back)
    6. +
  6. Leaflet page
    7. +
  7. Media (e.g. label side of CD)
    8. +
  8. Lead artist/lead performer/soloist
    9. +
  9. Artist/performer
    10. +
  10. Conductor
    11. +
  11. Band/Orchestra
    12. +
  12. Composer
    13. +
  13. Lyricist/text writer
    14. +
  14. Recording Location
    15. +
  15. During recording
    16. +
  16. During performance
    17. +
  17. Movie/video screen capture
    18. +
  18. A bright colored fish
    19. +
  19. Illustration
    20. +
  20. Band/artist logotype
    21. +
  21. Publisher/Studio logotype
    22. +
+ Others are reserved and should not be used. + There may only be one each of picture type 1 and 2 in a file.*/ + opus_int32 type; + /**The MIME type of the picture, in printable ASCII characters 0x20-0x7E. + The MIME type may also be "-->" to signify that the data part + is a URL pointing to the picture instead of the picture data itself. + In this case, a terminating NUL is appended to the URL string in #data, + but #data_length is set to the length of the string excluding that + terminating NUL.*/ + char *mime_type; + /**The description of the picture, in UTF-8.*/ + char *description; + /**The width of the picture in pixels.*/ + opus_uint32 width; + /**The height of the picture in pixels.*/ + opus_uint32 height; + /**The color depth of the picture in bits-per-pixel (not + bits-per-channel).*/ + opus_uint32 depth; + /**For indexed-color pictures (e.g., GIF), the number of colors used, or 0 + for non-indexed pictures.*/ + opus_uint32 colors; + /**The length of the picture data in bytes.*/ + opus_uint32 data_length; + /**The binary picture data.*/ + unsigned char *data; + /**The format of the picture data, if known. + One of +
    +
  • #OP_PIC_FORMAT_UNKNOWN,
    • +
  • #OP_PIC_FORMAT_URL,
    • +
  • #OP_PIC_FORMAT_JPEG,
    • +
  • #OP_PIC_FORMAT_PNG, or
    • +
  • #OP_PIC_FORMAT_GIF.
    • +
*/ + int format; +}; + /**\name Functions for manipulating header data These functions manipulate the #OpusHead and #OpusTags structures, @@ -315,7 +452,7 @@ ogg_int64_t opus_granule_sample(const OpusHead *_head,ogg_int64_t _gp) for validity. \param[in] _data The contents of the 'comment' header packet. \param _len The number of bytes of data in the 'info' header packet. - \retval 0 Success. + \retval 0 Success. \retval #OP_ENOTFORMAT If the data does not start with the "OpusTags" string. \retval #OP_EBADHEADER If the contents of the packet otherwise violate the @@ -324,6 +461,15 @@ ogg_int64_t opus_granule_sample(const OpusHead *_head,ogg_int64_t _gp) OP_WARN_UNUSED_RESULT int opus_tags_parse(OpusTags *_tags, const unsigned char *_data,size_t _len) OP_ARG_NONNULL(2); +/**Performs a deep copy of an #OpusTags structure. + \param _dst The #OpusTags structure to copy into. + If this function fails, the contents of this structure remain + untouched. + \param _src The #OpusTags structure to copy from. + \retval 0 Success. + \retval #OP_EFAULT If there wasn't enough memory to copy the tags.*/ +int opus_tags_copy(OpusTags *_dst,const OpusTags *_src) OP_ARG_NONNULL(1); + /**Initializes an #OpusTags structure. This should be called on a freshly allocated #OpusTags structure before attempting to use it. @@ -385,6 +531,24 @@ const char *opus_tags_query(const OpusTags *_tags,const char *_tag,int _count) int opus_tags_query_count(const OpusTags *_tags,const char *_tag) OP_ARG_NONNULL(1) OP_ARG_NONNULL(2); +/**Get the track gain from an R128_TRACK_GAIN tag, if one was specified. + This searches for the first R128_TRACK_GAIN tag with a valid signed, + 16-bit decimal integer value and returns the value. + This routine is exposed merely for convenience for applications which wish + to do something special with the track gain (i.e., display it). + If you simply wish to apply the track gain instead of the header gain, you + can use op_set_gain_offset() with an #OP_TRACK_GAIN type and no offset. + \param _tags An initialized #OpusTags structure. + \param[out] _gain_q8 The track gain, in 1/256ths of a dB. + This will lie in the range [-32768,32767], and should + be applied in addition to the header gain. + On error, no value is returned, and the previous + contents remain unchanged. + \return 0 on success, or a negative value on error. + \retval #OP_FALSE There was no track gain available in the given tags.*/ +int opus_tags_get_track_gain(const OpusTags *_tags,int *_gain_q8) + OP_ARG_NONNULL(1) OP_ARG_NONNULL(2); + /**Clears the #OpusTags structure. This should be called on an #OpusTags structure after it is no longer needed. @@ -392,6 +556,78 @@ int opus_tags_query_count(const OpusTags *_tags,const char *_tag) \param _tags The #OpusTags structure to clear.*/ void opus_tags_clear(OpusTags *_tags) OP_ARG_NONNULL(1); +/**Check if \a _comment is an instance of a \a _tag_name tag. + \see opus_tagncompare + \param _tag_name A NUL-terminated, case-insensitive, ASCII string containing + the name of the tag to check for (without the terminating + '=' character). + \param _comment The comment string to check. + \return An integer less than, equal to, or greater than zero if \a _comment + is found respectively, to be less than, to match, or be greater + than a "tag=value" string whose tag matches \a _tag_name.*/ +int opus_tagcompare(const char *_tag_name,const char *_comment); + +/**Check if \a _comment is an instance of a \a _tag_name tag. + This version is slightly more efficient than opus_tagcompare() if the length + of the tag name is already known (e.g., because it is a constant). + \see opus_tagcompare + \param _tag_name A case-insensitive ASCII string containing the name of the + tag to check for (without the terminating '=' character). + \param _tag_len The number of characters in the tag name. + This must be non-negative. + \param _comment The comment string to check. + \return An integer less than, equal to, or greater than zero if \a _comment + is found respectively, to be less than, to match, or be greater + than a "tag=value" string whose tag matches the first \a _tag_len + characters of \a _tag_name.*/ +int opus_tagncompare(const char *_tag_name,int _tag_len,const char *_comment); + +/**Parse a single METADATA_BLOCK_PICTURE tag. + This decodes the BASE64-encoded content of the tag and returns a structure + with the MIME type, description, image parameters (if known), and the + compressed image data. + If the MIME type indicates the presence of an image format we recognize + (JPEG, PNG, or GIF) and the actual image data contains the magic signature + associated with that format, then the OpusPictureTag::format field will be + set to the corresponding format. + This is provided as a convenience to avoid requiring applications to parse + the MIME type and/or do their own format detection for the commonly used + formats. + In this case, we also attempt to extract the image parameters directly from + the image data (overriding any that were present in the tag, which the + specification says applications are not meant to rely on). + The application must still provide its own support for actually decoding the + image data and, if applicable, retrieving that data from URLs. + \param[out] _pic Returns the parsed picture data. + No sanitation is done on the type, MIME type, or + description fields, so these might return invalid values. + The contents of this structure are left unmodified on + failure. + \param _tag The METADATA_BLOCK_PICTURE tag contents. + The leading "METADATA_BLOCK_PICTURE=" portion is optional, + to allow the function to be used on either directly on the + values in OpusTags::user_comments or on the return value + of opus_tags_query(). + \return 0 on success or a negative value on error. + \retval #OP_ENOTFORMAT The METADATA_BLOCK_PICTURE contents were not valid. + \retval #OP_EFAULT There was not enough memory to store the picture tag + contents.*/ +OP_WARN_UNUSED_RESULT int opus_picture_tag_parse(OpusPictureTag *_pic, + const char *_tag) OP_ARG_NONNULL(1) OP_ARG_NONNULL(2); + +/**Initializes an #OpusPictureTag structure. + This should be called on a freshly allocated #OpusPictureTag structure + before attempting to use it. + \param _pic The #OpusPictureTag structure to initialize.*/ +void opus_picture_tag_init(OpusPictureTag *_pic) OP_ARG_NONNULL(1); + +/**Clears the #OpusPictureTag structure. + This should be called on an #OpusPictureTag structure after it is no longer + needed. + It will free all memory used by the structure members. + \param _pic The #OpusPictureTag structure to clear.*/ +void opus_picture_tag_clear(OpusPictureTag *_pic) OP_ARG_NONNULL(1); + /*@}*/ /*@}*/ @@ -408,6 +644,8 @@ void opus_tags_clear(OpusTags *_tags) OP_ARG_NONNULL(1); They may be expanded in the future.*/ /*@{*/ +/**@cond PRIVATE*/ + /*These are the raw numbers used to define the request codes. They should not be used directly.*/ #define OP_SSL_SKIP_CERTIFICATE_CHECK_REQUEST (6464) @@ -415,6 +653,7 @@ void opus_tags_clear(OpusTags *_tags) OP_ARG_NONNULL(1); #define OP_HTTP_PROXY_PORT_REQUEST (6592) #define OP_HTTP_PROXY_USER_REQUEST (6656) #define OP_HTTP_PROXY_PASS_REQUEST (6720) +#define OP_GET_SERVER_INFO_REQUEST (6784) #define OP_URL_OPT(_request) ((_request)+(char *)0) @@ -422,6 +661,64 @@ void opus_tags_clear(OpusTags *_tags) OP_ARG_NONNULL(1); provided to one of the URL options.*/ #define OP_CHECK_INT(_x) ((void)((_x)==(opus_int32)0),(opus_int32)(_x)) #define OP_CHECK_CONST_CHAR_PTR(_x) ((_x)+((_x)-(const char *)(_x))) +#define OP_CHECK_SERVER_INFO_PTR(_x) ((_x)+((_x)-(OpusServerInfo *)(_x))) + +/**@endcond*/ + +/**HTTP/Shoutcast/Icecast server information associated with a URL.*/ +struct OpusServerInfo{ + /**The name of the server (icy-name/ice-name). + This is NULL if there was no icy-name or + ice-name header.*/ + char *name; + /**A short description of the server (icy-description/ice-description). + This is NULL if there was no icy-description or + ice-description header.*/ + char *description; + /**The genre the server falls under (icy-genre/ice-genre). + This is NULL if there was no icy-genre or + ice-genre header.*/ + char *genre; + /**The homepage for the server (icy-url/ice-url). + This is NULL if there was no icy-url or + ice-url header.*/ + char *url; + /**The software used by the origin server (Server). + This is NULL if there was no Server header.*/ + char *server; + /**The media type of the entity sent to the recepient (Content-Type). + This is NULL if there was no Content-Type + header.*/ + char *content_type; + /**The nominal stream bitrate in kbps (icy-br/ice-bitrate). + This is -1 if there was no icy-br or + ice-bitrate header.*/ + opus_int32 bitrate_kbps; + /**Flag indicating whether the server is public (1) or not + (0) (icy-pub/ice-public). + This is -1 if there was no icy-pub or + ice-public header.*/ + int is_public; + /**Flag indicating whether the server is using HTTPS instead of HTTP. + This is 0 unless HTTPS is being used. + This may not match the protocol used in the original URL if there were + redirections.*/ + int is_ssl; +}; + +/**Initializes an #OpusServerInfo structure. + All fields are set as if the corresponding header was not available. + \param _info The #OpusServerInfo structure to initialize. + \note If you use this function, you must link against libopusurl.*/ +void opus_server_info_init(OpusServerInfo *_info) OP_ARG_NONNULL(1); + +/**Clears the #OpusServerInfo structure. + This should be called on an #OpusServerInfo structure after it is no longer + needed. + It will free all memory used by the structure members. + \param _info The #OpusServerInfo structure to clear. + \note If you use this function, you must link against libopusurl.*/ +void opus_server_info_clear(OpusServerInfo *_info) OP_ARG_NONNULL(1); /**Skip the certificate check when connecting via TLS/SSL (https). \param _b opus_int32: Whether or not to skip the certificate @@ -467,7 +764,7 @@ void opus_tags_clear(OpusTags *_tags) OP_ARG_NONNULL(1); arguments. \hideinitializer*/ #define OP_HTTP_PROXY_USER(_user) \ - OP_URL_OPT(OP_HTTP_PROXY_USER_REQUEST),OP_CHECK_CONST_CHAR_PTR(_host) + OP_URL_OPT(OP_HTTP_PROXY_USER_REQUEST),OP_CHECK_CONST_CHAR_PTR(_user) /**Use the given password for authentication when proxying connections. All proxy parameters are ignored for non-http and non-https URLs. @@ -480,7 +777,28 @@ void opus_tags_clear(OpusTags *_tags) OP_ARG_NONNULL(1); arguments. \hideinitializer*/ #define OP_HTTP_PROXY_PASS(_pass) \ - OP_URL_OPT(OP_HTTP_PROXY_PASS_REQUEST),OP_CHECK_CONST_CHAR_PTR(_host) + OP_URL_OPT(OP_HTTP_PROXY_PASS_REQUEST),OP_CHECK_CONST_CHAR_PTR(_pass) + +/**Parse information about the streaming server (if any) and return it. + Very little validation is done. + In particular, OpusServerInfo::url may not be a valid URL, + OpusServerInfo::bitrate_kbps may not really be in kbps, and + OpusServerInfo::content_type may not be a valid MIME type. + The character set of the string fields is not specified anywhere, and should + not be assumed to be valid UTF-8. + \param _info OpusServerInfo *: Returns information about the server. + If there is any error opening the stream, the + contents of this structure remain + unmodified. + On success, fills in the structure with the + server information that was available, if + any. + After a successful return, the contents of + this structure should be freed by calling + opus_server_info_clear(). + \hideinitializer*/ +#define OP_GET_SERVER_INFO(_info) \ + OP_URL_OPT(OP_GET_SERVER_INFO_REQUEST),OP_CHECK_SERVER_INFO_PTR(_info) /*@}*/ /*@}*/ @@ -497,8 +815,6 @@ void opus_tags_clear(OpusTags *_tags) OP_ARG_NONNULL(1); block of memory, or URLs.*/ /*@{*/ -typedef struct OpusFileCallbacks OpusFileCallbacks; - /**Reads up to \a _nbytes bytes of data from \a _stream. \param _stream The stream to read from. \param[out] _ptr The buffer to store the data in. @@ -564,6 +880,10 @@ struct OpusFileCallbacks{ If there is an error opening the file, nothing will be filled in here. \param _path The path to the file to open. + On Windows, this string must be UTF-8 (to allow access to + files whose names cannot be represented in the current + MBCS code page). + All other systems use the native character encoding. \param _mode The mode to open the file in. \return A stream handle to use with the callbacks, or NULL on error.*/ @@ -597,6 +917,10 @@ OP_WARN_UNUSED_RESULT void *op_fdopen(OpusFileCallbacks *_cb, If there is an error opening the file, nothing will be filled in here. \param _path The path to the file to open. + On Windows, this string must be UTF-8 (to allow access + to files whose names cannot be represented in the + current MBCS code page). + All other systems use the native character encoding. \param _mode The mode to open the file in. \param _stream A stream previously returned by op_fopen(), op_fdopen(), or op_freopen(). @@ -624,6 +948,7 @@ OP_WARN_UNUSED_RESULT void *op_mem_stream_create(OpusFileCallbacks *_cb, takes a va_list instead of a variable number of arguments. It does not call the va_end macro, and because it invokes the va_arg macro, the value of \a _ap is undefined after the call. + \note If you use this function, you must link against libopusurl. \param[out] _cb The callbacks to use for this stream. If there is an error creating the stream, nothing will be filled in here. @@ -632,6 +957,10 @@ OP_WARN_UNUSED_RESULT void *op_mem_stream_create(OpusFileCallbacks *_cb, schemes are supported. Both and may be disabled at compile time, in which case opening such URLs will always fail. + Currently this only supports URIs. + IRIs should be converted to UTF-8 and URL-escaped, with + internationalized domain names encoded in punycode, + before passing them to this function. \param[in,out] _ap A list of the \ref url_options "optional flags" to use. This is a variable-length list of options terminated with NULL. @@ -640,7 +969,8 @@ OP_WARN_UNUSED_RESULT void *op_mem_stream_create(OpusFileCallbacks *_cb, OP_WARN_UNUSED_RESULT void *op_url_stream_vcreate(OpusFileCallbacks *_cb, const char *_url,va_list _ap) OP_ARG_NONNULL(1) OP_ARG_NONNULL(2); -/**Creates a stream that reads from the given URL using the specified proxy. +/**Creates a stream that reads from the given URL. + \note If you use this function, you must link against libopusurl. \param[out] _cb The callbacks to use for this stream. If there is an error creating the stream, nothing will be filled in here. @@ -649,6 +979,10 @@ OP_WARN_UNUSED_RESULT void *op_url_stream_vcreate(OpusFileCallbacks *_cb, are supported. Both and may be disabled at compile time, in which case opening such URLs will always fail. + Currently this only supports URIs. + IRIs should be converted to UTF-8 and URL-escaped, with + internationalized domain names encoded in punycode, before + passing them to this function. \param ... The \ref url_options "optional flags" to use. This is a variable-length list of options terminated with NULL. @@ -731,12 +1065,17 @@ OP_WARN_UNUSED_RESULT OggOpusFile *op_open_memory(const unsigned char *_data, takes a va_list instead of a variable number of arguments. It does not call the va_end macro, and because it invokes the va_arg macro, the value of \a _ap is undefined after the call. + \note If you use this function, you must link against libopusurl. \param _url The URL to open. Currently only the , , and schemes are supported. Both and may be disabled at compile time, in which case opening such URLs will always fail. + Currently this only supports URIs. + IRIs should be converted to UTF-8 and URL-escaped, + with internationalized domain names encoded in + punycode, before passing them to this function. \param[out] _error Returns 0 on success, or a failure code on error. You may pass in NULL if you don't want the failure code. @@ -751,13 +1090,16 @@ OP_WARN_UNUSED_RESULT OggOpusFile *op_vopen_url(const char *_url, int *_error,va_list _ap) OP_ARG_NONNULL(1); /**Open a stream from a URL. - However, this approach will not work for live streams or HTTP/1.0 servers - (which do not support Range requets). + \note If you use this function, you must link against libopusurl. \param _url The URL to open. Currently only the , , and schemes are supported. Both and may be disabled at compile time, in which case opening such URLs will always fail. + Currently this only supports URIs. + IRIs should be converted to UTF-8 and URL-escaped, with + internationalized domain names encoded in punycode, + before passing them to this function. \param[out] _error Returns 0 on success, or a failure code on error. You may pass in NULL if you don't want the failure code. @@ -841,7 +1183,11 @@ OP_WARN_UNUSED_RESULT OggOpusFile *op_open_url(const char *_url,
The first or last timestamp in a link failed basic validity checks.
- \return A freshly opened \c OggOpusFile, or NULL on error.*/ + \return A freshly opened \c OggOpusFile, or NULL on error. + libopusfile does not take ownership of the source + if the call fails. + The calling application is responsible for closing the source if + this call returns an error.*/ OP_WARN_UNUSED_RESULT OggOpusFile *op_open_callbacks(void *_source, const OpusFileCallbacks *_cb,const unsigned char *_initial_data, size_t _initial_bytes,int *_error) OP_ARG_NONNULL(2); @@ -876,6 +1222,7 @@ OP_WARN_UNUSED_RESULT OggOpusFile *op_test_memory(const unsigned char *_data, takes a va_list instead of a variable number of arguments. It does not call the va_end macro, and because it invokes the va_arg macro, the value of \a _ap is undefined after the call. + \note If you use this function, you must link against libopusurl. \see op_test_url \see op_test_callbacks \param _url The URL to open. @@ -884,6 +1231,10 @@ OP_WARN_UNUSED_RESULT OggOpusFile *op_test_memory(const unsigned char *_data, Both and may be disabled at compile time, in which case opening such URLs will always fail. + Currently this only supports URIs. + IRIs should be converted to UTF-8 and URL-escaped, + with internationalized domain names encoded in + punycode, before passing them to this function. \param[out] _error Returns 0 on success, or a failure code on error. You may pass in NULL if you don't want the failure code. @@ -898,12 +1249,17 @@ OP_WARN_UNUSED_RESULT OggOpusFile *op_vtest_url(const char *_url, int *_error,va_list _ap) OP_ARG_NONNULL(1); /**Partially open a stream from a URL. + \note If you use this function, you must link against libopusurl. \see op_test_callbacks \param _url The URL to open. Currently only the , , and schemes are supported. Both and may be disabled at compile time, in which case opening such URLs will always fail. + Currently this only supports URIs. + IRIs should be converted to UTF-8 and URL-escaped, with + internationalized domain names encoded in punycode, + before passing them to this function. \param[out] _error Returns 0 on success, or a failure code on error. You may pass in NULL if you don't want the failure code. @@ -974,7 +1330,11 @@ OP_WARN_UNUSED_RESULT OggOpusFile *op_test_url(const char *_url, the failure code. See op_open_callbacks() for a full list of failure codes. - \return A partially opened \c OggOpusFile, or NULL on error.*/ + \return A partially opened \c OggOpusFile, or NULL on error. + libopusfile does not take ownership of the source + if the call fails. + The calling application is responsible for closing the source if + this call returns an error.*/ OP_WARN_UNUSED_RESULT OggOpusFile *op_test_callbacks(void *_source, const OpusFileCallbacks *_cb,const unsigned char *_initial_data, size_t _initial_bytes,int *_error) OP_ARG_NONNULL(2); @@ -1046,7 +1406,7 @@ void op_free(OggOpusFile *_of); This function may be called on partially-opened streams. \param _of The \c OggOpusFile whose seekable status is to be returned. \return A non-zero value if seekable, and 0 if unseekable.*/ -int op_seekable(OggOpusFile *_of) OP_ARG_NONNULL(1); +int op_seekable(const OggOpusFile *_of) OP_ARG_NONNULL(1); /**Returns the number of links in this chained stream. This function may be called on partially-opened streams, but it will always @@ -1054,9 +1414,9 @@ int op_seekable(OggOpusFile *_of) OP_ARG_NONNULL(1); The actual number of links is not known until the stream is fully opened. \param _of The \c OggOpusFile from which to retrieve the link count. \return For fully-open seekable sources, this returns the total number of - links in the whole stream. + links in the whole stream, which will be at least 1. For partially-open or unseekable sources, this always returns 1.*/ -int op_link_count(OggOpusFile *_of) OP_ARG_NONNULL(1); +int op_link_count(const OggOpusFile *_of) OP_ARG_NONNULL(1); /**Get the serial number of the given link in a (possibly-chained) Ogg Opus stream. @@ -1071,7 +1431,7 @@ int op_link_count(OggOpusFile *_of) OP_ARG_NONNULL(1); the serial number of the last link. If the source is not seekable, this always returns the serial number of the current link.*/ -opus_uint32 op_serialno(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); +opus_uint32 op_serialno(const OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); /**Get the channel count of the given link in a (possibly-chained) Ogg Opus stream. @@ -1088,7 +1448,7 @@ opus_uint32 op_serialno(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); the channel count of the last link. If the source is not seekable, this always returns the channel count of the current link.*/ -int op_channel_count(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); +int op_channel_count(const OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); /**Get the total (compressed) size of the stream, or of an individual link in a (possibly-chained) Ogg Opus stream, including all headers and Ogg muxing @@ -1106,7 +1466,7 @@ int op_channel_count(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); \retval #OP_EINVAL The source is not seekable (so we can't know the length), \a _li wasn't less than the total number of links in the stream, or the stream was only partially open.*/ -opus_int64 op_raw_total(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); +opus_int64 op_raw_total(const OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); /**Get the total PCM length (number of samples at 48 kHz) of the stream, or of an individual link in a (possibly-chained) Ogg Opus stream. @@ -1124,7 +1484,7 @@ opus_int64 op_raw_total(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); \retval #OP_EINVAL The source is not seekable (so we can't know the length), \a _li wasn't less than the total number of links in the stream, or the stream was only partially open.*/ -ogg_int64_t op_pcm_total(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); +ogg_int64_t op_pcm_total(const OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); /**Get the ID header information for the given link in a (possibly chained) Ogg Opus stream. @@ -1140,7 +1500,7 @@ ogg_int64_t op_pcm_total(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); information for the current link is always returned, if available. \return The contents of the ID header for the given link.*/ -const OpusHead *op_head(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); +const OpusHead *op_head(const OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); /**Get the comment header information for the given link in a (possibly chained) Ogg Opus stream. @@ -1158,7 +1518,7 @@ const OpusHead *op_head(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); \return The contents of the comment header for the given link, or NULL if this is an unseekable stream that encountered an invalid link.*/ -const OpusTags *op_tags(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); +const OpusTags *op_tags(const OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); /**Retrieve the index of the current link. This is the link that produced the data most recently read by @@ -1175,7 +1535,7 @@ const OpusTags *op_tags(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); each time a new link is encountered (even though op_link_count() always returns 1). \retval #OP_EINVAL The stream was only partially open.*/ -int op_current_link(OggOpusFile *_of) OP_ARG_NONNULL(1); +int op_current_link(const OggOpusFile *_of) OP_ARG_NONNULL(1); /**Computes the bitrate for a given link in a (possibly chained) Ogg Opus stream. @@ -1188,7 +1548,7 @@ int op_current_link(OggOpusFile *_of) OP_ARG_NONNULL(1); \retval #OP_EINVAL The stream was only partially open, the stream was not seekable, or \a _li was larger than the number of links.*/ -opus_int32 op_bitrate(OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); +opus_int32 op_bitrate(const OggOpusFile *_of,int _li) OP_ARG_NONNULL(1); /**Compute the instantaneous bitrate, measured as the ratio of bits to playable samples decoded since a) the last call to op_bitrate_instant(), b) the last @@ -1207,7 +1567,7 @@ opus_int32 op_bitrate_instant(OggOpusFile *_of) OP_ARG_NONNULL(1); \param _of The \c OggOpusFile from which to retrieve the position indicator. \return The byte position that is currently being read from. \retval #OP_EINVAL The stream was only partially open.*/ -opus_int64 op_raw_tell(OggOpusFile *_of) OP_ARG_NONNULL(1); +opus_int64 op_raw_tell(const OggOpusFile *_of) OP_ARG_NONNULL(1); /**Obtain the PCM offset of the next sample to be read. If the stream is not properly timestamped, this might not increment by the @@ -1216,7 +1576,7 @@ opus_int64 op_raw_tell(OggOpusFile *_of) OP_ARG_NONNULL(1); \param _of The \c OggOpusFile from which to retrieve the PCM offset. \return The PCM offset of the next sample to be read. \retval #OP_EINVAL The stream was only partially open.*/ -ogg_int64_t op_pcm_tell(OggOpusFile *_of) OP_ARG_NONNULL(1); +ogg_int64_t op_pcm_tell(const OggOpusFile *_of) OP_ARG_NONNULL(1); /*@}*/ /*@}*/ @@ -1303,11 +1663,12 @@ int op_pcm_seek(OggOpusFile *_of,ogg_int64_t _pcm_offset) OP_ARG_NONNULL(1); clipping and other issues which might be avoided entirely if, e.g., you scale down the volume at some other stage. However, if you intend to direct consume 16-bit samples, the conversion in - libopusfile provides noise-shaping dithering API. + libopusfile provides noise-shaping dithering and, if compiled + against libopus 1.1 or later, soft-clipping prevention. libopusfile can also be configured at compile time to use the fixed-point libopus API. - If so, the floating-point API may also be disabled. + If so, libopusfile's floating-point API may also be disabled. In that configuration, nothing in libopusfile will use any floating-point operations, to simplify support on devices without an adequate FPU. @@ -1316,20 +1677,125 @@ int op_pcm_seek(OggOpusFile *_of,ogg_int64_t _pcm_offset) OP_ARG_NONNULL(1); not check the error return code from op_read_float() or its associated functions. If the remote peer does not close the connection gracefully (with a TLS - "close notify" message), these functions will return OP_EREAD instead of 0 + "close notify" message), these functions will return #OP_EREAD instead of 0 when they reach the end of the file. If you are reading from an URL (particularly if seeking is not supported), you should make sure to check for this error and warn the user appropriately.*/ /*@{*/ +/**Indicates that the decoding callback should produce signed 16-bit + native-endian output samples.*/ +#define OP_DEC_FORMAT_SHORT (7008) +/**Indicates that the decoding callback should produce 32-bit native-endian + float samples.*/ +#define OP_DEC_FORMAT_FLOAT (7040) + +/**Indicates that the decoding callback did not decode anything, and that + libopusfile should decode normally instead.*/ +#define OP_DEC_USE_DEFAULT (6720) + +/**Called to decode an Opus packet. + This should invoke the functional equivalent of opus_multistream_decode() or + opus_multistream_decode_float(), except that it returns 0 on success + instead of the number of decoded samples (which is known a priori). + \param _ctx The application-provided callback context. + \param _decoder The decoder to use to decode the packet. + \param[out] _pcm The buffer to decode into. + This will always have enough room for \a _nchannels of + \a _nsamples samples, which should be placed into this + buffer interleaved. + \param _op The packet to decode. + This will always have its granule position set to a valid + value. + \param _nsamples The number of samples expected from the packet. + \param _nchannels The number of channels expected from the packet. + \param _format The desired sample output format. + This is either #OP_DEC_FORMAT_SHORT or + #OP_DEC_FORMAT_FLOAT. + \param _li The index of the link from which this packet was decoded. + \return A non-negative value on success, or a negative value on error. + The error codes should be the same as those returned by + opus_multistream_decode() or opus_multistream_decode_float(). + \retval 0 Decoding was successful. + The application has filled the buffer with + exactly \a _nsamples*\a + _nchannels samples in the requested + format. + \retval #OP_DEC_USE_DEFAULT No decoding was done. + libopusfile should decode normally + instead.*/ +typedef int (*op_decode_cb_func)(void *_ctx,OpusMSDecoder *_decoder,void *_pcm, + const ogg_packet *_op,int _nsamples,int _nchannels,int _format,int _li); + +/**Sets the packet decode callback function. + This is called once for each packet that needs to be decoded. + A call to this function is no guarantee that the audio will eventually be + delivered to the application. + Some or all of the data from the packet may be discarded (i.e., at the + beginning or end of a link, or after a seek), however the callback is + required to provide all of it. + \param _of The \c OggOpusFile on which to set the decode callback. + \param _decode_cb The callback function to call. + This may be NULL to disable calling the + callback. + \param _ctx The application-provided context pointer to pass to the + callback on each call.*/ +void op_set_decode_callback(OggOpusFile *_of, + op_decode_cb_func _decode_cb,void *_ctx) OP_ARG_NONNULL(1); + +/**Gain offset type that indicates that the provided offset is relative to the + header gain. + This is the default.*/ +#define OP_HEADER_GAIN (0) + +/**Gain offset type that indicates that the provided offset is relative to the + R128_TRACK_GAIN value (if any), in addition to the header gain.*/ +#define OP_TRACK_GAIN (3008) + +/**Gain offset type that indicates that the provided offset should be used as + the gain directly, without applying any the header or track gains.*/ +#define OP_ABSOLUTE_GAIN (3009) + +/**Sets the gain to be used for decoded output. + By default, the gain in the header is applied with no additional offset. + The total gain (including header gain and/or track gain, if applicable, and + this offset), will be clamped to [-32768,32767]/256 dB. + This is more than enough to saturate or underflow 16-bit PCM. + \note The new gain will not be applied to any already buffered, decoded + output. + This means you cannot change it sample-by-sample, as at best it will be + updated packet-by-packet. + It is meant for setting a target volume level, rather than applying smooth + fades, etc. + \param _of The \c OggOpusFile on which to set the gain offset. + \param _gain_type One of #OP_HEADER_GAIN, #OP_TRACK_GAIN, or + #OP_ABSOLUTE_GAIN. + \param _gain_offset_q8 The gain offset to apply, in 1/256ths of a dB. + \return 0 on success or a negative value on error. + \retval #OP_EINVAL The \a _gain_type was unrecognized.*/ +int op_set_gain_offset(OggOpusFile *_of, + int _gain_type,opus_int32 _gain_offset_q8) OP_ARG_NONNULL(1); + +/**Sets whether or not dithering is enabled for 16-bit decoding. + By default, when libopusfile is compiled to use floating-point + internally, calling op_read() or op_read_stereo() will first decode to + float, and then convert to fixed-point using noise-shaping dithering. + This flag can be used to disable that dithering. + When the application uses op_read_float() or op_read_float_stereo(), or when + the library has been compiled to decode directly to fixed point, this flag + has no effect. + \param _of The \c OggOpusFile on which to enable or disable dithering. + \param _enabled A non-zero value to enable dithering, or 0 to disable it.*/ +void op_set_dither_enabled(OggOpusFile *_of,int _enabled) OP_ARG_NONNULL(1); + /**Reads more samples from the stream. \note Although \a _buf_size must indicate the total number of values that can be stored in \a _pcm, the return value is the number of samples per channel. This is done because
    -
  1. The channel count cannot be known a prior (reading more samples might +
  2. The channel count cannot be known a priori (reading more samples might advance us into the next link, with a different channel count), so \a _buf_size cannot also be in units of samples per channel,
  3. Returning the samples per channel matches the libopus API @@ -1346,14 +1812,14 @@ int op_pcm_seek(OggOpusFile *_of,ogg_int64_t _pcm_offset) OP_ARG_NONNULL(1);
\param _of The \c OggOpusFile from which to read. \param[out] _pcm A buffer in which to store the output PCM samples, as - signed native-endian 16-bit values with a nominal - range of [-32768,32767). + signed native-endian 16-bit values at 48 kHz + with a nominal range of [-32768,32767). Multiple channels are interleaved using the Vorbis channel ordering. This must have room for at least \a _buf_size values. \param _buf_size The number of values that can be stored in \a _pcm. - It is reccommended that this be large enough for at + It is recommended that this be large enough for at least 120 ms of data at 48 kHz per channel (5760 values per channel). Smaller buffers will simply return less data, possibly @@ -1411,7 +1877,7 @@ OP_WARN_UNUSED_RESULT int op_read(OggOpusFile *_of, can be stored in \a _pcm, the return value is the number of samples per channel.
    -
  1. The channel count cannot be known a prior (reading more samples might +
  2. The channel count cannot be known a priori (reading more samples might advance us into the next link, with a different channel count), so \a _buf_size cannot also be in units of samples per channel,
  3. Returning the samples per channel matches the libopus API @@ -1428,14 +1894,14 @@ OP_WARN_UNUSED_RESULT int op_read(OggOpusFile *_of,
\param _of The \c OggOpusFile from which to read. \param[out] _pcm A buffer in which to store the output PCM samples as - signed floats with a nominal range of + signed floats at 48 kHz with a nominal range of [-1.0,1.0]. Multiple channels are interleaved using the Vorbis channel ordering. This must have room for at least \a _buf_size floats. \param _buf_size The number of floats that can be stored in \a _pcm. - It is reccommended that this be large enough for at + It is recommended that this be large enough for at least 120 ms of data at 48 kHz per channel (5760 samples per channel). Smaller buffers will simply return less data, possibly @@ -1497,13 +1963,13 @@ OP_WARN_UNUSED_RESULT int op_read_float(OggOpusFile *_of, op_read(). \param _of The \c OggOpusFile from which to read. \param[out] _pcm A buffer in which to store the output PCM samples, as - signed native-endian 16-bit values with a nominal - range of [-32768,32767). + signed native-endian 16-bit values at 48 kHz + with a nominal range of [-32768,32767). The left and right channels are interleaved in the buffer. This must have room for at least \a _buf_size values. \param _buf_size The number of values that can be stored in \a _pcm. - It is reccommended that this be large enough for at + It is recommended that this be large enough for at least 120 ms of data at 48 kHz per channel (11520 values total). Smaller buffers will simply return less data, possibly @@ -1558,13 +2024,13 @@ OP_WARN_UNUSED_RESULT int op_read_stereo(OggOpusFile *_of, op_read_float(). \param _of The \c OggOpusFile from which to read. \param[out] _pcm A buffer in which to store the output PCM samples, as - signed floats with a nominal range of + signed floats at 48 kHz with a nominal range of [-1.0,1.0]. The left and right channels are interleaved in the buffer. This must have room for at least \a _buf_size values. \param _buf_size The number of values that can be stored in \a _pcm. - It is reccommended that this be large enough for at + It is recommended that this be large enough for at least 120 ms of data at 48 kHz per channel (11520 values total). Smaller buffers will simply return less data, possibly diff --git a/code/opusfile-0.2/src/http.c b/code/opusfile-0.5/src/http.c similarity index 86% rename from code/opusfile-0.2/src/http.c rename to code/opusfile-0.5/src/http.c index 09580c6f..4a9eaf59 100644 --- a/code/opusfile-0.2/src/http.c +++ b/code/opusfile-0.5/src/http.c @@ -9,6 +9,10 @@ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ * * * ********************************************************************/ +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + #include "internal.h" #include #include @@ -40,7 +44,8 @@ RFC 6066: Transport Layer Security (TLS) Extensions: Extension Definitions RFC 6125: Representation and Verification of Domain-Based Application Service Identity within Internet Public Key Infrastructure Using X.509 (PKIX) - Certificates in the Context of Transport Layer Security (TLS)*/ + Certificates in the Context of Transport Layer Security (TLS) + RFC 6555: Happy Eyeballs: Success with Dual-Stack Hosts*/ typedef struct OpusParsedURL OpusParsedURL; typedef struct OpusStringBuf OpusStringBuf; @@ -56,7 +61,7 @@ static char *op_string_range_dup(const char *_start,const char *_end){ if(OP_UNLIKELY(len>=INT_MAX))return NULL; ret=(char *)_ogg_malloc(sizeof(*ret)*(len+1)); if(OP_LIKELY(ret!=NULL)){ - memcpy(ret,_start,sizeof(*ret)*(len)); + ret=(char *)memcpy(ret,_start,sizeof(*ret)*(len)); ret[len]='\0'; } return ret; @@ -205,18 +210,144 @@ static const char *op_parse_file_url(const char *_src){ } #if defined(OP_ENABLE_HTTP) -# include -# include -# include +# if defined(_WIN32) +# include +# include +# include +# include "winerrno.h" + +typedef SOCKET op_sock; + +# define OP_INVALID_SOCKET (INVALID_SOCKET) + +/*Vista and later support WSAPoll(), but we don't want to rely on that. + Instead we re-implement it badly using select(). + Unfortunately, they define a conflicting struct pollfd, so we only define our + own if it looks like that one has not already been defined.*/ +# if !defined(POLLIN) +/*Equivalent to POLLIN.*/ +# define POLLRDNORM (0x0100) +/*Priority band data can be read.*/ +# define POLLRDBAND (0x0200) +/*There is data to read.*/ +# define POLLIN (POLLRDNORM|POLLRDBAND) +/* There is urgent data to read.*/ +# define POLLPRI (0x0400) +/*Equivalent to POLLOUT.*/ +# define POLLWRNORM (0x0010) +/*Writing now will not block.*/ +# define POLLOUT (POLLWRNORM) +/*Priority data may be written.*/ +# define POLLWRBAND (0x0020) +/*Error condition (output only).*/ +# define POLLERR (0x0001) +/*Hang up (output only).*/ +# define POLLHUP (0x0002) +/*Invalid request: fd not open (output only).*/ +# define POLLNVAL (0x0004) + +struct pollfd{ + /*File descriptor.*/ + op_sock fd; + /*Requested events.*/ + short events; + /*Returned events.*/ + short revents; +}; +# endif + +/*But Winsock never defines nfds_t (it's simply hard-coded to ULONG).*/ +typedef unsigned long nfds_t; + +/*The usage of FD_SET() below is O(N^2). + This is okay because select() is limited to 64 sockets in Winsock, anyway. + In practice, we only ever call it with one or two sockets.*/ +static int op_poll_win32(struct pollfd *_fds,nfds_t _nfds,int _timeout){ + struct timeval tv; + fd_set ifds; + fd_set ofds; + fd_set efds; + nfds_t i; + int ret; + FD_ZERO(&ifds); + FD_ZERO(&ofds); + FD_ZERO(&efds); + for(i=0;i<_nfds;i++){ + _fds[i].revents=0; + if(_fds[i].events&POLLIN)FD_SET(_fds[i].fd,&ifds); + if(_fds[i].events&POLLOUT)FD_SET(_fds[i].fd,&ofds); + FD_SET(_fds[i].fd,&efds); + } + if(_timeout>=0){ + tv.tv_sec=_timeout/1000; + tv.tv_usec=(_timeout%1000)*1000; + } + ret=select(-1,&ifds,&ofds,&efds,_timeout<0?NULL:&tv); + if(ret>0){ + for(i=0;i<_nfds;i++){ + if(FD_ISSET(_fds[i].fd,&ifds))_fds[i].revents|=POLLIN; + if(FD_ISSET(_fds[i].fd,&ofds))_fds[i].revents|=POLLOUT; + /*This isn't correct: there are several different things that might have + happened to a fd in efds, but I don't know a good way to distinguish + them without more context from the caller. + It's okay, because we don't actually check any of these bits, we just + need _some_ bit set.*/ + if(FD_ISSET(_fds[i].fd,&efds))_fds[i].revents|=POLLHUP; + } + } + return ret; +} + +/*We define op_errno() to make it clear that it's not an l-value like normal + errno is.*/ +# define op_errno() (WSAGetLastError()?WSAGetLastError()-WSABASEERR:0) +# define op_reset_errno() (WSASetLastError(0)) + +/*The remaining functions don't get an op_ prefix even though they only + operate on sockets, because we don't use non-socket I/O here, and this + minimizes the changes needed to deal with Winsock.*/ +# define close(_fd) closesocket(_fd) +/*This relies on sizeof(u_long)==sizeof(int), which is always true on both + Win32 and Win64.*/ +# define ioctl(_fd,_req,_arg) ioctlsocket(_fd,_req,(u_long *)(_arg)) +# define getsockopt(_fd,_level,_name,_val,_len) \ + getsockopt(_fd,_level,_name,(char *)(_val),_len) +# define setsockopt(_fd,_level,_name,_val,_len) \ + setsockopt(_fd,_level,_name,(const char *)(_val),_len) +# define poll(_fds,_nfds,_timeout) op_poll_win32(_fds,_nfds,_timeout) + +# if defined(_MSC_VER) +typedef ptrdiff_t ssize_t; +# endif + +/*Load certificates from the built-in certificate store.*/ +int SSL_CTX_set_default_verify_paths_win32(SSL_CTX *_ssl_ctx); +# define SSL_CTX_set_default_verify_paths \ + SSL_CTX_set_default_verify_paths_win32 + +# else +/*Normal Berkeley sockets.*/ +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include + +typedef int op_sock; + +# define OP_INVALID_SOCKET (-1) + +# define op_errno() (errno) +# define op_reset_errno() (errno=0) + +# endif # include -# include -# include -# include -# include -# include -# include -# include -# include # include /*The maximum number of simultaneous connections. @@ -228,6 +359,11 @@ static const char *op_parse_file_url(const char *_src){ when seeking, and time out rapidly.*/ # define OP_NCONNS_MAX (4) +/*The amount of time before we attempt to re-resolve the host. + This is 10 minutes, as recommended in RFC 6555 for expiring cached connection + results for dual-stack hosts.*/ +# define OP_RESOLVE_CACHE_TIMEOUT_MS (10*60*(opus_int32)1000) + /*The number of redirections at which we give up. The value here is the current default in Firefox. RFC 2068 mandated a maximum of 5, but RFC 2616 relaxed that to "a client @@ -486,6 +622,8 @@ static void op_sb_clear(OpusStringBuf *_sb){ _ogg_free(_sb->buf); } +/*Make sure we have room for at least _capacity characters (plus 1 more for the + terminating NUL).*/ static int op_sb_ensure_capacity(OpusStringBuf *_sb,int _capacity){ char *buf; int cbuf; @@ -503,6 +641,8 @@ static int op_sb_ensure_capacity(OpusStringBuf *_sb,int _capacity){ return 0; } +/*Increase the capacity of the buffer, but not to more than _max_size + characters (plus 1 more for the terminating NUL).*/ static int op_sb_grow(OpusStringBuf *_sb,int _max_size){ char *buf; int cbuf; @@ -581,27 +721,33 @@ static struct addrinfo *op_resolve(const char *_host,unsigned _port){ char service[6]; memset(&hints,0,sizeof(hints)); hints.ai_socktype=SOCK_STREAM; +#if !defined(_WIN32) hints.ai_flags=AI_NUMERICSERV; +#endif OP_ASSERT(_port<=65535U); sprintf(service,"%u",_port); if(OP_LIKELY(!getaddrinfo(_host,service,&hints,&addrs)))return addrs; return NULL; } -static int op_sock_set_nonblocking(int _fd,int _nonblocking){ +static int op_sock_set_nonblocking(op_sock _fd,int _nonblocking){ +#if !defined(_WIN32) int flags; flags=fcntl(_fd,F_GETFL); if(OP_UNLIKELY(flags<0))return flags; if(_nonblocking)flags|=O_NONBLOCK; else flags&=~O_NONBLOCK; return fcntl(_fd,F_SETFL,flags); +#else + return ioctl(_fd,FIONBIO,&_nonblocking); +#endif } /*Disable/enable write coalescing if we can. We always send whole requests at once and always parse the response headers before sending another one, so normally write coalescing just causes added delay.*/ -static void op_sock_set_tcp_nodelay(int _fd,int _nodelay){ +static void op_sock_set_tcp_nodelay(op_sock _fd,int _nodelay){ # if defined(TCP_NODELAY)&&(defined(IPPROTO_TCP)||defined(SOL_TCP)) # if defined(IPPROTO_TCP) # define OP_SO_LEVEL IPPROTO_TCP @@ -615,6 +761,14 @@ static void op_sock_set_tcp_nodelay(int _fd,int _nodelay){ # endif } +#if defined(_WIN32) +static void op_init_winsock(){ + static LONG count; + static WSADATA wsadata; + if(InterlockedIncrement(&count)==1)WSAStartup(0x0202,&wsadata); +} +#endif + /*A single physical connection to an HTTP server. We may have several of these open at once.*/ struct OpusHTTPConn{ @@ -640,7 +794,7 @@ struct OpusHTTPConn{ /*The estimated throughput of this connection, in bytes/s.*/ opus_int64 read_rate; /*The socket we're reading from.*/ - int fd; + op_sock fd; /*The number of remaining requests we are allowed on this connection.*/ int nrequests_left; /*The chunk size to use for pipelining requests.*/ @@ -651,13 +805,13 @@ static void op_http_conn_init(OpusHTTPConn *_conn){ _conn->next_pos=-1; _conn->ssl_conn=NULL; _conn->next=NULL; - _conn->fd=-1; + _conn->fd=OP_INVALID_SOCKET; } static void op_http_conn_clear(OpusHTTPConn *_conn){ if(_conn->ssl_conn!=NULL)SSL_free(_conn->ssl_conn); /*SSL frees the BIO for us.*/ - if(_conn->fd>=0)close(_conn->fd); + if(_conn->fd!=OP_INVALID_SOCKET)close(_conn->fd); } /*The global stream state.*/ @@ -683,6 +837,8 @@ struct OpusHTTPStream{ struct sockaddr_in v4; struct sockaddr_in6 v6; } addr; + /*The last time we re-resolved the host.*/ + struct timeb resolve_time; /*A buffer used to build HTTP requests.*/ OpusStringBuf request; /*A buffer used to build proxy CONNECT requests.*/ @@ -694,6 +850,10 @@ struct OpusHTTPStream{ opus_int64 content_length; /*The position indicator used when no connection is active.*/ opus_int64 pos; + /*The host we actually connected to.*/ + char *connect_host; + /*The port we actually connected to.*/ + unsigned connect_port; /*The connection we're currently reading from. This can be -1 if no connection is active.*/ int cur_conni; @@ -713,7 +873,7 @@ struct OpusHTTPStream{ static void op_http_stream_init(OpusHTTPStream *_stream){ OpusHTTPConn **pnext; - int ci; + int ci; pnext=&_stream->free_head; for(ci=0;ciconns+ci); @@ -727,6 +887,7 @@ static void op_http_stream_init(OpusHTTPStream *_stream){ op_sb_init(&_stream->request); op_sb_init(&_stream->proxy_connect); op_sb_init(&_stream->response); + _stream->connect_host=NULL; _stream->seekable=0; } @@ -744,19 +905,13 @@ static void op_http_conn_close(OpusHTTPStream *_stream,OpusHTTPConn *_conn, However, we will not wait if this would block (it's not worth the savings from session resumption to do so). Clients (that's us) MAY resume a TLS session that ended with an incomplete - close, according to RFC 2818, so that's no reason to make sure the server - shut things down gracefully. - It also says "client implementations MUST treat any premature closes as - errors and the data received as potentially truncated," but libopusfile - treats errors and potentially truncated data in unseekable streams just - like a normal EOF. - We warn about this in the docs, and give some suggestions if you truly want - to avoid truncation attacks.*/ + close, according to RFC 2818, so there's no reason to make sure the server + shut things down gracefully.*/ if(_gracefully&&_conn->ssl_conn!=NULL)SSL_shutdown(_conn->ssl_conn); op_http_conn_clear(_conn); _conn->next_pos=-1; _conn->ssl_conn=NULL; - _conn->fd=-1; + _conn->fd=OP_INVALID_SOCKET; OP_ASSERT(*_pnext==_conn); *_pnext=_conn->next; _conn->next=_stream->free_head; @@ -772,6 +927,7 @@ static void op_http_stream_clear(OpusHTTPStream *_stream){ op_sb_clear(&_stream->response); op_sb_clear(&_stream->proxy_connect); op_sb_clear(&_stream->request); + if(_stream->connect_host!=_stream->url.host)_ogg_free(_stream->connect_host); op_parsed_url_clear(&_stream->url); } @@ -802,14 +958,14 @@ static int op_http_conn_write_fully(OpusHTTPConn *_conn, } else{ ssize_t ret; - errno=0; - ret=write(fd.fd,_buf,_buf_size); + op_reset_errno(); + ret=send(fd.fd,_buf,_buf_size,0); if(ret>0){ _buf+=ret; _buf_size-=ret; continue; } - err=errno; + err=op_errno(); if(err!=EAGAIN&&err!=EWOULDBLOCK)return OP_FALSE; fd.events=POLLOUT; } @@ -839,11 +995,11 @@ static int op_http_conn_estimate_available(OpusHTTPConn *_conn){ static opus_int32 op_time_diff_ms(const struct timeb *_end, const struct timeb *_start){ opus_int64 dtime; - dtime=_end->time-_start->time; + dtime=_end->time-(opus_int64)_start->time; OP_ASSERT(_end->millitm<1000); OP_ASSERT(_start->millitm<1000); - if(OP_UNLIKELY(dtime>(0x7FFFFFFF-1000)/1000))return 0x7FFFFFFF; - if(OP_UNLIKELY(dtime<(-0x7FFFFFFF+999)/1000))return -0x7FFFFFFF-1; + if(OP_UNLIKELY(dtime>(OP_INT32_MAX-1000)/1000))return OP_INT32_MAX; + if(OP_UNLIKELY(dtime<(OP_INT32_MIN+1000)/1000))return OP_INT32_MIN; return (opus_int32)dtime*1000+_end->millitm-_start->millitm; } @@ -855,7 +1011,7 @@ static void op_http_conn_read_rate_update(OpusHTTPConn *_conn){ opus_int64 read_rate; read_delta_bytes=_conn->read_bytes; if(read_delta_bytes<=0)return; - OP_ALWAYS_TRUE(!ftime(&read_time)); + ftime(&read_time); read_delta_ms=op_time_diff_ms(&read_time,&_conn->read_time); read_rate=_conn->read_rate; read_delta_ms=OP_MAX(read_delta_ms,1); @@ -881,6 +1037,9 @@ static int op_http_conn_read(OpusHTTPConn *_conn, fd.fd=_conn->fd; ssl_conn=_conn->ssl_conn; nread=nread_unblocked=0; + /*RFC 2818 says "client implementations MUST treat any premature closes as + errors and the data received as potentially truncated," so we make very + sure to report read errors upwards.*/ do{ int err; if(ssl_conn!=NULL){ @@ -912,8 +1071,8 @@ static int op_http_conn_read(OpusHTTPConn *_conn, } else{ ssize_t ret; - errno=0; - ret=read(fd.fd,_buf+nread,_buf_size-nread); + op_reset_errno(); + ret=recv(fd.fd,_buf+nread,_buf_size-nread,0); OP_ASSERT(ret<=_buf_size-nread); if(ret>0){ /*Read some data. @@ -925,7 +1084,7 @@ static int op_http_conn_read(OpusHTTPConn *_conn, /*If we already read some data or the connection was closed, return right now.*/ if(ret==0||nread>0)break; - err=errno; + err=op_errno(); if(err!=EAGAIN&&err!=EWOULDBLOCK)return OP_EREAD; fd.events=POLLIN; } @@ -944,8 +1103,7 @@ static int op_http_conn_read(OpusHTTPConn *_conn, /*Tries to look at the pending data for a connection without consuming it. [out] _buf: Returns the data at which we're peeking. _buf_size: The size of the buffer.*/ -static int op_http_conn_peek(OpusHTTPConn *_conn, - char *_buf,int _buf_size){ +static int op_http_conn_peek(OpusHTTPConn *_conn,char *_buf,int _buf_size){ struct pollfd fd; SSL *ssl_conn; int ret; @@ -964,11 +1122,11 @@ static int op_http_conn_peek(OpusHTTPConn *_conn, else return 0; } else{ - errno=0; + op_reset_errno(); ret=(int)recv(fd.fd,_buf,_buf_size,MSG_PEEK); /*Either saw some data or the connection was closed.*/ if(ret>=0)return ret; - err=errno; + err=op_errno(); if(err!=EAGAIN&&err!=EWOULDBLOCK)return 0; fd.events=POLLIN; } @@ -1192,24 +1350,22 @@ static int op_http_get_next_header(char **_header,char **_cdr,char **_s){ static opus_int64 op_http_parse_nonnegative_int64(const char **_next, const char *_cdr){ const char *next; - opus_int64 content_length; + opus_int64 ret; int i; next=_cdr+strspn(_cdr,OP_HTTP_DIGIT); *_next=next; if(OP_UNLIKELY(next<=_cdr))return OP_FALSE; while(*_cdr=='0')_cdr++; if(OP_UNLIKELY(next-_cdr>19))return OP_EIMPL; - content_length=0; + ret=0; for(i=0;i(OP_INT64_MAX-9)/10+(digit<=7))){ - return OP_EIMPL; - } - content_length=content_length*10+digit; + if(OP_UNLIKELY(ret>(OP_INT64_MAX-9)/10+(digit<=7)))return OP_EIMPL; + ret=ret*10+digit; } - return content_length; + return ret; } static opus_int64 op_http_parse_content_length(const char *_cdr){ @@ -1295,7 +1451,7 @@ static int op_http_parse_connection(char *_cdr){ typedef int (*op_ssl_step_func)(SSL *_ssl_conn); /*Try to run an SSL function to completion (blocking if necessary).*/ -static int op_do_ssl_step(SSL *_ssl_conn,int _fd,op_ssl_step_func _step){ +static int op_do_ssl_step(SSL *_ssl_conn,op_sock _fd,op_ssl_step_func _step){ struct pollfd fd; fd.fd=_fd; for(;;){ @@ -1389,8 +1545,8 @@ static BIO_METHOD op_bio_retry_method={ /*Establish a CONNECT tunnel and pipeline the start of the TLS handshake for proxying https URL requests.*/ -int op_http_conn_establish_tunnel(OpusHTTPStream *_stream, - OpusHTTPConn *_conn,int _fd,SSL *_ssl_conn,BIO *_ssl_bio){ +static int op_http_conn_establish_tunnel(OpusHTTPStream *_stream, + OpusHTTPConn *_conn,op_sock _fd,SSL *_ssl_conn,BIO *_ssl_bio){ BIO *retry_bio; char *status_code; char *next; @@ -1507,8 +1663,7 @@ static struct addrinfo *op_inet_pton(const char *_host){ /*Verify the server's hostname matches the certificate they presented using the procedure from Section 6 of RFC 6125. Return: 0 if the certificate doesn't match, and a non-zero value if it does.*/ -static int op_http_verify_hostname(OpusHTTPStream *_stream, - SSL *_ssl_conn){ +static int op_http_verify_hostname(OpusHTTPStream *_stream,SSL *_ssl_conn){ X509 *peer_cert; STACK_OF(GENERAL_NAME) *san_names; char *host; @@ -1576,7 +1731,7 @@ static int op_http_verify_hostname(OpusHTTPStream *_stream, equivalent) of a URI or deriving the application service type from the scheme of a URI) ..." We don't have a way to check (without relying on DNS records, which might - be subverted), if this address is fully-qualified. + be subverted) if this address is fully-qualified. This is particularly problematic when using a CONNECT tunnel, as it is the server that does DNS lookup, not us. However, we are certain that if the hostname has no '.', it is definitely @@ -1659,8 +1814,8 @@ static int op_http_verify_hostname(OpusHTTPStream *_stream, } /*Perform the TLS handshake on a new connection.*/ -int op_http_conn_start_tls(OpusHTTPStream *_stream,OpusHTTPConn *_conn, - int _fd,SSL *_ssl_conn){ +static int op_http_conn_start_tls(OpusHTTPStream *_stream,OpusHTTPConn *_conn, + op_sock _fd,SSL *_ssl_conn){ SSL_SESSION *ssl_session; BIO *ssl_bio; int skip_certificate_check; @@ -1724,9 +1879,10 @@ int op_http_conn_start_tls(OpusHTTPStream *_stream,OpusHTTPConn *_conn, OP_FALSE If the connection failed and there were no more addresses left to try. *_addr will be set to NULL in this case.*/ -static int op_sock_connect_next(int _fd, - struct addrinfo **_addr,int _ai_family){ - struct addrinfo *addr; +static int op_sock_connect_next(op_sock _fd, + const struct addrinfo **_addr,int _ai_family){ + const struct addrinfo *addr; + int err; addr=*_addr; for(;;){ /*Move to the next address of the requested type.*/ @@ -1735,7 +1891,9 @@ static int op_sock_connect_next(int _fd, /*No more: failure.*/ if(addr==NULL)return OP_FALSE; if(connect(_fd,addr->ai_addr,addr->ai_addrlen)>=0)return 1; - if(OP_LIKELY(errno==EINPROGRESS))return 0; + err=op_errno(); + /*Winsock will set WSAEWOULDBLOCK.*/ + if(OP_LIKELY(err==EINPROGRESS||err==EWOULDBLOCK))return 0; addr=addr->ai_next; } } @@ -1743,36 +1901,30 @@ static int op_sock_connect_next(int _fd, /*The number of address families to try connecting to simultaneously.*/ # define OP_NPROTOS (2) -static int op_http_connect(OpusHTTPStream *_stream,OpusHTTPConn *_conn, - struct addrinfo *_addrs,struct timeb *_start_time){ - struct addrinfo *addr; - struct addrinfo *addrs[OP_NPROTOS]; - struct pollfd fds[OP_NPROTOS]; - int ai_family; - int nprotos; - int ret; - int pi; - int pj; +static int op_http_connect_impl(OpusHTTPStream *_stream,OpusHTTPConn *_conn, + const struct addrinfo *_addrs,struct timeb *_start_time){ + const struct addrinfo *addr; + const struct addrinfo *addrs[OP_NPROTOS]; + struct pollfd fds[OP_NPROTOS]; + int ai_family; + int nprotos; + int ret; + int pi; + int pj; for(pi=0;piai_next){ - /*Give IPv6 a slight edge by putting it first in the list.*/ - if(addr->ai_family==AF_INET6){ + one that succeeds. + Start by finding the first address from each family. + We order the first connection attempts in the same order the address + families were returned in the DNS records in accordance with RFC 6555.*/ + for(addr=_addrs,nprotos=0;addr!=NULL&&nprotosai_next){ + if(addr->ai_family==AF_INET6||addr->ai_family==AF_INET){ OP_ASSERT(addr->ai_addrlen<=sizeof(struct sockaddr_in6)); - if(addrs[0]==NULL)addrs[0]=addr; - } - else if(addr->ai_family==AF_INET){ OP_ASSERT(addr->ai_addrlen<=sizeof(struct sockaddr_in)); - if(addrs[1]==NULL)addrs[1]=addr; - } - } - /*Consolidate the list of addresses.*/ - for(pi=nprotos=0;piai_family==addr->ai_family)break; + if(pifree_head=_conn->next; _conn->next=_stream->lru_head; _stream->lru_head=_conn; - OP_ALWAYS_TRUE(!ftime(_start_time)); + ftime(_start_time); *&_conn->read_time=*_start_time; _conn->read_bytes=0; _conn->read_rate=0; - /*Try to start a connection to each protocol.*/ + /*Try to start a connection to each protocol. + RFC 6555 says it is RECOMMENDED that connection attempts be paced + 150...250 ms apart "to balance human factors against network load", but + that "stateful algorithms" (that's us) "are expected to be more + aggressive". + We are definitely more aggressive: we don't pace at all.*/ for(pi=0;piai_family; fds[pi].fd=socket(ai_family,SOCK_STREAM,addrs[pi]->ai_protocol); fds[pi].events=POLLOUT; - if(OP_LIKELY(fds[pi].fd>=0)){ + if(OP_LIKELY(fds[pi].fd!=OP_INVALID_SOCKET)){ if(OP_LIKELY(op_sock_set_nonblocking(fds[pi].fd,1)>=0)){ ret=op_sock_connect_next(fds[pi].fd,addrs+pi,ai_family); if(OP_UNLIKELY(ret>0)){ @@ -1820,7 +1977,7 @@ static int op_http_connect(OpusHTTPStream *_stream,OpusHTTPConn *_conn, /*Some platforms will return the pending error in &err and return 0. Others will put it in errno and return -1.*/ ret=getsockopt(fds[pi].fd,SOL_SOCKET,SO_ERROR,&err,&errlen); - if(ret<0)err=errno; + if(ret<0)err=op_errno(); /*Success!*/ if(err==0||err==EISCONN)break; /*Move on to the next address for this protocol.*/ @@ -1862,7 +2019,7 @@ static int op_http_connect(OpusHTTPStream *_stream,OpusHTTPConn *_conn, SSL_free(ssl_conn); } close(fds[pi].fd); - _conn->fd=-1; + _conn->fd=OP_INVALID_SOCKET; return OP_FALSE; } /*Just a normal non-SSL connection.*/ @@ -1876,6 +2033,29 @@ static int op_http_connect(OpusHTTPStream *_stream,OpusHTTPConn *_conn, return 0; } +static int op_http_connect(OpusHTTPStream *_stream,OpusHTTPConn *_conn, + const struct addrinfo *_addrs,struct timeb *_start_time){ + struct timeb resolve_time; + struct addrinfo *new_addrs; + int ret; + /*Re-resolve the host if we need to (RFC 6555 says we MUST do so + occasionally).*/ + new_addrs=NULL; + ftime(&resolve_time); + if(_addrs!=&_stream->addr_info||op_time_diff_ms(&resolve_time, + &_stream->resolve_time)>=OP_RESOLVE_CACHE_TIMEOUT_MS){ + new_addrs=op_resolve(_stream->connect_host,_stream->connect_port); + if(OP_LIKELY(new_addrs!=NULL)){ + _addrs=new_addrs; + *&_stream->resolve_time=*&resolve_time; + } + else if(OP_LIKELY(_addrs==NULL))return OP_FALSE; + } + ret=op_http_connect_impl(_stream,_conn,_addrs,_start_time); + if(new_addrs!=NULL)freeaddrinfo(new_addrs); + return ret; +} + # define OP_BASE64_LENGTH(_len) (((_len)+2)/3*4) static const char BASE64_TABLE[64]={ @@ -1897,15 +2077,15 @@ static char *op_base64_encode(char *_dst,const char *_src,int _len){ s1=_src[3*i+1]; s2=_src[3*i+2]; _dst[4*i+0]=BASE64_TABLE[s0>>2]; - _dst[4*i+1]=BASE64_TABLE[s0&3<<4|s1>>4]; - _dst[4*i+2]=BASE64_TABLE[s1&15<<2|s2>>6]; + _dst[4*i+1]=BASE64_TABLE[(s0&3)<<4|s1>>4]; + _dst[4*i+2]=BASE64_TABLE[(s1&15)<<2|s2>>6]; _dst[4*i+3]=BASE64_TABLE[s2&63]; } _len-=3*i; if(_len==1){ s0=_src[3*i+0]; _dst[4*i+0]=BASE64_TABLE[s0>>2]; - _dst[4*i+1]=BASE64_TABLE[s0&3<<4]; + _dst[4*i+1]=BASE64_TABLE[(s0&3)<<4]; _dst[4*i+2]='='; _dst[4*i+3]='='; i++; @@ -1914,8 +2094,8 @@ static char *op_base64_encode(char *_dst,const char *_src,int _len){ s0=_src[3*i+0]; s1=_src[3*i+1]; _dst[4*i+0]=BASE64_TABLE[s0>>2]; - _dst[4*i+1]=BASE64_TABLE[s0&3<<4|s1>>4]; - _dst[4*i+2]=BASE64_TABLE[s1&15<<2]; + _dst[4*i+1]=BASE64_TABLE[(s0&3)<<4|s1>>4]; + _dst[4*i+2]=BASE64_TABLE[(s1&15)<<2]; _dst[4*i+3]='='; i++; } @@ -1990,50 +2170,33 @@ static int op_http_allow_pipelining(const char *_server){ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, int _skip_certificate_check,const char *_proxy_host,unsigned _proxy_port, - const char *_proxy_user,const char *_proxy_pass){ + const char *_proxy_user,const char *_proxy_pass,OpusServerInfo *_info){ struct addrinfo *addrs; - const char *last_host; - unsigned last_port; int nredirs; int ret; - if(_proxy_host!=NULL&&OP_UNLIKELY(_proxy_port>65535U))return OP_EINVAL; - last_host=NULL; - /*We shouldn't have to initialize last_port, but gcc is too dumb to figure - out that last_host!=NULL implies we've already taken one trip through the - loop.*/ - last_port=0; +#if defined(_WIN32) + op_init_winsock(); +#endif ret=op_parse_url(&_stream->url,_url); if(OP_UNLIKELY(ret<0))return ret; + if(_proxy_host!=NULL){ + if(OP_UNLIKELY(_proxy_port>65535U))return OP_EINVAL; + _stream->connect_host=op_string_dup(_proxy_host); + _stream->connect_port=_proxy_port; + } + else{ + _stream->connect_host=_stream->url.host; + _stream->connect_port=_stream->url.port; + } + addrs=NULL; for(nredirs=0;nredirsurl.host; - port=_stream->url.port; - } - else{ - host=_proxy_host; - port=_proxy_port; - } - /*If connecting to the same place as last time, don't re-resolve it.*/ - addrs=NULL; - if(last_host!=NULL){ - if(strcmp(last_host,host)==0&&last_port==port)addrs=&_stream->addr_info; - else if(_stream->ssl_session!=NULL){ - /*Forget any cached SSL session from the last host.*/ - SSL_SESSION_free(_stream->ssl_session); - _stream->ssl_session=NULL; - } - if(last_host!=_proxy_host)_ogg_free((void *)last_host); - } - last_host=host; - last_port=port; + OpusParsedURL next_url; + struct timeb start_time; + struct timeb end_time; + char *next; + char *status_code; + int minor_version_pos; + int v1_1_compat; /*Initialize the SSL library if necessary.*/ if(OP_URL_IS_SSL(&_stream->url)&&_stream->ssl_ctx==NULL){ SSL_CTX *ssl_ctx; @@ -2069,6 +2232,7 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, if(_proxy_host!=NULL){ /*We need to establish a CONNECT tunnel to handle https proxying. Build the request we'll send to do so.*/ + _stream->proxy_connect.nbuf=0; ret=op_sb_append(&_stream->proxy_connect,"CONNECT ",8); ret|=op_sb_append_string(&_stream->proxy_connect,_stream->url.host); ret|=op_sb_append_port(&_stream->proxy_connect,_stream->url.port); @@ -2096,12 +2260,7 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, } } /*Actually make the connection.*/ - if(addrs!=&_stream->addr_info){ - addrs=op_resolve(host,port); - if(OP_UNLIKELY(addrs==NULL))return OP_FALSE; - } ret=op_http_connect(_stream,_stream->conns+0,addrs,&start_time); - if(addrs!=&_stream->addr_info)freeaddrinfo(addrs); if(OP_UNLIKELY(ret<0))return ret; /*Build the request to send.*/ _stream->request.nbuf=0; @@ -2162,20 +2321,22 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, if(OP_UNLIKELY(ret<0))return ret; ret=op_http_conn_read_response(_stream->conns+0,&_stream->response); if(OP_UNLIKELY(ret<0))return ret; - OP_ALWAYS_TRUE(!ftime(&end_time)); + ftime(&end_time); next=op_http_parse_status_line(&v1_1_compat,&status_code, _stream->response.buf); if(OP_UNLIKELY(next==NULL))return OP_FALSE; if(status_code[0]=='2'){ opus_int64 content_length; opus_int64 range_length; - int pipeline; + int pipeline_supported; + int pipeline_disabled; /*We only understand 20x codes.*/ if(status_code[1]!='0')return OP_FALSE; content_length=-1; range_length=-1; - /*Pipelining is disabled by default.*/ - pipeline=0; + /*Pipelining must be explicitly enabled.*/ + pipeline_supported=0; + pipeline_disabled=0; for(;;){ char *header; char *cdr; @@ -2200,7 +2361,7 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, ret=op_http_parse_content_range(&range_first,&range_last, &range_length,cdr); if(OP_UNLIKELY(ret<0))return ret; - /*"A response with satus code 206 (Partial Content) MUST NOTE + /*"A response with satus code 206 (Partial Content) MUST NOT include a Content-Range field with a byte-range-resp-spec of '*'."*/ if(status_code[2]=='6' @@ -2233,9 +2394,9 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, error out and reconnect, adding lots of latency.*/ ret=op_http_parse_connection(cdr); if(OP_UNLIKELY(ret<0))return ret; - pipeline-=ret; + pipeline_disabled|=ret; } - else if(strcmp(header,"server")){ + else if(strcmp(header,"server")==0){ /*If we got a Server response header, and it wasn't from a known-bad server, enable pipelining, as long as it's at least HTTP/1.1. According to RFC 2145, the server is supposed to respond with the @@ -2243,7 +2404,51 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, suspected that we incorrectly implement the HTTP specification. So it should send back at least HTTP/1.1, despite our HTTP/1.0 request.*/ - pipeline+=v1_1_compat&&op_http_allow_pipelining(cdr); + pipeline_supported=v1_1_compat; + if(v1_1_compat)pipeline_disabled|=!op_http_allow_pipelining(cdr); + if(_info!=NULL&&_info->server==NULL)_info->server=op_string_dup(cdr); + } + /*Collect station information headers if the caller requested it. + If there's more than one copy of a header, the first one wins.*/ + else if(_info!=NULL){ + if(strcmp(header,"content-type")==0){ + if(_info->content_type==NULL){ + _info->content_type=op_string_dup(cdr); + } + } + else if(header[0]=='i'&&header[1]=='c' + &&(header[2]=='e'||header[2]=='y')&&header[3]=='-'){ + if(strcmp(header+4,"name")==0){ + if(_info->name==NULL)_info->name=op_string_dup(cdr); + } + else if(strcmp(header+4,"description")==0){ + if(_info->description==NULL)_info->description=op_string_dup(cdr); + } + else if(strcmp(header+4,"genre")==0){ + if(_info->genre==NULL)_info->genre=op_string_dup(cdr); + } + else if(strcmp(header+4,"url")==0){ + if(_info->url==NULL)_info->url=op_string_dup(cdr); + } + else if(strcmp(header,"icy-br")==0 + ||strcmp(header,"ice-bitrate")==0){ + if(_info->bitrate_kbps<0){ + opus_int64 bitrate_kbps; + /*Just re-using this function to parse a random unsigned + integer field.*/ + bitrate_kbps=op_http_parse_content_length(cdr); + if(bitrate_kbps>=0&&bitrate_kbps<=OP_INT32_MAX){ + _info->bitrate_kbps=(opus_int32)bitrate_kbps; + } + } + } + else if(strcmp(header,"icy-pub")==0 + ||strcmp(header,"ice-public")==0){ + if(_info->is_public<0&&(cdr[0]=='0'||cdr[0]=='1')&&cdr[1]=='\0'){ + _info->is_public=cdr[0]-'0'; + } + } + } } } switch(status_code[2]){ @@ -2278,15 +2483,16 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, default:return OP_FALSE; } _stream->content_length=content_length; - _stream->pipeline=pipeline>0; + _stream->pipeline=pipeline_supported&&!pipeline_disabled; /*Pipelining requires HTTP/1.1 persistent connections.*/ - if(pipeline)_stream->request.buf[minor_version_pos]='1'; + if(_stream->pipeline)_stream->request.buf[minor_version_pos]='1'; _stream->conns[0].pos=0; _stream->conns[0].end_pos=_stream->seekable?content_length:-1; _stream->conns[0].chunk_size=-1; _stream->cur_conni=0; _stream->connect_rate=op_time_diff_ms(&end_time,&start_time); _stream->connect_rate=OP_MAX(_stream->connect_rate,1); + if(_info!=NULL)_info->is_ssl=OP_URL_IS_SSL(&_stream->url); /*The URL has been successfully opened.*/ return 0; } @@ -2306,7 +2512,9 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, /*302 Found*/ case '2': /*307 Temporary Redirect*/ - case '7':break; + case '7': + /*308 Permanent Redirect (defined by draft-reschke-http-status-308-07).*/ + case '8':break; /*305 Use Proxy: "The Location field gives the URI of the proxy." TODO: This shouldn't actually be that hard to do.*/ case '5':return OP_EIMPL; @@ -2314,7 +2522,7 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, originally requested resource." 304 Not Modified: "The 304 response MUST NOT contain a message-body." 306 (Unused) - 308...309 are not yet defined, so we don't know how to handle them.*/ + 309 is not yet defined, so we don't know how to handle it.*/ default:return OP_FALSE; } _url=NULL; @@ -2327,15 +2535,33 @@ static int op_http_stream_open(OpusHTTPStream *_stream,const char *_url, if(strcmp(header,"location")==0&&OP_LIKELY(_url==NULL))_url=cdr; } if(OP_UNLIKELY(_url==NULL))return OP_FALSE; - /*Don't free last_host if it came from the last URL.*/ - if(last_host!=_proxy_host)_stream->url.host=NULL; - op_parsed_url_clear(&_stream->url); - ret=op_parse_url(&_stream->url,_url); - if(OP_UNLIKELY(ret<0)){ - if(ret==OP_EINVAL)ret=OP_FALSE; - if(last_host!=_proxy_host)_ogg_free((void *)last_host); - return ret; + ret=op_parse_url(&next_url,_url); + if(OP_UNLIKELY(ret<0))return ret; + if(_proxy_host==NULL||_stream->ssl_session!=NULL){ + if(strcmp(_stream->url.host,next_url.host)==0 + &&_stream->url.port==next_url.port){ + /*Try to skip re-resolve when connecting to the same host.*/ + addrs=&_stream->addr_info; + } + else{ + if(_stream->ssl_session!=NULL){ + /*Forget any cached SSL session from the last host.*/ + SSL_SESSION_free(_stream->ssl_session); + _stream->ssl_session=NULL; + } + } } + if(_proxy_host==NULL){ + OP_ASSERT(_stream->connect_host==_stream->url.host); + _stream->connect_host=next_url.host; + _stream->connect_port=next_url.port; + } + /*Always try to skip re-resolve for proxy connections.*/ + else addrs=&_stream->addr_info; + op_parsed_url_clear(&_stream->url); + *&_stream->url=*&next_url; + /*TODO: On servers/proxies that support pipelining, we might be able to + re-use this connection.*/ op_http_conn_close(_stream,_stream->conns+0,&_stream->lru_head,1); } /*Redirection limit reached.*/ @@ -2500,7 +2726,7 @@ static int op_http_conn_handle_response(OpusHTTPStream *_stream, } /*Open a new connection that will start reading at byte offset _pos. - _pos: The byte offset to start readiny from. + _pos: The byte offset to start reading from. _chunk_size: The number of bytes to ask for in the initial request, or -1 to request the rest of the resource. This may be more bytes than remain, in which case it will be @@ -2518,7 +2744,7 @@ static int op_http_conn_open_pos(OpusHTTPStream *_stream, if(OP_UNLIKELY(ret<0))return ret; ret=op_http_conn_handle_response(_stream,_conn); if(OP_UNLIKELY(ret!=0))return OP_FALSE; - OP_ALWAYS_TRUE(!ftime(&end_time)); + ftime(&end_time); _stream->cur_conni=_conn-_stream->conns; OP_ASSERT(_stream->cur_conni>=0&&_stream->cur_conni=0&&next_end-next_pos<=0x7FFFFFFF); + ||next_end-next_pos>=0&&next_end-next_pos<=OP_INT32_MAX); ret=op_http_conn_open_pos(_stream,_conn,next_pos, next_end<0?-1:(opus_int32)(next_end-next_pos)); if(OP_UNLIKELY(ret<0))return OP_EREAD; @@ -2811,7 +3037,7 @@ static int op_http_stream_seek(void *_stream,opus_int64 _offset,int _whence){ op_http_conn_read_rate_update(stream->conns+ci); *&seek_time=*&stream->conns[ci].read_time; } - else OP_ALWAYS_TRUE(!ftime(&seek_time)); + else ftime(&seek_time); /*If we seeked past the end of the stream, just disable the active connection.*/ if(pos>=content_length){ @@ -2979,12 +3205,33 @@ static const OpusFileCallbacks OP_HTTP_CALLBACKS={ }; #endif +void opus_server_info_init(OpusServerInfo *_info){ + _info->name=NULL; + _info->description=NULL; + _info->genre=NULL; + _info->url=NULL; + _info->server=NULL; + _info->content_type=NULL; + _info->bitrate_kbps=-1; + _info->is_public=-1; + _info->is_ssl=0; +} + +void opus_server_info_clear(OpusServerInfo *_info){ + _ogg_free(_info->content_type); + _ogg_free(_info->server); + _ogg_free(_info->url); + _ogg_free(_info->genre); + _ogg_free(_info->description); + _ogg_free(_info->name); +} + /*The actual URL stream creation function. This one isn't extensible like the application-level interface, but because it isn't public, we're free to change it in the future.*/ static void *op_url_stream_create_impl(OpusFileCallbacks *_cb,const char *_url, int _skip_certificate_check,const char *_proxy_host,unsigned _proxy_port, - const char *_proxy_user,const char *_proxy_pass){ + const char *_proxy_user,const char *_proxy_pass,OpusServerInfo *_info){ const char *path; /*Check to see if this is a valid file: URL.*/ path=op_parse_file_url(_url); @@ -3006,7 +3253,7 @@ static void *op_url_stream_create_impl(OpusFileCallbacks *_cb,const char *_url, if(OP_UNLIKELY(stream==NULL))return NULL; op_http_stream_init(stream); ret=op_http_stream_open(stream,_url,_skip_certificate_check, - _proxy_host,_proxy_port,_proxy_user,_proxy_pass); + _proxy_host,_proxy_port,_proxy_user,_proxy_pass,_info); if(OP_UNLIKELY(ret<0)){ op_http_stream_clear(stream); _ogg_free(stream); @@ -3021,22 +3268,25 @@ static void *op_url_stream_create_impl(OpusFileCallbacks *_cb,const char *_url, (void)_proxy_port; (void)_proxy_user; (void)_proxy_pass; + (void)_info; return NULL; #endif } void *op_url_stream_vcreate(OpusFileCallbacks *_cb, const char *_url,va_list _ap){ - int skip_certificate_check; - const char *proxy_host; - opus_int32 proxy_port; - const char *proxy_user; - const char *proxy_pass; + int skip_certificate_check; + const char *proxy_host; + opus_int32 proxy_port; + const char *proxy_user; + const char *proxy_pass; + OpusServerInfo *pinfo; skip_certificate_check=0; proxy_host=NULL; proxy_port=8080; proxy_user=NULL; proxy_pass=NULL; + pinfo=NULL; for(;;){ ptrdiff_t request; request=va_arg(_ap,char *)-(char *)NULL; @@ -3059,17 +3309,83 @@ void *op_url_stream_vcreate(OpusFileCallbacks *_cb, case OP_HTTP_PROXY_PASS_REQUEST:{ proxy_pass=va_arg(_ap,const char *); }break; + case OP_GET_SERVER_INFO_REQUEST:{ + pinfo=va_arg(_ap,OpusServerInfo *); + }break; /*Some unknown option.*/ default:return NULL; } } + /*If the caller has requested server information, proxy it to a local copy to + simplify error handling.*/ + if(pinfo!=NULL){ + OpusServerInfo info; + void *ret; + opus_server_info_init(&info); + ret=op_url_stream_create_impl(_cb,_url,skip_certificate_check, + proxy_host,proxy_port,proxy_user,proxy_pass,&info); + if(ret!=NULL)*pinfo=*&info; + else opus_server_info_clear(&info); + return ret; + } return op_url_stream_create_impl(_cb,_url,skip_certificate_check, - proxy_host,proxy_port,proxy_user,proxy_pass); + proxy_host,proxy_port,proxy_user,proxy_pass,NULL); } void *op_url_stream_create(OpusFileCallbacks *_cb, const char *_url,...){ - va_list ap; + va_list ap; + void *ret; va_start(ap,_url); - return op_url_stream_vcreate(_cb,_url,ap); + ret=op_url_stream_vcreate(_cb,_url,ap); + va_end(ap); + return ret; +} + +/*Convenience routines to open/test URLs in a single step.*/ + +OggOpusFile *op_vopen_url(const char *_url,int *_error,va_list _ap){ + OpusFileCallbacks cb; + OggOpusFile *of; + void *source; + source=op_url_stream_vcreate(&cb,_url,_ap); + if(OP_UNLIKELY(source==NULL)){ + if(_error!=NULL)*_error=OP_EFAULT; + return NULL; + } + of=op_open_callbacks(source,&cb,NULL,0,_error); + if(OP_UNLIKELY(of==NULL))(*cb.close)(source); + return of; +} + +OggOpusFile *op_open_url(const char *_url,int *_error,...){ + OggOpusFile *ret; + va_list ap; + va_start(ap,_error); + ret=op_vopen_url(_url,_error,ap); + va_end(ap); + return ret; +} + +OggOpusFile *op_vtest_url(const char *_url,int *_error,va_list _ap){ + OpusFileCallbacks cb; + OggOpusFile *of; + void *source; + source=op_url_stream_vcreate(&cb,_url,_ap); + if(OP_UNLIKELY(source==NULL)){ + if(_error!=NULL)*_error=OP_EFAULT; + return NULL; + } + of=op_test_callbacks(source,&cb,NULL,0,_error); + if(OP_UNLIKELY(of==NULL))(*cb.close)(source); + return of; +} + +OggOpusFile *op_test_url(const char *_url,int *_error,...){ + OggOpusFile *ret; + va_list ap; + va_start(ap,_error); + ret=op_vtest_url(_url,_error,ap); + va_end(ap); + return ret; } diff --git a/code/opusfile-0.5/src/info.c b/code/opusfile-0.5/src/info.c new file mode 100644 index 00000000..6cf98516 --- /dev/null +++ b/code/opusfile-0.5/src/info.c @@ -0,0 +1,683 @@ +/******************************************************************** + * * + * THIS FILE IS PART OF THE libopusfile SOFTWARE CODEC SOURCE CODE. * + * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS * + * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE * + * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. * + * * + * THE libopusfile SOURCE CODE IS (C) COPYRIGHT 2012 * + * by the Xiph.Org Foundation and contributors http://www.xiph.org/ * + * * + ********************************************************************/ +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "internal.h" +#include +#include + +static unsigned op_parse_uint16le(const unsigned char *_data){ + return _data[0]|_data[1]<<8; +} + +static int op_parse_int16le(const unsigned char *_data){ + int ret; + ret=_data[0]|_data[1]<<8; + return (ret^0x8000)-0x8000; +} + +static opus_uint32 op_parse_uint32le(const unsigned char *_data){ + return _data[0]|_data[1]<<8|_data[2]<<16|_data[3]<<24; +} + +static opus_uint32 op_parse_uint32be(const unsigned char *_data){ + return _data[3]|_data[2]<<8|_data[1]<<16|_data[0]<<24; +} + +int opus_head_parse(OpusHead *_head,const unsigned char *_data,size_t _len){ + OpusHead head; + if(_len<8)return OP_ENOTFORMAT; + if(memcmp(_data,"OpusHead",8)!=0)return OP_ENOTFORMAT; + if(_len<9)return OP_EBADHEADER; + head.version=_data[8]; + if(head.version>15)return OP_EVERSION; + if(_len<19)return OP_EBADHEADER; + head.channel_count=_data[9]; + head.pre_skip=op_parse_uint16le(_data+10); + head.input_sample_rate=op_parse_uint32le(_data+12); + head.output_gain=op_parse_int16le(_data+16); + head.mapping_family=_data[18]; + if(head.mapping_family==0){ + if(head.channel_count<1||head.channel_count>2)return OP_EBADHEADER; + if(head.version<=1&&_len>19)return OP_EBADHEADER; + head.stream_count=1; + head.coupled_count=head.channel_count-1; + if(_head!=NULL){ + _head->mapping[0]=0; + _head->mapping[1]=1; + } + } + else if(head.mapping_family==1){ + size_t size; + int ci; + if(head.channel_count<1||head.channel_count>8)return OP_EBADHEADER; + size=21+head.channel_count; + if(_lensize)return OP_EBADHEADER; + head.stream_count=_data[19]; + if(head.stream_count<1)return OP_EBADHEADER; + head.coupled_count=_data[20]; + if(head.coupled_count>head.stream_count)return OP_EBADHEADER; + for(ci=0;ci=head.stream_count+head.coupled_count + &&_data[21+ci]!=255){ + return OP_EBADHEADER; + } + } + if(_head!=NULL)memcpy(_head->mapping,_data+21,head.channel_count); + } + /*General purpose players should not attempt to play back content with + channel mapping family 255.*/ + else if(head.mapping_family==255)return OP_EIMPL; + /*No other channel mapping families are currently defined.*/ + else return OP_EBADHEADER; + if(_head!=NULL)memcpy(_head,&head,head.mapping-(unsigned char *)&head); + return 0; +} + +void opus_tags_init(OpusTags *_tags){ + memset(_tags,0,sizeof(*_tags)); +} + +void opus_tags_clear(OpusTags *_tags){ + int ci; + for(ci=_tags->comments;ci-->0;)_ogg_free(_tags->user_comments[ci]); + _ogg_free(_tags->user_comments); + _ogg_free(_tags->comment_lengths); + _ogg_free(_tags->vendor); +} + +/*Ensure there's room for up to _ncomments comments.*/ +static int op_tags_ensure_capacity(OpusTags *_tags,size_t _ncomments){ + char **user_comments; + int *comment_lengths; + size_t size; + if(OP_UNLIKELY(_ncomments>=(size_t)INT_MAX))return OP_EFAULT; + size=sizeof(*_tags->comment_lengths)*(_ncomments+1); + if(size/sizeof(*_tags->comment_lengths)!=_ncomments+1)return OP_EFAULT; + comment_lengths=(int *)_ogg_realloc(_tags->comment_lengths,size); + if(OP_UNLIKELY(comment_lengths==NULL))return OP_EFAULT; + comment_lengths[_ncomments]=0; + _tags->comment_lengths=comment_lengths; + size=sizeof(*_tags->user_comments)*(_ncomments+1); + if(size/sizeof(*_tags->user_comments)!=_ncomments+1)return OP_EFAULT; + user_comments=(char **)_ogg_realloc(_tags->user_comments,size); + if(OP_UNLIKELY(user_comments==NULL))return OP_EFAULT; + user_comments[_ncomments]=NULL; + _tags->user_comments=user_comments; + return 0; +} + +/*Duplicate a (possibly non-NUL terminated) string with a known length.*/ +static char *op_strdup_with_len(const char *_s,size_t _len){ + size_t size; + char *ret; + size=sizeof(*ret)*(_len+1); + if(OP_UNLIKELY(size<_len))return NULL; + ret=(char *)_ogg_malloc(size); + if(OP_LIKELY(ret!=NULL)){ + ret=(char *)memcpy(ret,_s,sizeof(*ret)*_len); + ret[_len]='\0'; + } + return ret; +} + +/*The actual implementation of opus_tags_parse(). + Unlike the public API, this function requires _tags to already be + initialized, modifies its contents before success is guaranteed, and assumes + the caller will clear it on error.*/ +static int opus_tags_parse_impl(OpusTags *_tags, + const unsigned char *_data,size_t _len){ + opus_uint32 count; + size_t len; + int ncomments; + int ci; + len=_len; + if(len<8)return OP_ENOTFORMAT; + if(memcmp(_data,"OpusTags",8)!=0)return OP_ENOTFORMAT; + if(len<16)return OP_EBADHEADER; + _data+=8; + len-=8; + count=op_parse_uint32le(_data); + _data+=4; + len-=4; + if(count>len)return OP_EBADHEADER; + if(_tags!=NULL){ + _tags->vendor=op_strdup_with_len((char *)_data,count); + if(_tags->vendor==NULL)return OP_EFAULT; + } + _data+=count; + len-=count; + if(len<4)return OP_EBADHEADER; + count=op_parse_uint32le(_data); + _data+=4; + len-=4; + /*Check to make sure there's minimally sufficient data left in the packet.*/ + if(count>len>>2)return OP_EBADHEADER; + /*Check for overflow (the API limits this to an int).*/ + if(count>(opus_uint32)INT_MAX-1)return OP_EFAULT; + if(_tags!=NULL){ + int ret; + ret=op_tags_ensure_capacity(_tags,count); + if(ret<0)return ret; + } + ncomments=(int)count; + for(ci=0;cilen>>2)return OP_EBADHEADER; + count=op_parse_uint32le(_data); + _data+=4; + len-=4; + if(count>len)return OP_EBADHEADER; + /*Check for overflow (the API limits this to an int).*/ + if(count>(opus_uint32)INT_MAX)return OP_EFAULT; + if(_tags!=NULL){ + _tags->user_comments[ci]=op_strdup_with_len((char *)_data,count); + if(_tags->user_comments[ci]==NULL)return OP_EFAULT; + _tags->comment_lengths[ci]=(int)count; + _tags->comments=ci+1; + } + _data+=count; + len-=count; + } + return 0; +} + +int opus_tags_parse(OpusTags *_tags,const unsigned char *_data,size_t _len){ + if(_tags!=NULL){ + OpusTags tags; + int ret; + opus_tags_init(&tags); + ret=opus_tags_parse_impl(&tags,_data,_len); + if(ret<0)opus_tags_clear(&tags); + else *_tags=*&tags; + return ret; + } + else return opus_tags_parse_impl(NULL,_data,_len); +} + +/*The actual implementation of opus_tags_copy(). + Unlike the public API, this function requires _dst to already be + initialized, modifies its contents before success is guaranteed, and assumes + the caller will clear it on error.*/ +static int opus_tags_copy_impl(OpusTags *_dst,const OpusTags *_src){ + char *vendor; + int ncomments; + int ret; + int ci; + vendor=_src->vendor; + _dst->vendor=op_strdup_with_len(vendor,strlen(vendor)); + if(OP_UNLIKELY(_dst->vendor==NULL))return OP_EFAULT; + ncomments=_src->comments; + ret=op_tags_ensure_capacity(_dst,ncomments); + if(OP_UNLIKELY(ret<0))return ret; + for(ci=0;cicomment_lengths[ci]; + OP_ASSERT(len>=0); + _dst->user_comments[ci]=op_strdup_with_len(_src->user_comments[ci],len); + if(OP_UNLIKELY(_dst->user_comments[ci]==NULL))return OP_EFAULT; + _dst->comment_lengths[ci]=len; + _dst->comments=ci+1; + } + return 0; +} + +int opus_tags_copy(OpusTags *_dst,const OpusTags *_src){ + OpusTags dst; + int ret; + opus_tags_init(&dst); + ret=opus_tags_copy_impl(&dst,_src); + if(OP_UNLIKELY(ret<0))opus_tags_clear(&dst); + else *_dst=*&dst; + return 0; +} + +int opus_tags_add(OpusTags *_tags,const char *_tag,const char *_value){ + char *comment; + int tag_len; + int value_len; + int ncomments; + int ret; + ncomments=_tags->comments; + ret=op_tags_ensure_capacity(_tags,ncomments+1); + if(OP_UNLIKELY(ret<0))return ret; + tag_len=strlen(_tag); + value_len=strlen(_value); + /*+2 for '=' and '\0'.*/ + _tags->comment_lengths[ncomments]=0; + _tags->user_comments[ncomments]=comment= + (char *)_ogg_malloc(sizeof(*comment)*(tag_len+value_len+2)); + if(OP_UNLIKELY(comment==NULL))return OP_EFAULT; + _tags->comment_lengths[ncomments]=tag_len+value_len+1; + memcpy(comment,_tag,sizeof(*comment)*tag_len); + comment[tag_len]='='; + memcpy(comment+tag_len+1,_value,sizeof(*comment)*(value_len+1)); + return 0; +} + +int opus_tags_add_comment(OpusTags *_tags,const char *_comment){ + int comment_len; + int ncomments; + int ret; + ncomments=_tags->comments; + ret=op_tags_ensure_capacity(_tags,ncomments+1); + if(OP_UNLIKELY(ret<0))return ret; + comment_len=(int)strlen(_comment); + _tags->comment_lengths[ncomments]=0; + _tags->user_comments[ncomments]=op_strdup_with_len(_comment,comment_len); + if(OP_UNLIKELY(_tags->user_comments[ncomments]==NULL))return OP_EFAULT; + _tags->comment_lengths[ncomments]=comment_len; + return 0; +} + +int opus_tagcompare(const char *_tag_name,const char *_comment){ + return opus_tagncompare(_tag_name,strlen(_tag_name),_comment); +} + +int opus_tagncompare(const char *_tag_name,int _tag_len,const char *_comment){ + int ret; + OP_ASSERT(_tag_len>=0); + ret=op_strncasecmp(_tag_name,_comment,_tag_len); + return ret?ret:'='-_comment[_tag_len]; +} + +const char *opus_tags_query(const OpusTags *_tags,const char *_tag,int _count){ + char **user_comments; + int tag_len; + int found; + int ncomments; + int ci; + tag_len=strlen(_tag); + ncomments=_tags->comments; + user_comments=_tags->user_comments; + found=0; + for(ci=0;cicomments; + user_comments=_tags->user_comments; + found=0; + for(ci=0;ciuser_comments; + ncomments=_tags->comments; + /*Look for the first valid R128_TRACK_GAIN tag and use that.*/ + for(ci=0;ci='0'&&*p<='9'){ + gain_q8=10*gain_q8+*p-'0'; + if(gain_q8>32767-negative)break; + p++; + } + /*This didn't look like a signed 16-bit decimal integer. + Not a valid R128_TRACK_GAIN tag.*/ + if(*p!='\0')continue; + *_gain_q8=(int)(gain_q8+negative^negative); + return 0; + } + } + return OP_FALSE; +} + +static int op_is_jpeg(const unsigned char *_buf,size_t _buf_sz){ + return _buf_sz>=11&&memcmp(_buf,"\xFF\xD8\xFF\xE0",4)==0 + &&(_buf[4]<<8|_buf[5])>=16&&memcmp(_buf+6,"JFIF",5)==0; +} + +/*Tries to extract the width, height, bits per pixel, and palette size of a + JPEG. + On failure, simply leaves its outputs unmodified.*/ +static void op_extract_jpeg_params(const unsigned char *_buf,size_t _buf_sz, + opus_uint32 *_width,opus_uint32 *_height, + opus_uint32 *_depth,opus_uint32 *_colors,int *_has_palette){ + if(op_is_jpeg(_buf,_buf_sz)){ + size_t offs; + offs=2; + for(;;){ + size_t segment_len; + int marker; + while(offs<_buf_sz&&_buf[offs]!=0xFF)offs++; + while(offs<_buf_sz&&_buf[offs]==0xFF)offs++; + marker=_buf[offs]; + offs++; + /*If we hit EOI* (end of image), or another SOI* (start of image), + or SOS (start of scan), then stop now.*/ + if(offs>=_buf_sz||(marker>=0xD8&&marker<=0xDA))break; + /*RST* (restart markers): skip (no segment length).*/ + else if(marker>=0xD0&&marker<=0xD7)continue; + /*Read the length of the marker segment.*/ + if(_buf_sz-offs<2)break; + segment_len=_buf[offs]<<8|_buf[offs+1]; + if(segment_len<2||_buf_sz-offs0xC0&&marker<0xD0&&(marker&3)!=0)){ + /*Found a SOFn (start of frame) marker segment:*/ + if(segment_len>=8){ + *_height=_buf[offs+3]<<8|_buf[offs+4]; + *_width=_buf[offs+5]<<8|_buf[offs+6]; + *_depth=_buf[offs+2]*_buf[offs+7]; + *_colors=0; + *_has_palette=0; + } + break; + } + /*Other markers: skip the whole marker segment.*/ + offs+=segment_len; + } + } +} + +static int op_is_png(const unsigned char *_buf,size_t _buf_sz){ + return _buf_sz>=8&&memcmp(_buf,"\x89PNG\x0D\x0A\x1A\x0A",8)==0; +} + +/*Tries to extract the width, height, bits per pixel, and palette size of a + PNG. + On failure, simply leaves its outputs unmodified.*/ +static void op_extract_png_params(const unsigned char *_buf,size_t _buf_sz, + opus_uint32 *_width,opus_uint32 *_height, + opus_uint32 *_depth,opus_uint32 *_colors,int *_has_palette){ + if(op_is_png(_buf,_buf_sz)){ + size_t offs; + offs=8; + while(_buf_sz-offs>=12){ + ogg_uint32_t chunk_len; + chunk_len=op_parse_uint32be(_buf+offs); + if(chunk_len>_buf_sz-(offs+12))break; + else if(chunk_len==13&&memcmp(_buf+offs+4,"IHDR",4)==0){ + int color_type; + *_width=op_parse_uint32be(_buf+offs+8); + *_height=op_parse_uint32be(_buf+offs+12); + color_type=_buf[offs+17]; + if(color_type==3){ + *_depth=24; + *_has_palette=1; + } + else{ + int sample_depth; + sample_depth=_buf[offs+16]; + if(color_type==0)*_depth=sample_depth; + else if(color_type==2)*_depth=sample_depth*3; + else if(color_type==4)*_depth=sample_depth*2; + else if(color_type==6)*_depth=sample_depth*4; + *_colors=0; + *_has_palette=0; + break; + } + } + else if(*_has_palette>0&&memcmp(_buf+offs+4,"PLTE",4)==0){ + *_colors=chunk_len/3; + break; + } + offs+=12+chunk_len; + } + } +} + +static int op_is_gif(const unsigned char *_buf,size_t _buf_sz){ + return _buf_sz>=6&&(memcmp(_buf,"GIF87a",6)==0||memcmp(_buf,"GIF89a",6)==0); +} + +/*Tries to extract the width, height, bits per pixel, and palette size of a + GIF. + On failure, simply leaves its outputs unmodified.*/ +static void op_extract_gif_params(const unsigned char *_buf,size_t _buf_sz, + opus_uint32 *_width,opus_uint32 *_height, + opus_uint32 *_depth,opus_uint32 *_colors,int *_has_palette){ + if(op_is_gif(_buf,_buf_sz)&&_buf_sz>=14){ + *_width=_buf[6]|_buf[7]<<8; + *_height=_buf[8]|_buf[9]<<8; + /*libFLAC hard-codes the depth to 24.*/ + *_depth=24; + *_colors=1<<((_buf[10]&7)+1); + *_has_palette=1; + } +} + +/*The actual implementation of opus_picture_tag_parse(). + Unlike the public API, this function requires _pic to already be + initialized, modifies its contents before success is guaranteed, and assumes + the caller will clear it on error.*/ +static int opus_picture_tag_parse_impl(OpusPictureTag *_pic,const char *_tag, + unsigned char *_buf,size_t _buf_sz,size_t _base64_sz){ + opus_int32 picture_type; + opus_uint32 mime_type_length; + char *mime_type; + opus_uint32 description_length; + char *description; + opus_uint32 width; + opus_uint32 height; + opus_uint32 depth; + opus_uint32 colors; + opus_uint32 data_length; + opus_uint32 file_width; + opus_uint32 file_height; + opus_uint32 file_depth; + opus_uint32 file_colors; + int format; + int has_palette; + int colors_set; + size_t i; + /*Decode the BASE64 data.*/ + for(i=0;i<_base64_sz;i++){ + opus_uint32 value; + int j; + value=0; + for(j=0;j<4;j++){ + unsigned c; + unsigned d; + c=(unsigned char)_tag[4*i+j]; + if(c=='+')d=62; + else if(c=='/')d=63; + else if(c>='0'&&c<='9')d=52+c-'0'; + else if(c>='a'&&c<='z')d=26+c-'a'; + else if(c>='A'&&c<='Z')d=c-'A'; + else if(c=='='&&3*i+j>_buf_sz)d=0; + else return OP_ENOTFORMAT; + value=value<<6|d; + } + _buf[3*i]=(unsigned char)(value>>16); + if(3*i+1<_buf_sz){ + _buf[3*i+1]=(unsigned char)(value>>8); + if(3*i+2<_buf_sz)_buf[3*i+2]=(unsigned char)value; + } + } + i=0; + picture_type=op_parse_uint32be(_buf+i); + i+=4; + /*Extract the MIME type.*/ + mime_type_length=op_parse_uint32be(_buf+i); + i+=4; + if(mime_type_length>_buf_sz-32)return OP_ENOTFORMAT; + mime_type=(char *)_ogg_malloc(sizeof(*_pic->mime_type)*(mime_type_length+1)); + if(mime_type==NULL)return OP_EFAULT; + memcpy(mime_type,_buf+i,sizeof(*mime_type)*mime_type_length); + mime_type[mime_type_length]='\0'; + _pic->mime_type=mime_type; + i+=mime_type_length; + /*Extract the description string.*/ + description_length=op_parse_uint32be(_buf+i); + i+=4; + if(description_length>_buf_sz-mime_type_length-32)return OP_ENOTFORMAT; + description= + (char *)_ogg_malloc(sizeof(*_pic->mime_type)*(description_length+1)); + if(description==NULL)return OP_EFAULT; + memcpy(description,_buf+i,sizeof(*description)*description_length); + description[description_length]='\0'; + _pic->description=description; + i+=description_length; + /*Extract the remaining fields.*/ + width=op_parse_uint32be(_buf+i); + i+=4; + height=op_parse_uint32be(_buf+i); + i+=4; + depth=op_parse_uint32be(_buf+i); + i+=4; + colors=op_parse_uint32be(_buf+i); + i+=4; + /*If one of these is set, they all must be, but colors==0 is a valid value.*/ + colors_set=width!=0||height!=0||depth!=0||colors!=0; + if(width==0||height==0||depth==0&&colors_set)return OP_ENOTFORMAT; + data_length=op_parse_uint32be(_buf+i); + i+=4; + if(data_length>_buf_sz-i)return OP_ENOTFORMAT; + /*Trim extraneous data so we don't copy it below.*/ + _buf_sz=i+data_length; + /*Attempt to determine the image format.*/ + format=OP_PIC_FORMAT_UNKNOWN; + if(mime_type_length==3&&strcmp(mime_type,"-->")==0){ + format=OP_PIC_FORMAT_URL; + /*Picture type 1 must be a 32x32 PNG.*/ + if(picture_type==1&&(width!=0||height!=0)&&(width!=32||height!=32)){ + return OP_ENOTFORMAT; + } + /*Append a terminating NUL for the convenience of our callers.*/ + _buf[_buf_sz++]='\0'; + } + else{ + if(mime_type_length==10 + &&op_strncasecmp(mime_type,"image/jpeg",mime_type_length)==0){ + if(op_is_jpeg(_buf+i,data_length))format=OP_PIC_FORMAT_JPEG; + } + else if(mime_type_length==9 + &&op_strncasecmp(mime_type,"image/png",mime_type_length)==0){ + if(op_is_png(_buf+i,data_length))format=OP_PIC_FORMAT_PNG; + } + else if(mime_type_length==9 + &&op_strncasecmp(mime_type,"image/gif",mime_type_length)==0){ + if(op_is_gif(_buf+i,data_length))format=OP_PIC_FORMAT_GIF; + } + else if(mime_type_length==0||(mime_type_length==6 + &&op_strncasecmp(mime_type,"image/",mime_type_length)==0)){ + if(op_is_jpeg(_buf+i,data_length))format=OP_PIC_FORMAT_JPEG; + else if(op_is_png(_buf+i,data_length))format=OP_PIC_FORMAT_PNG; + else if(op_is_gif(_buf+i,data_length))format=OP_PIC_FORMAT_GIF; + } + file_width=file_height=file_depth=file_colors=0; + has_palette=-1; + switch(format){ + case OP_PIC_FORMAT_JPEG:{ + op_extract_jpeg_params(_buf+i,data_length, + &file_width,&file_height,&file_depth,&file_colors,&has_palette); + }break; + case OP_PIC_FORMAT_PNG:{ + op_extract_png_params(_buf+i,data_length, + &file_width,&file_height,&file_depth,&file_colors,&has_palette); + }break; + case OP_PIC_FORMAT_GIF:{ + op_extract_gif_params(_buf+i,data_length, + &file_width,&file_height,&file_depth,&file_colors,&has_palette); + }break; + } + if(has_palette>=0){ + /*If we successfully extracted these parameters from the image, override + any declared values.*/ + width=file_width; + height=file_height; + depth=file_depth; + colors=file_colors; + } + /*Picture type 1 must be a 32x32 PNG.*/ + if(picture_type==1&&(format!=OP_PIC_FORMAT_PNG||width!=32||height!=32)){ + return OP_ENOTFORMAT; + } + } + /*Adjust _buf_sz instead of using data_length to capture the terminating NUL + for URLs.*/ + _buf_sz-=i; + memmove(_buf,_buf+i,sizeof(*_buf)*_buf_sz); + _buf=(unsigned char *)_ogg_realloc(_buf,_buf_sz); + if(_buf_sz>0&&_buf==NULL)return OP_EFAULT; + _pic->type=picture_type; + _pic->width=width; + _pic->height=height; + _pic->depth=depth; + _pic->colors=colors; + _pic->data_length=data_length; + _pic->data=_buf; + _pic->format=format; + return 0; +} + +int opus_picture_tag_parse(OpusPictureTag *_pic,const char *_tag){ + OpusPictureTag pic; + unsigned char *buf; + size_t base64_sz; + size_t buf_sz; + size_t tag_length; + int ret; + if(opus_tagncompare("METADATA_BLOCK_PICTURE",22,_tag)==0)_tag+=23; + /*Figure out how much BASE64-encoded data we have.*/ + tag_length=strlen(_tag); + if(tag_length&3)return OP_ENOTFORMAT; + base64_sz=tag_length>>2; + buf_sz=3*base64_sz; + if(buf_sz<32)return OP_ENOTFORMAT; + if(_tag[tag_length-1]=='=')buf_sz--; + if(_tag[tag_length-2]=='=')buf_sz--; + if(buf_sz<32)return OP_ENOTFORMAT; + /*Allocate an extra byte to allow appending a terminating NUL to URL data.*/ + buf=(unsigned char *)_ogg_malloc(sizeof(*buf)*(buf_sz+1)); + if(buf==NULL)return OP_EFAULT; + opus_picture_tag_init(&pic); + ret=opus_picture_tag_parse_impl(&pic,_tag,buf,buf_sz,base64_sz); + if(ret<0){ + opus_picture_tag_clear(&pic); + _ogg_free(buf); + } + else *_pic=*&pic; + return ret; +} + +void opus_picture_tag_init(OpusPictureTag *_pic){ + memset(_pic,0,sizeof(*_pic)); +} + +void opus_picture_tag_clear(OpusPictureTag *_pic){ + _ogg_free(_pic->description); + _ogg_free(_pic->mime_type); + _ogg_free(_pic->data); +} diff --git a/code/opusfile-0.2/src/internal.c b/code/opusfile-0.5/src/internal.c similarity index 96% rename from code/opusfile-0.2/src/internal.c rename to code/opusfile-0.5/src/internal.c index 2d2e3c85..96c80def 100644 --- a/code/opusfile-0.2/src/internal.c +++ b/code/opusfile-0.5/src/internal.c @@ -9,6 +9,10 @@ * by the Xiph.Org Foundation and contributors http://www.xiph.org/ * * * ********************************************************************/ +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + #include "internal.h" #if defined(OP_ENABLE_ASSERTIONS) diff --git a/code/opusfile-0.2/src/internal.h b/code/opusfile-0.5/src/internal.h similarity index 85% rename from code/opusfile-0.2/src/internal.h rename to code/opusfile-0.5/src/internal.h index 79416ae3..08114919 100644 --- a/code/opusfile-0.2/src/internal.h +++ b/code/opusfile-0.5/src/internal.h @@ -32,10 +32,22 @@ # include typedef struct OggOpusLink OggOpusLink; + # if defined(OP_FIXED_POINT) + typedef opus_int16 op_sample; + # else + typedef float op_sample; + +/*We're using this define to test for libopus 1.1 or later until libopus + provides a better mechanism.*/ +# if defined(OPUS_GET_EXPERT_FRAME_DURATION_REQUEST) +/*Enable soft clipping prevention in 16-bit decodes.*/ +# define OP_SOFT_CLIP (1) +# endif + # endif # if OP_GNUC_PREREQ(4,2) @@ -80,8 +92,10 @@ void op_fatal_impl(const char *_str,const char *_file,int _line); # define OP_ALWAYS_TRUE(_cond) ((void)(_cond)) # endif -# define OP_INT64_MAX ((ogg_int64_t)0x7FFFFFFFFFFFFFFFLL) +# define OP_INT64_MAX (2*(((ogg_int64_t)1<<62)-1)|1) # define OP_INT64_MIN (-OP_INT64_MAX-1) +# define OP_INT32_MAX (2*(((ogg_int32_t)1<<30)-1)|1) +# define OP_INT32_MIN (-OP_INT32_MAX-1) # define OP_MIN(_a,_b) ((_a)<(_b)?(_a):(_b)) # define OP_MAX(_a,_b) ((_a)>(_b)?(_a):(_b)) @@ -90,7 +104,7 @@ void op_fatal_impl(const char *_str,const char *_file,int _line); /*Advance a file offset by the given amount, clamping against OP_INT64_MAX. This is used to advance a known offset by things like OP_CHUNK_SIZE or OP_PAGE_SIZE_MAX, while making sure to avoid signed overflow. - It assumes that both _offset and _amount are positive.*/ + It assumes that both _offset and _amount are non-negative.*/ #define OP_ADV_OFFSET(_offset,_amount) \ (OP_MIN(_offset,OP_INT64_MAX-(_amount))+(_amount)) @@ -189,6 +203,10 @@ struct OggOpusFile{ int op_count; /*Central working state for the packet-to-PCM decoder.*/ OpusMSDecoder *od; + /*The application-provided packet decode callback.*/ + op_decode_cb_func decode_cb; + /*The application-provided packet decode callback context.*/ + void *decode_cb_ctx; /*The stream count used to initialize the decoder.*/ int od_stream_count; /*The coupled stream count used to initialize the decoder.*/ @@ -203,12 +221,26 @@ struct OggOpusFile{ int od_buffer_pos; /*The number of valid samples in the decoded buffer.*/ int od_buffer_size; - /*Internal state for dithering float->short output.*/ + /*The type of gain offset to apply. + One of OP_HEADER_GAIN, OP_TRACK_GAIN, or OP_ABSOLUTE_GAIN.*/ + int gain_type; + /*The offset to apply to the gain.*/ + opus_int32 gain_offset_q8; + /*Internal state for soft clipping and dithering float->short output.*/ #if !defined(OP_FIXED_POINT) +# if defined(OP_SOFT_CLIP) + float clip_state[OP_NCHANNELS_MAX]; +# endif float dither_a[OP_NCHANNELS_MAX*4]; float dither_b[OP_NCHANNELS_MAX*4]; - int dither_mute; opus_uint32 dither_seed; + int dither_mute; + int dither_disabled; + /*The number of channels represented by the internal state. + This gets set to 0 whenever anything that would prevent state propagation + occurs (switching between the float/short APIs, or between the + stereo/multistream APIs).*/ + int state_channel_count; #endif }; diff --git a/code/opusfile-0.2/src/opusfile.c b/code/opusfile-0.5/src/opusfile.c similarity index 86% rename from code/opusfile-0.2/src/opusfile.c rename to code/opusfile-0.5/src/opusfile.c index 812c2c45..392ddb29 100644 --- a/code/opusfile-0.2/src/opusfile.c +++ b/code/opusfile-0.5/src/opusfile.c @@ -14,6 +14,10 @@ last mod: $Id: vorbisfile.c 17573 2010-10-27 14:53:59Z xiphmont $ ********************************************************************/ +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + #include "internal.h" #include #include @@ -163,7 +167,7 @@ static int op_seek_helper(OggOpusFile *_of,opus_int64 _offset){ /*Get the current position indicator of the underlying source. This should be the same as the value reported by tell().*/ -static opus_int64 op_position(OggOpusFile *_of){ +static opus_int64 op_position(const OggOpusFile *_of){ /*The current position indicator is _not_ simply offset. We may also have unprocessed, buffered data in the sync state.*/ return _of->offset+_of->oy.fill-_of->oy.returned; @@ -182,9 +186,8 @@ static opus_int64 op_position(OggOpusFile *_of){ OP_BADLINK: We hit end-of-file before reaching _boundary.*/ static opus_int64 op_get_next_page(OggOpusFile *_of,ogg_page *_og, opus_int64 _boundary){ - for(;;){ + while(_boundary<=0||_of->offset<_boundary){ int more; - if(_boundary>0&&_of->offset>=_boundary)return OP_FALSE; more=ogg_sync_pageseek(&_of->oy,_og); /*Skipped (-more) bytes.*/ if(OP_UNLIKELY(more<0))_of->offset-=more; @@ -211,17 +214,19 @@ static opus_int64 op_get_next_page(OggOpusFile *_of,ogg_page *_og, } else{ /*Got a page. - Return the offset at the page beginning, advance the internal offset - past the page end.*/ + Return the page start offset and advance the internal offset past the + page end.*/ opus_int64 page_offset; page_offset=_of->offset; _of->offset+=more; + OP_ASSERT(page_offset>=0); return page_offset; } } + return OP_FALSE; } -static int op_add_serialno(ogg_page *_og, +static int op_add_serialno(const ogg_page *_og, ogg_uint32_t **_serialnos,int *_nserialnos,int *_cserialnos){ ogg_uint32_t *serialnos; int nserialnos; @@ -235,7 +240,8 @@ static int op_add_serialno(ogg_page *_og, if(OP_UNLIKELY(cserialnos>INT_MAX-1>>1))return OP_EFAULT; cserialnos=2*cserialnos+1; OP_ASSERT(nserialnos=0 implies we entered the if(page_gp!=-1) block at least once.*/ + gp=-1; chunk_size=OP_CHUNK_SIZE; do{ int left_link; @@ -422,7 +431,7 @@ static opus_int64 op_get_last_page(OggOpusFile *_of,ogg_int64_t *_gp, opus_int64 llret; ogg_uint32_t serialno; llret=op_get_next_page(_of,&og,end); - if(OP_UNLIKELY(llretop[op_count].packet, _of->op[op_count].bytes); if(OP_LIKELY(_durations[op_count]>0)){ @@ -990,7 +1003,7 @@ static int op_find_final_pcm_offset(OggOpusFile *_of, /*Rescale the number _x from the range [0,_from] to [0,_to]. _from and _to must be positive.*/ -opus_int64 op_rescale64(opus_int64 _x,opus_int64 _from,opus_int64 _to){ +static opus_int64 op_rescale64(opus_int64 _x,opus_int64 _from,opus_int64 _to){ opus_int64 frac; opus_int64 ret; int i; @@ -1124,7 +1137,7 @@ static int op_bisect_forward_serialno(OggOpusFile *_of, if(OP_UNLIKELY(clinks>INT_MAX-1>>1))return OP_EFAULT; clinks=2*clinks+1; OP_ASSERT(nlinkslinks=links; } @@ -1134,7 +1147,7 @@ static int op_bisect_forward_serialno(OggOpusFile *_of, (potentially not a page we care about).*/ /*Scan the seek records we already have to save us some bisection.*/ for(sri=0;srilinks=links; /*We also don't need these anymore.*/ _ogg_free(*_serialnos); @@ -1282,12 +1295,49 @@ static int op_bisect_forward_serialno(OggOpusFile *_of, return 0; } +static void op_update_gain(OggOpusFile *_of){ + OpusHead *head; + opus_int32 gain_q8; + int li; + /*If decode isn't ready, then we'll apply the gain when we initialize the + decoder.*/ + if(_of->ready_stategain_offset_q8; + li=_of->seekable?_of->cur_link:0; + head=&_of->links[li].head; + /*We don't have to worry about overflow here because the header gain and + track gain must lie in the range [-32768,32767], and the user-supplied + offset has been pre-clamped to [-98302,98303].*/ + switch(_of->gain_type){ + case OP_TRACK_GAIN:{ + int track_gain_q8; + track_gain_q8=0; + opus_tags_get_track_gain(&_of->links[li].tags,&track_gain_q8); + gain_q8+=track_gain_q8; + } + /*Fall through.*/ + case OP_HEADER_GAIN:gain_q8+=head->output_gain;break; + case OP_ABSOLUTE_GAIN:break; + default:OP_ASSERT(0); + } + gain_q8=OP_CLAMP(-32768,gain_q8,32767); + OP_ASSERT(_of->od!=NULL); +#if defined(OPUS_SET_GAIN) + opus_multistream_decoder_ctl(_of->od,OPUS_SET_GAIN(gain_q8)); +#else +/*A fallback that works with both float and fixed-point is a bunch of work, + so just force people to use a sufficiently new version. + This is deployed well enough at this point that this shouldn't be a burden.*/ +# error "libopus 1.0.1 or later required" +#endif +} + static int op_make_decode_ready(OggOpusFile *_of){ - OpusHead *head; - int li; - int stream_count; - int coupled_count; - int channel_count; + const OpusHead *head; + int li; + int stream_count; + int coupled_count; + int channel_count; if(_of->ready_state>OP_STREAMSET)return 0; if(OP_UNLIKELY(_of->ready_stateseekable?_of->cur_link:0; @@ -1313,23 +1363,16 @@ static int op_make_decode_ready(OggOpusFile *_of){ _of->od_channel_count=channel_count; memcpy(_of->od_mapping,head->mapping,sizeof(*head->mapping)*channel_count); } -#if defined(OPUS_SET_GAIN) - opus_multistream_decoder_ctl(_of->od,OPUS_SET_GAIN(head->output_gain)); -#else -/*A fallback that works with both float and fixed-point is a bunch of work, - so just force people to use a sufficiently new version. - This is deployed well enough at this point that this shouldn't be a burden.*/ -# error "libopus 1.0.1 or later required" -#endif _of->ready_state=OP_INITSET; _of->bytes_tracked=0; _of->samples_tracked=0; #if !defined(OP_FIXED_POINT) - _of->dither_mute=65; + _of->state_channel_count=0; /*Use the serial number for the PRNG seed to get repeatable output for straight play-throughs.*/ _of->dither_seed=_of->links[li].serialno; #endif + op_update_gain(_of); return 0; } @@ -1368,7 +1411,15 @@ static int op_open_seekable2(OggOpusFile *_of){ int start_op_count; int ret; /*We're partially open and have a first link header state in storage in _of. - Save off that stream state so we can come back to it.*/ + Save off that stream state so we can come back to it. + It would be simpler to just dump all this state and seek back to + links[0].data_offset when we're done. + But we do the extra work to allow us to seek back to _exactly_ the same + stream position we're at now. + This allows, e.g., the HTTP backend to continue reading from the original + connection (if it's still available), instead of opening a new one. + This means we can open and start playing a normal Opus file with a single + link and reasonable packet sizes using only two HTTP requests.*/ start_op_count=_of->op_count; /*This is a bit too large to put on the stack unconditionally.*/ op_start=(ogg_packet *)_ogg_malloc(sizeof(*op_start)*start_op_count); @@ -1468,14 +1519,12 @@ static int op_open1(OggOpusFile *_of, seekable=_cb->seek!=NULL&&(*_cb->seek)(_source,0,SEEK_CUR)!=-1; /*If seek is implemented, tell must also be implemented.*/ if(seekable){ + opus_int64 pos; if(OP_UNLIKELY(_of->callbacks.tell==NULL))return OP_EINVAL; - else{ - opus_int64 pos; - pos=(*_of->callbacks.tell)(_of->source); - /*If the current position is not equal to the initial bytes consumed, - absolute seeking will not work.*/ - if(OP_UNLIKELY(pos!=(opus_int64)_initial_bytes))return OP_EINVAL; - } + pos=(*_of->callbacks.tell)(_of->source); + /*If the current position is not equal to the initial bytes consumed, + absolute seeking will not work.*/ + if(OP_UNLIKELY(pos!=(opus_int64)_initial_bytes))return OP_EINVAL; } _of->seekable=seekable; /*Don't seek yet. @@ -1506,12 +1555,7 @@ static int op_open1(OggOpusFile *_of, if(!seekable)_of->cur_link++; pog=&og; } - if(OP_UNLIKELY(ret<0)){ - /*Don't auto-close the stream on failure.*/ - _of->callbacks.close=NULL; - op_clear(_of); - } - else _of->ready_state=OP_PARTOPEN; + if(OP_LIKELY(ret>=0))_of->ready_state=OP_PARTOPEN; return ret; } @@ -1548,6 +1592,9 @@ OggOpusFile *op_test_callbacks(void *_source,const OpusFileCallbacks *_cb, if(_error!=NULL)*_error=0; return of; } + /*Don't auto-close the stream on failure.*/ + of->callbacks.close=NULL; + op_clear(of); _ogg_free(of); } if(_error!=NULL)*_error=ret; @@ -1594,18 +1641,6 @@ OggOpusFile *op_open_memory(const unsigned char *_data,size_t _size, _error); } -OggOpusFile *op_vopen_url(const char *_url,int *_error,va_list _ap){ - OpusFileCallbacks cb; - return op_open_close_on_failure(op_url_stream_vcreate(&cb,_url,_ap),&cb, - _error); -} - -OggOpusFile *op_open_url(const char *_url,int *_error,...){ - va_list ap; - va_start(ap,_error); - return op_vopen_url(_url,_error,ap); -} - /*Convenience routine to clean up from failure for the open functions that create their own streams.*/ static OggOpusFile *op_test_close_on_failure(void *_source, @@ -1632,18 +1667,6 @@ OggOpusFile *op_test_memory(const unsigned char *_data,size_t _size, _error); } -OggOpusFile *op_vtest_url(const char *_url,int *_error,va_list _ap){ - OpusFileCallbacks cb; - return op_test_close_on_failure(op_url_stream_vcreate(&cb,_url,_ap),&cb, - _error); -} - -OggOpusFile *op_test_url(const char *_url,int *_error,...){ - va_list ap; - va_start(ap,_error); - return op_vtest_url(_url,_error,ap); -} - int op_test_open(OggOpusFile *_of){ int ret; if(OP_UNLIKELY(_of->ready_state!=OP_PARTOPEN))return OP_EINVAL; @@ -1661,25 +1684,25 @@ void op_free(OggOpusFile *_of){ } } -int op_seekable(OggOpusFile *_of){ +int op_seekable(const OggOpusFile *_of){ return _of->seekable; } -int op_link_count(OggOpusFile *_of){ +int op_link_count(const OggOpusFile *_of){ return _of->nlinks; } -ogg_uint32_t op_serialno(OggOpusFile *_of,int _li){ +ogg_uint32_t op_serialno(const OggOpusFile *_of,int _li){ if(OP_UNLIKELY(_li>=_of->nlinks))_li=_of->nlinks-1; if(!_of->seekable)_li=0; return _of->links[_li<0?_of->cur_link:_li].serialno; } -int op_channel_count(OggOpusFile *_of,int _li){ +int op_channel_count(const OggOpusFile *_of,int _li){ return op_head(_of,_li)->channel_count; } -opus_int64 op_raw_total(OggOpusFile *_of,int _li){ +opus_int64 op_raw_total(const OggOpusFile *_of,int _li){ if(OP_UNLIKELY(_of->ready_stateseekable) ||OP_UNLIKELY(_li>=_of->nlinks)){ @@ -1690,7 +1713,7 @@ opus_int64 op_raw_total(OggOpusFile *_of,int _li){ -_of->links[_li].offset; } -ogg_int64_t op_pcm_total(OggOpusFile *_of,int _li){ +ogg_int64_t op_pcm_total(const OggOpusFile *_of,int _li){ OggOpusLink *links; ogg_int64_t diff; int nlinks; @@ -1720,13 +1743,13 @@ ogg_int64_t op_pcm_total(OggOpusFile *_of,int _li){ return diff-links[_li].head.pre_skip; } -const OpusHead *op_head(OggOpusFile *_of,int _li){ +const OpusHead *op_head(const OggOpusFile *_of,int _li){ if(OP_UNLIKELY(_li>=_of->nlinks))_li=_of->nlinks-1; if(!_of->seekable)_li=0; return &_of->links[_li<0?_of->cur_link:_li].head; } -const OpusTags *op_tags(OggOpusFile *_of,int _li){ +const OpusTags *op_tags(const OggOpusFile *_of,int _li){ if(OP_UNLIKELY(_li>=_of->nlinks))_li=_of->nlinks-1; if(!_of->seekable){ if(_of->ready_stateready_state!=OP_PARTOPEN){ @@ -1738,7 +1761,7 @@ const OpusTags *op_tags(OggOpusFile *_of,int _li){ return &_of->links[_li].tags; } -int op_current_link(OggOpusFile *_of){ +int op_current_link(const OggOpusFile *_of){ if(OP_UNLIKELY(_of->ready_statecur_link; } @@ -1749,11 +1772,13 @@ static opus_int32 op_calc_bitrate(opus_int64 _bytes,ogg_int64_t _samples){ /*These rates are absurd, but let's handle them anyway.*/ if(OP_UNLIKELY(_bytes>(OP_INT64_MAX-(_samples>>1))/(48000*8))){ ogg_int64_t den; - if(OP_UNLIKELY(_bytes/(0x7FFFFFFFF/(48000*8))>=_samples))return 0x7FFFFFFF; + if(OP_UNLIKELY(_bytes/(OP_INT32_MAX/(48000*8))>=_samples)){ + return OP_INT32_MAX; + } den=_samples/(48000*8); return (opus_int32)((_bytes+(den>>1))/den); } - if(OP_UNLIKELY(_samples<=0))return 0x7FFFFFFF; + if(OP_UNLIKELY(_samples<=0))return OP_INT32_MAX; /*This can't actually overflow in normal operation: even with a pre-skip of 545 2.5 ms frames with 8 streams running at 1282*8+1 bytes per packet (1275 byte frames + Opus framing overhead + Ogg lacing values), that all @@ -1761,10 +1786,11 @@ static opus_int32 op_calc_bitrate(opus_int64 _bytes,ogg_int64_t _samples){ The only way to get bitrates larger than that is with excessive Opus padding, more encoded streams than output channels, or lots and lots of Ogg pages with no packets on them.*/ - return (opus_int32)OP_MIN((_bytes*48000*8+(_samples>>1))/_samples,0x7FFFFFFF); + return (opus_int32)OP_MIN((_bytes*48000*8+(_samples>>1))/_samples, + OP_INT32_MAX); } -opus_int32 op_bitrate(OggOpusFile *_of,int _li){ +opus_int32 op_bitrate(const OggOpusFile *_of,int _li){ if(OP_UNLIKELY(_of->ready_stateseekable) ||OP_UNLIKELY(_li>=_of->nlinks)){ return OP_EINVAL; @@ -1800,11 +1826,8 @@ static int op_fetch_and_process_page(OggOpusFile *_of, int seekable; int cur_link; int ret; - if(OP_LIKELY(_of->ready_state>=OP_INITSET) - &&OP_LIKELY(_of->op_pos<_of->op_count)){ - /*We're ready to decode and have at least one packet available already.*/ - return 1; - } + /*We shouldn't get here if we have unprocessed packets.*/ + OP_ASSERT(_of->ready_stateop_pos>=_of->op_count); if(!_readp)return 0; seekable=_of->seekable; links=_of->links; @@ -2059,7 +2082,7 @@ static int op_fetch_and_process_page(OggOpusFile *_of, } int op_raw_seek(OggOpusFile *_of,opus_int64 _pos){ - int ret; + int ret; if(OP_UNLIKELY(_of->ready_stateseekable))return OP_ENOSEEK; @@ -2090,10 +2113,10 @@ int op_raw_seek(OggOpusFile *_of,opus_int64 _pos){ position in an individual link.*/ static ogg_int64_t op_get_granulepos(const OggOpusFile *_of, ogg_int64_t _pcm_offset,int *_li){ - OggOpusLink *links; - ogg_int64_t duration; - int nlinks; - int li; + const OggOpusLink *links; + ogg_int64_t duration; + int nlinks; + int li; OP_ASSERT(_pcm_offset>=0); nlinks=_of->nlinks; links=_of->links; @@ -2128,6 +2151,12 @@ static ogg_int64_t op_get_granulepos(const OggOpusFile *_of, Two minutes seems to be a good default.*/ #define OP_CUR_TIME_THRESH (120*48*(opus_int32)1000) +/*Note: The OP_SMALL_FOOTPRINT #define doesn't (currently) save much code size, + but it's meant to serve as documentation for portions of the seeking + algorithm that are purely optional, to aid others learning from/porting this + code to other contexts.*/ +/*#define OP_SMALL_FOOTPRINT (1)*/ + /*Search within link _li for the page with the highest granule position preceding (or equal to) _target_gp. There is a danger here: missing pages or incorrect frame number information @@ -2135,27 +2164,27 @@ static ogg_int64_t op_get_granulepos(const OggOpusFile *_of, Account for that (and report it as an error condition).*/ static int op_pcm_seek_page(OggOpusFile *_of, ogg_int64_t _target_gp,int _li){ - OggOpusLink *link; - ogg_page og; - ogg_int64_t pcm_pre_skip; - ogg_int64_t pcm_start; - ogg_int64_t pcm_end; - ogg_int64_t best_gp; - ogg_int64_t diff; - ogg_uint32_t serialno; - opus_int32 pre_skip; - opus_int32 cur_discard_count; - opus_int64 begin; - opus_int64 end; - opus_int64 boundary; - opus_int64 best; - opus_int64 page_offset; - opus_int64 d[3]; - int force_bisect; - int ret; + const OggOpusLink *link; + ogg_page og; + ogg_int64_t pcm_pre_skip; + ogg_int64_t pcm_start; + ogg_int64_t pcm_end; + ogg_int64_t best_gp; + ogg_int64_t diff; + ogg_uint32_t serialno; + opus_int32 pre_skip; + opus_int64 begin; + opus_int64 end; + opus_int64 boundary; + opus_int64 best; + opus_int64 page_offset; + opus_int64 d0; + opus_int64 d1; + opus_int64 d2; + int force_bisect; + int ret; _of->bytes_tracked=0; _of->samples_tracked=0; - /*New search algorithm by HB (Nicholas Vinen).*/ link=_of->links+_li; best_gp=pcm_start=link->pcm_start; pcm_end=link->pcm_end; @@ -2165,7 +2194,8 @@ static int op_pcm_seek_page(OggOpusFile *_of, /*We discard the first 80 ms of data after a seek, so seek back that much farther. If we can't, simply seek to the beginning of the link.*/ - if(OP_UNLIKELY(op_granpos_add(&_target_gp,_target_gp,-80*48)<0)){ + if(OP_UNLIKELY(op_granpos_add(&_target_gp,_target_gp,-80*48)<0) + ||OP_UNLIKELY(op_granpos_cmp(_target_gp,pcm_start)<0)){ _target_gp=pcm_start; } /*Special case seeking to the start of the link.*/ @@ -2174,6 +2204,7 @@ static int op_pcm_seek_page(OggOpusFile *_of, if(op_granpos_cmp(_target_gp,pcm_pre_skip)<0)end=boundary=begin; else{ end=boundary=link->end_offset; +#if !defined(OP_SMALL_FOOTPRINT) /*If we were decoding from this link, we can narrow the range a bit.*/ if(_li==_of->cur_link&&_of->ready_state>=OP_INITSET){ opus_int64 offset; @@ -2187,15 +2218,15 @@ static int op_pcm_seek_page(OggOpusFile *_of, offset=_of->offset; if(op_count>0&&OP_LIKELY(offset<=end)){ ogg_int64_t gp; - gp=_of->op[op_count-1].granulepos; /*Make sure the timestamp is valid. The granule position might be -1 if we collected the packets from a page without a granule position after reporting a hole.*/ + gp=_of->op[op_count-1].granulepos; if(OP_LIKELY(gp!=-1)&&OP_LIKELY(op_granpos_cmp(pcm_start,gp)<0) &&OP_LIKELY(op_granpos_cmp(pcm_end,gp)>0)){ OP_ALWAYS_TRUE(!op_granpos_diff(&diff,gp,_target_gp)); /*We only actually use the current time if either - a) We can cut off more than half the range, or + a) We can cut off at least half the range, or b) We're seeking sufficiently close to the current position that it's likely to be informative. Otherwise it appears using the whole link range to estimate the @@ -2207,18 +2238,44 @@ static int op_pcm_seek_page(OggOpusFile *_of, best_gp=pcm_start=gp; } } - else if(offset-begin<=end-begin>>1||diffop[0].granulepos, + op_get_packet_duration(_of->op[0].packet,_of->op[0].bytes))); + if(op_granpos_cmp(prev_page_gp,_target_gp)<=0){ + /*Don't call op_decode_clear(), because it will dump our + packets.*/ + _of->op_pos=0; + _of->od_buffer_size=0; + _of->prev_packet_gp=prev_page_gp; + _of->ready_state=OP_STREAMSET; + return op_make_decode_ready(_of); + } + /*No such luck. + Check if we can cut off at least half the range, though.*/ + if(offset-begin<=end-begin>>1||diff>1; - d[1]=d[2]>>1; - d[2]=end-begin>>1; + d0=d1>>1; + d1=d2>>1; + d2=end-begin>>1; if(force_bisect)bisect=begin+(end-begin>>1); else{ ogg_int64_t diff2; @@ -2307,7 +2364,7 @@ static int op_pcm_seek_page(OggOpusFile *_of, boundary=next_boundary; /*If we're not making much progress shrinking the interval size, start forcing straight bisection to limit the worst case.*/ - force_bisect=end-begin>d[0]*2; + force_bisect=end-begin>d0*2; /*Don't let pcm_end get out of range! That could happen with an invalid timestamp.*/ if(OP_LIKELY(op_granpos_cmp(pcm_end,gp)>0) @@ -2321,7 +2378,8 @@ static int op_pcm_seek_page(OggOpusFile *_of, } } /*Found our page. - Seek right after it and update prev_packet_gp and cur_discard_count. + Seek to the end of it and update prev_packet_gp. + Our caller will set cur_discard_count. This is an easier case than op_raw_seek(), as we don't need to keep any packets from the page we found.*/ /*Seek, if necessary.*/ @@ -2330,21 +2388,10 @@ static int op_pcm_seek_page(OggOpusFile *_of, ret=op_seek_helper(_of,best); if(OP_UNLIKELY(ret<0))return ret; } - /*By default, discard 80 ms of data after a seek, unless we seek - into the pre-skip region.*/ - cur_discard_count=80*48; - OP_ALWAYS_TRUE(!op_granpos_diff(&diff,best_gp,pcm_start)); - OP_ASSERT(diff>=0); - /*If we start at the beginning of the pre-skip region, or we're at least - 80 ms from the end of the pre-skip region, we discard to the end of the - pre-skip region. - Otherwise, we still use the 80 ms default, which will discard past the end - of the pre-skip region.*/ - if(diff<=OP_MAX(0,pre_skip-80*48))cur_discard_count=pre_skip-(int)diff; + OP_ASSERT(op_granpos_cmp(best_gp,pcm_start)>=0); _of->cur_link=_li; _of->ready_state=OP_STREAMSET; _of->prev_packet_gp=best_gp; - _of->cur_discard_count=cur_discard_count; ogg_stream_reset_serialno(&_of->os,serialno); ret=op_fetch_and_process_page(_of,page_offset<0?NULL:&og,page_offset,1,0,1); if(OP_UNLIKELY(ret<=0))return OP_EBADLINK; @@ -2356,31 +2403,60 @@ static int op_pcm_seek_page(OggOpusFile *_of, } int op_pcm_seek(OggOpusFile *_of,ogg_int64_t _pcm_offset){ - OggOpusLink *link; - ogg_int64_t pcm_start; - ogg_int64_t target_gp; - ogg_int64_t prev_packet_gp; - ogg_int64_t skip; - ogg_int64_t diff; - int op_count; - int op_pos; - int ret; - int li; + const OggOpusLink *link; + ogg_int64_t pcm_start; + ogg_int64_t target_gp; + ogg_int64_t prev_packet_gp; + ogg_int64_t skip; + ogg_int64_t diff; + int op_count; + int op_pos; + int ret; + int li; if(OP_UNLIKELY(_of->ready_stateseekable))return OP_ENOSEEK; if(OP_UNLIKELY(_pcm_offset<0))return OP_EINVAL; target_gp=op_get_granulepos(_of,_pcm_offset,&li); if(OP_UNLIKELY(target_gp==-1))return OP_EINVAL; - ret=op_pcm_seek_page(_of,target_gp,li); - /*Now skip samples until we actually get to our target.*/ link=_of->links+li; pcm_start=link->pcm_start; OP_ALWAYS_TRUE(!op_granpos_diff(&_pcm_offset,target_gp,pcm_start)); +#if !defined(OP_SMALL_FOOTPRINT) + /*For small (90 ms or less) forward seeks within the same link, just decode + forward. + This also optimizes the case of seeking to the current position.*/ + if(li==_of->cur_link&&_of->ready_state>=OP_INITSET){ + ogg_int64_t gp; + gp=_of->prev_packet_gp; + if(OP_LIKELY(gp!=-1)){ + int nbuffered; + nbuffered=OP_MAX(_of->od_buffer_size-_of->od_buffer_pos,0); + OP_ALWAYS_TRUE(!op_granpos_add(&gp,gp,-nbuffered)); + /*We do _not_ add cur_discard_count to gp. + Otherwise the total amount to discard could grow without bound, and it + would be better just to do a full seek.*/ + if(OP_LIKELY(!op_granpos_diff(&diff,gp,pcm_start))){ + ogg_int64_t discard_count; + discard_count=_pcm_offset-diff; + /*We use a threshold of 90 ms instead of 80, since 80 ms is the + _minimum_ we would have discarded after a full seek. + Assuming 20 ms frames (the default), we'd discard 90 ms on average.*/ + if(discard_count>=0&&OP_UNLIKELY(discard_count<90*48)){ + _of->cur_discard_count=(opus_int32)discard_count; + return 0; + } + } + } + } +#endif + ret=op_pcm_seek_page(_of,target_gp,li); + if(OP_UNLIKELY(ret<0))return ret; + /*Now skip samples until we actually get to our target.*/ /*Figure out where we should skip to.*/ if(_pcm_offset<=link->head.pre_skip)skip=0; else skip=OP_MAX(_pcm_offset-80*48,0); OP_ASSERT(_pcm_offset-skip>=0); - OP_ASSERT(_pcm_offset-skip<0x7FFFFFFF-120*48); + OP_ASSERT(_pcm_offset-skipop_count; @@ -2406,7 +2482,7 @@ int op_pcm_seek(OggOpusFile *_of,ogg_int64_t _pcm_offset){ /*We skipped too far. Either the timestamps were illegal or there was a hole in the data.*/ if(diff>skip)return OP_EBADLINK; - OP_ASSERT(_pcm_offset-diff<0x7FFFFFFF); + OP_ASSERT(_pcm_offset-diffready_stateoffset; } @@ -2425,10 +2501,10 @@ opus_int64 op_raw_tell(OggOpusFile *_of){ For unseekable sources, this gets reset to 0 at the beginning of each link.*/ static ogg_int64_t op_get_pcm_offset(const OggOpusFile *_of, ogg_int64_t _gp,int _li){ - OggOpusLink *links; - ogg_int64_t pcm_offset; - ogg_int64_t delta; - int li; + const OggOpusLink *links; + ogg_int64_t pcm_offset; + ogg_int64_t delta; + int li; links=_of->links; pcm_offset=0; OP_ASSERT(_li<_of->nlinks); @@ -2443,15 +2519,23 @@ static ogg_int64_t op_get_pcm_offset(const OggOpusFile *_of, _gp=links[_li].pcm_end; } if(OP_LIKELY(op_granpos_cmp(_gp,links[_li].pcm_start)>0)){ - OP_ALWAYS_TRUE(!op_granpos_diff(&delta,_gp,links[_li].pcm_start)); + if(OP_UNLIKELY(op_granpos_diff(&delta,_gp,links[_li].pcm_start)<0)){ + /*This means an unseekable stream claimed to have a page from more than + 2 billion days after we joined.*/ + OP_ASSERT(!_of->seekable); + return OP_INT64_MAX; + } if(deltadecode_cb=_decode_cb; + _of->decode_cb_ctx=_ctx; +} + +int op_set_gain_offset(OggOpusFile *_of, + int _gain_type,opus_int32 _gain_offset_q8){ + if(_gain_type!=OP_HEADER_GAIN&&_gain_type!=OP_TRACK_GAIN + &&_gain_type!=OP_ABSOLUTE_GAIN){ + return OP_EINVAL; + } + _of->gain_type=_gain_type; + /*The sum of header gain and track gain lies in the range [-65536,65534]. + These bounds allow the offset to set the final value to anywhere in the + range [-32768,32767], which is what we'll clamp it to before applying.*/ + _of->gain_offset_q8=OP_CLAMP(-98302,_gain_offset_q8,98303); + op_update_gain(_of); + return 0; +} + +void op_set_dither_enabled(OggOpusFile *_of,int _enabled){ +#if !defined(OP_FIXED_POINT) + _of->dither_disabled=!_enabled; + if(!_enabled)_of->dither_mute=65; +#endif +} + /*Allocate the decoder scratch buffer. This is done lazily, since if the user provides large enough buffers, we'll never need it.*/ static int op_init_buffer(OggOpusFile *_of){ int nchannels_max; if(_of->seekable){ - OggOpusLink *links; - int nlinks; - int li; + const OggOpusLink *links; + int nlinks; + int li; links=_of->links; nlinks=_of->nlinks; nchannels_max=1; @@ -2490,6 +2602,39 @@ static int op_init_buffer(OggOpusFile *_of){ return 0; } +/*Decode a single packet into the target buffer.*/ +static int op_decode(OggOpusFile *_of,op_sample *_pcm, + const ogg_packet *_op,int _nsamples,int _nchannels){ + int ret; + /*First we try using the application-provided decode callback.*/ + if(_of->decode_cb!=NULL){ +#if defined(OP_FIXED_POINT) + ret=(*_of->decode_cb)(_of->decode_cb_ctx,_of->od,_pcm,_op, + _nsamples,_nchannels,OP_DEC_FORMAT_SHORT,_of->cur_link); +#else + ret=(*_of->decode_cb)(_of->decode_cb_ctx,_of->od,_pcm,_op, + _nsamples,_nchannels,OP_DEC_FORMAT_FLOAT,_of->cur_link); +#endif + } + else ret=OP_DEC_USE_DEFAULT; + /*If the application didn't want to handle decoding, do it ourselves.*/ + if(ret==OP_DEC_USE_DEFAULT){ +#if defined(OP_FIXED_POINT) + ret=opus_multistream_decode(_of->od, + _op->packet,_op->bytes,_pcm,_nsamples,0); +#else + ret=opus_multistream_decode_float(_of->od, + _op->packet,_op->bytes,_pcm,_nsamples,0); +#endif + OP_ASSERT(ret<0||ret==_nsamples); + } + /*If the application returned a positive value other than 0 or + OP_DEC_USE_DEFAULT, fail.*/ + else if(OP_UNLIKELY(ret>0))return OP_EBADPACKET; + if(OP_UNLIKELY(ret<0))return OP_EBADPACKET; + return ret; +} + /*Read more samples from the stream, using the same API as op_read() or op_read_float().*/ static int op_read_native(OggOpusFile *_of, @@ -2506,10 +2651,8 @@ static int op_read_native(OggOpusFile *_of, od_buffer_pos=_of->od_buffer_pos; nsamples=_of->od_buffer_size-od_buffer_pos; /*If we have buffered samples, return them.*/ - if(OP_UNLIKELY(nsamples>0)){ - if(OP_UNLIKELY(nsamples*nchannels>_buf_size)){ - nsamples=_buf_size/nchannels; - } + if(nsamples>0){ + if(nsamples*nchannels>_buf_size)nsamples=_buf_size/nchannels; memcpy(_pcm,_of->od_buffer+nchannels*od_buffer_pos, sizeof(*_pcm)*nchannels*nsamples); od_buffer_pos+=nsamples; @@ -2520,11 +2663,11 @@ static int op_read_native(OggOpusFile *_of, /*If we have buffered packets, decode one.*/ op_pos=_of->op_pos; if(OP_LIKELY(op_pos<_of->op_count)){ - ogg_packet *pop; - ogg_int64_t diff; - opus_int32 cur_discard_count; - int duration; - int trimmed_duration; + const ogg_packet *pop; + ogg_int64_t diff; + opus_int32 cur_discard_count; + int duration; + int trimmed_duration; pop=_of->op+op_pos++; _of->op_pos=op_pos; cur_discard_count=_of->cur_discard_count; @@ -2553,15 +2696,8 @@ static int op_read_native(OggOpusFile *_of, if(OP_UNLIKELY(ret<0))return ret; buf=_of->od_buffer; } -#if defined(OP_FIXED_POINT) - ret=opus_multistream_decode(_of->od, - pop->packet,pop->bytes,buf,120*48,0); -#else - ret=opus_multistream_decode_float(_of->od, - pop->packet,pop->bytes,buf,120*48,0); -#endif - if(OP_UNLIKELY(ret<0))return OP_EBADPACKET; - OP_ASSERT(ret==duration); + ret=op_decode(_of,buf,pop,duration,nchannels); + if(OP_UNLIKELY(ret<0))return ret; /*Perform pre-skip/pre-roll.*/ od_buffer_pos=(int)OP_MIN(trimmed_duration,cur_discard_count); cur_discard_count-=od_buffer_pos; @@ -2572,31 +2708,22 @@ static int op_read_native(OggOpusFile *_of, what was decoded.*/ _of->bytes_tracked+=pop->bytes; _of->samples_tracked+=trimmed_duration-od_buffer_pos; - /*Don't grab another page yet.*/ - if(OP_LIKELY(od_buffer_posod,pop->packet,pop->bytes, - _pcm,_buf_size/nchannels,0); -#else - ret=opus_multistream_decode_float(_of->od,pop->packet,pop->bytes, - _pcm,_buf_size/nchannels,0); -#endif - if(OP_UNLIKELY(ret<0))return OP_EBADPACKET; - OP_ASSERT(ret==duration); + ret=op_decode(_of,_pcm,pop,duration,nchannels); + if(OP_UNLIKELY(ret<0))return ret; if(OP_LIKELY(trimmed_duration>0)){ /*Perform pre-skip/pre-roll.*/ od_buffer_pos=(int)OP_MIN(trimmed_duration,cur_discard_count); cur_discard_count-=od_buffer_pos; _of->cur_discard_count=cur_discard_count; - if(OP_UNLIKELY(od_buffer_pos>0) - &&OP_LIKELY(od_buffer_pos0) + &&OP_UNLIKELY(od_buffer_pos>0)){ + memmove(_pcm,_pcm+od_buffer_pos*nchannels, + sizeof(*_pcm)*trimmed_duration*nchannels); + } /*Update bitrate tracking based on the actual samples we used from what was decoded.*/ _of->bytes_tracked+=pop->bytes; @@ -2607,6 +2734,9 @@ static int op_read_native(OggOpusFile *_of, } } } + /*Don't grab another page yet. + This one might have more packets, or might have buffered data now.*/ + continue; } } /*Suck in another page.*/ @@ -2619,12 +2749,15 @@ static int op_read_native(OggOpusFile *_of, } } +/*A generic filter to apply to the decoded audio data. + _src is non-const because we will destructively modify the contents of the + source buffer that we consume in some cases.*/ typedef int (*op_read_filter_func)(OggOpusFile *_of,void *_dst,int _dst_sz, op_sample *_src,int _nsamples,int _nchannels); /*Decode some samples and then apply a custom filter to them. This is used to convert to different output formats.*/ -static int op_read_native_filter(OggOpusFile *_of,void *_dst,int _dst_sz, +static int op_filter_read_native(OggOpusFile *_of,void *_dst,int _dst_sz, op_read_filter_func _filter,int *_li){ int ret; /*Ensure we have some decoded samples in our buffer.*/ @@ -2648,11 +2781,46 @@ static int op_read_native_filter(OggOpusFile *_of,void *_dst,int _dst_sz, return ret; } -#if defined(OP_FIXED_POINT) +#if !defined(OP_FIXED_POINT)||!defined(OP_DISABLE_FLOAT_API) -int op_read(OggOpusFile *_of,opus_int16 *_pcm,int _buf_size,int *_li){ - return op_read_native(_of,_pcm,_buf_size,_li); -} +/*Matrices for downmixing from the supported channel counts to stereo. + The matrices with 5 or more channels are normalized to a total volume of 2.0, + since most mixes sound too quiet if normalized to 1.0 (as there is generally + little volume in the side/rear channels).*/ +static const float OP_STEREO_DOWNMIX[OP_NCHANNELS_MAX-2][OP_NCHANNELS_MAX][2]={ + /*3.0*/ + { + {0.5858F,0.0F},{0.4142F,0.4142F},{0.0F,0.5858F} + }, + /*quadrophonic*/ + { + {0.4226F,0.0F},{0.0F,0.4226F},{0.366F,0.2114F},{0.2114F,0.336F} + }, + /*5.0*/ + { + {0.651F,0.0F},{0.46F,0.46F},{0.0F,0.651F},{0.5636F,0.3254F}, + {0.3254F,0.5636F} + }, + /*5.1*/ + { + {0.529F,0.0F},{0.3741F,0.3741F},{0.0F,0.529F},{0.4582F,0.2645F}, + {0.2645F,0.4582F},{0.3741F,0.3741F} + }, + /*6.1*/ + { + {0.4553F,0.0F},{0.322F,0.322F},{0.0F,0.4553F},{0.3943F,0.2277F}, + {0.2277F,0.3943F},{0.2788F,0.2788F},{0.322F,0.322F} + }, + /*7.1*/ + { + {0.3886F,0.0F},{0.2748F,0.2748F},{0.0F,0.3886F},{0.3366F,0.1943F}, + {0.1943F,0.3366F},{0.3366F,0.1943F},{0.1943F,0.3366F},{0.2748F,0.2748F} + } +}; + +#endif + +#if defined(OP_FIXED_POINT) /*Matrices for downmixing from the supported channel counts to stereo. The matrices with 5 or more channels are normalized to a total volume of 2.0, @@ -2690,9 +2858,13 @@ static const opus_int16 OP_STEREO_DOWNMIX_Q14 } }; +int op_read(OggOpusFile *_of,opus_int16 *_pcm,int _buf_size,int *_li){ + return op_read_native(_of,_pcm,_buf_size,_li); +} + static int op_stereo_filter(OggOpusFile *_of,void *_dst,int _dst_sz, op_sample *_src,int _nsamples,int _nchannels){ - _of=_of; + (void)_of; _nsamples=OP_MIN(_nsamples,_dst_sz>>1); if(_nchannels==2)memcpy(_dst,_src,_nsamples*2*sizeof(*_src)); else{ @@ -2714,6 +2886,7 @@ static int op_stereo_filter(OggOpusFile *_of,void *_dst,int _dst_sz, l+=OP_STEREO_DOWNMIX_Q14[_nchannels-3][ci][0]*s; r+=OP_STEREO_DOWNMIX_Q14[_nchannels-3][ci][1]*s; } + /*TODO: For 5 or more channels, we should do soft clipping here.*/ dst[2*i+0]=(opus_int16)OP_CLAMP(-32768,l+8192>>14,32767); dst[2*i+1]=(opus_int16)OP_CLAMP(-32768,r+8192>>14,32767); } @@ -2723,7 +2896,7 @@ static int op_stereo_filter(OggOpusFile *_of,void *_dst,int _dst_sz, } int op_read_stereo(OggOpusFile *_of,opus_int16 *_pcm,int _buf_size){ - return op_read_native_filter(_of,_pcm,_buf_size,op_stereo_filter,NULL); + return op_filter_read_native(_of,_pcm,_buf_size,op_stereo_filter,NULL); } # if !defined(OP_DISABLE_FLOAT_API) @@ -2732,7 +2905,7 @@ static int op_short2float_filter(OggOpusFile *_of,void *_dst,int _dst_sz, op_sample *_src,int _nsamples,int _nchannels){ float *dst; int i; - _of=_of; + (void)_of; dst=(float *)_dst; if(OP_UNLIKELY(_nsamples*_nchannels>_dst_sz))_nsamples=_dst_sz/_nchannels; _dst_sz=_nsamples*_nchannels; @@ -2741,33 +2914,51 @@ static int op_short2float_filter(OggOpusFile *_of,void *_dst,int _dst_sz, } int op_read_float(OggOpusFile *_of,float *_pcm,int _buf_size,int *_li){ - return op_read_native_filter(_of,_pcm,_buf_size,op_short2float_filter,_li); + return op_filter_read_native(_of,_pcm,_buf_size,op_short2float_filter,_li); } static int op_short2float_stereo_filter(OggOpusFile *_of, void *_dst,int _dst_sz,op_sample *_src,int _nsamples,int _nchannels){ float *dst; + int i; dst=(float *)_dst; _nsamples=OP_MIN(_nsamples,_dst_sz>>1); if(_nchannels==1){ - int i; _nsamples=op_short2float_filter(_of,dst,_nsamples,_src,_nsamples,1); for(i=_nsamples;i-->0;)dst[2*i+0]=dst[2*i+1]=dst[i]; - return _nsamples; } - /*It would be better to convert to floats and then downmix (so that we don't - risk clipping with more than 5 channels), but that would require a large - stack buffer, which is probably not a good idea if you're using the - fixed-point build.*/ - if(_nchannels>2){ - _nsamples=op_stereo_filter(_of,_src,_nsamples*2, - _src,_nsamples,_nchannels); + else if(_nchannels<5){ + /*For 3 or 4 channels, we can downmix in fixed point without risk of + clipping.*/ + if(_nchannels>2){ + _nsamples=op_stereo_filter(_of,_src,_nsamples*2, + _src,_nsamples,_nchannels); + } + return op_short2float_filter(_of,dst,_dst_sz,_src,_nsamples,2); } - return op_short2float_filter(_of,dst,_dst_sz,_src,_nsamples,2); + else{ + /*For 5 or more channels, we convert to floats and then downmix (so that we + don't risk clipping).*/ + for(i=0;i<_nsamples;i++){ + float l; + float r; + int ci; + l=r=0; + for(ci=0;ci<_nchannels;ci++){ + float s; + s=(1.0F/32768)*_src[_nchannels*i+ci]; + l+=OP_STEREO_DOWNMIX[_nchannels-3][ci][0]*s; + r+=OP_STEREO_DOWNMIX[_nchannels-3][ci][1]*s; + } + dst[2*i+0]=l; + dst[2*i+1]=r; + } + } + return _nsamples; } int op_read_float_stereo(OggOpusFile *_of,float *_pcm,int _buf_size){ - return op_read_native_filter(_of,_pcm,_buf_size, + return op_filter_read_native(_of,_pcm,_buf_size, op_short2float_stereo_filter,NULL); } @@ -2801,7 +2992,7 @@ static opus_uint32 op_rand(opus_uint32 _seed){ suppression. This process can increase the peak level of the signal (in theory by the peak error of 1.5 +20 dB, though that is unobservably rare). - To avoid clipping, the signal is attenuated by a couple thousands of a dB. + To avoid clipping, the signal is attenuated by a couple thousandths of a dB. Initially, the approach taken here was to only attenuate by the 99.9th percentile, making clipping rare but not impossible (like SoX), but the limited gain of the filter means that the worst case was only two @@ -2809,9 +3000,9 @@ static opus_uint32 op_rand(opus_uint32 _seed){ The attenuation is probably also helpful to prevent clipping in the DAC reconstruction filters or downstream resampling, in any case.*/ -#define OP_GAIN (32753.0F) +# define OP_GAIN (32753.0F) -#define OP_PRNG_GAIN (1.0F/0xFFFFFFFF) +# define OP_PRNG_GAIN (1.0F/0xFFFFFFFF) /*48 kHz noise shaping filter, sd=2.34.*/ @@ -2823,118 +3014,95 @@ static const float OP_FCOEF_A[4]={ 0.9030F,0.0116F,-0.5853F,-0.2571F }; -static void op_shaped_dither16(OggOpusFile *_of,opus_int16 *_dst, - const float *_src,int _nsamples,int _nchannels){ - opus_uint32 seed; - int mute; - int i; - mute=_of->dither_mute; - seed=_of->dither_seed; - /*In order to avoid replacing digital silence with quiet dither noise, we - mute if the output has been silent for a while.*/ - if(mute>64)memset(_of->dither_a,0,sizeof(*_of->dither_a)*4*_nchannels); - for(i=0;i<_nsamples;i++){ - int silent; - int ci; - silent=1; - for(ci=0;ci<_nchannels;ci++){ - float r; - float s; - float err; - int si; - int j; - s=_src[_nchannels*i+ci]; - silent&=s==0; - s*=OP_GAIN; - err=0; - for(j=0;j<4;j++){ - err+=OP_FCOEF_B[j]*_of->dither_b[ci*4+j] - -OP_FCOEF_A[j]*_of->dither_a[ci*4+j]; - } - for(j=3;j-->0;)_of->dither_a[ci*4+j+1]=_of->dither_a[ci*4+j]; - for(j=3;j-->0;)_of->dither_b[ci*4+j+1]=_of->dither_b[ci*4+j]; - _of->dither_a[ci*4]=err; - s-=err; - if(mute>16)r=0; - else{ - seed=op_rand(seed); - r=seed*OP_PRNG_GAIN; - seed=op_rand(seed); - r-=seed*OP_PRNG_GAIN; - } - /*Clamp in float out of paranoia that the input will be > 96 dBFS and - wrap if the integer is clamped.*/ - si=op_float2int(OP_CLAMP(-32768,s+r,32767)); - _dst[_nchannels*i+ci]=(opus_int16)si; - /*Including clipping in the noise shaping is generally disastrous: the - futile effort to restore the clipped energy results in more clipping. - However, small amounts---at the level which could normally be created - by dither and rounding---are harmless and can even reduce clipping - somewhat due to the clipping sometimes reducing the dither + rounding - error.*/ - _of->dither_b[ci*4]=mute>16?0:OP_CLAMP(-1.5F,si-s,1.5F); - } - mute++; - if(!silent)mute=0; - } - _of->dither_mute=OP_MIN(mute,65); - _of->dither_seed=seed; -} - static int op_float2short_filter(OggOpusFile *_of,void *_dst,int _dst_sz, - op_sample *_src,int _nsamples,int _nchannels){ + float *_src,int _nsamples,int _nchannels){ opus_int16 *dst; + int ci; + int i; dst=(opus_int16 *)_dst; if(OP_UNLIKELY(_nsamples*_nchannels>_dst_sz))_nsamples=_dst_sz/_nchannels; - op_shaped_dither16(_of,dst,_src,_nsamples,_nchannels); +# if defined(OP_SOFT_CLIP) + if(_of->state_channel_count!=_nchannels){ + for(ci=0;ci<_nchannels;ci++)_of->clip_state[ci]=0; + } + opus_pcm_soft_clip(_src,_nsamples,_nchannels,_of->clip_state); +# endif + if(_of->dither_disabled){ + for(i=0;i<_nchannels*_nsamples;i++){ + dst[i]=op_float2int(OP_CLAMP(-32768,32768.0F*_src[i],32767)); + } + } + else{ + opus_uint32 seed; + int mute; + seed=_of->dither_seed; + mute=_of->dither_mute; + if(_of->state_channel_count!=_nchannels)mute=65; + /*In order to avoid replacing digital silence with quiet dither noise, we + mute if the output has been silent for a while.*/ + if(mute>64)memset(_of->dither_a,0,sizeof(*_of->dither_a)*4*_nchannels); + for(i=0;i<_nsamples;i++){ + int silent; + silent=1; + for(ci=0;ci<_nchannels;ci++){ + float r; + float s; + float err; + int si; + int j; + s=_src[_nchannels*i+ci]; + silent&=s==0; + s*=OP_GAIN; + err=0; + for(j=0;j<4;j++){ + err+=OP_FCOEF_B[j]*_of->dither_b[ci*4+j] + -OP_FCOEF_A[j]*_of->dither_a[ci*4+j]; + } + for(j=3;j-->0;)_of->dither_a[ci*4+j+1]=_of->dither_a[ci*4+j]; + for(j=3;j-->0;)_of->dither_b[ci*4+j+1]=_of->dither_b[ci*4+j]; + _of->dither_a[ci*4]=err; + s-=err; + if(mute>16)r=0; + else{ + seed=op_rand(seed); + r=seed*OP_PRNG_GAIN; + seed=op_rand(seed); + r-=seed*OP_PRNG_GAIN; + } + /*Clamp in float out of paranoia that the input will be > 96 dBFS and + wrap if the integer is clamped.*/ + si=op_float2int(OP_CLAMP(-32768,s+r,32767)); + dst[_nchannels*i+ci]=(opus_int16)si; + /*Including clipping in the noise shaping is generally disastrous: the + futile effort to restore the clipped energy results in more clipping. + However, small amounts---at the level which could normally be created + by dither and rounding---are harmless and can even reduce clipping + somewhat due to the clipping sometimes reducing the dither + rounding + error.*/ + _of->dither_b[ci*4]=mute>16?0:OP_CLAMP(-1.5F,si-s,1.5F); + } + mute++; + if(!silent)mute=0; + } + _of->dither_mute=OP_MIN(mute,65); + _of->dither_seed=seed; + } + _of->state_channel_count=_nchannels; return _nsamples; } int op_read(OggOpusFile *_of,opus_int16 *_pcm,int _buf_size,int *_li){ - return op_read_native_filter(_of,_pcm,_buf_size,op_float2short_filter,_li); + return op_filter_read_native(_of,_pcm,_buf_size,op_float2short_filter,_li); } int op_read_float(OggOpusFile *_of,float *_pcm,int _buf_size,int *_li){ + _of->state_channel_count=0; return op_read_native(_of,_pcm,_buf_size,_li); } -/*Matrices for downmixing from the supported channel counts to stereo. - The matrices with 5 or more channels are normalized to a total volume of 2.0, - since most mixes sound too quiet if normalized to 1.0 (as there is generally - little volume in the side/rear channels).*/ -static const float OP_STEREO_DOWNMIX[OP_NCHANNELS_MAX-2][OP_NCHANNELS_MAX][2]={ - /*3.0*/ - { - {0.5858F,0.0F},{0.4142F,0.4142F},{0.0F,0.5858F} - }, - /*quadrophonic*/ - { - {0.4226F,0.0F},{0.0F,0.4226F},{0.366F,0.2114F},{0.2114F,0.336F} - }, - /*5.0*/ - { - {0.651F,0.0F},{0.46F,0.46F},{0.0F,0.651F},{0.5636F,0.3254F}, - {0.3254F,0.5636F} - }, - /*5.1*/ - { - {0.529F,0.0F},{0.3741F,0.3741F},{0.0F,0.529F},{0.4582F,0.2645F}, - {0.2645F,0.4582F},{0.3741F,0.3741F} - }, - /*6.1*/ - { - {0.4553F,0.0F},{0.322F,0.322F},{0.0F,0.4553F},{0.3943F,0.2277F}, - {0.2277F,0.3943F},{0.2788F,0.2788F},{0.322F,0.322F} - }, - /*7.1*/ - { - {0.3886F,0.0F},{0.2748F,0.2748F},{0.0F,0.3886F},{0.3366F,0.1943F}, - {0.1943F,0.3366F},{0.3366F,0.1943F},{0.1943F,0.3366F},{0.2748F,0.2748F} - } -}; - static int op_stereo_filter(OggOpusFile *_of,void *_dst,int _dst_sz, op_sample *_src,int _nsamples,int _nchannels){ + (void)_of; _nsamples=OP_MIN(_nsamples,_dst_sz>>1); if(_nchannels==2)memcpy(_dst,_src,_nsamples*2*sizeof(*_src)); else{ @@ -2966,29 +3134,30 @@ static int op_float2short_stereo_filter(OggOpusFile *_of, void *_dst,int _dst_sz,op_sample *_src,int _nsamples,int _nchannels){ opus_int16 *dst; dst=(opus_int16 *)_dst; - _nsamples=OP_MIN(_nsamples,_dst_sz>>1); if(_nchannels==1){ int i; - op_shaped_dither16(_of,dst,_src,_nsamples,1); + _nsamples=op_float2short_filter(_of,dst,_dst_sz>>1,_src,_nsamples,1); for(i=_nsamples;i-->0;)dst[2*i+0]=dst[2*i+1]=dst[i]; } else{ if(_nchannels>2){ + _nsamples=OP_MIN(_nsamples,_dst_sz>>1); _nsamples=op_stereo_filter(_of,_src,_nsamples*2, _src,_nsamples,_nchannels); } - op_shaped_dither16(_of,dst,_src,_nsamples,_nchannels); + _nsamples=op_float2short_filter(_of,dst,_dst_sz,_src,_nsamples,2); } return _nsamples; } int op_read_stereo(OggOpusFile *_of,opus_int16 *_pcm,int _buf_size){ - return op_read_native_filter(_of,_pcm,_buf_size, + return op_filter_read_native(_of,_pcm,_buf_size, op_float2short_stereo_filter,NULL); } int op_read_float_stereo(OggOpusFile *_of,float *_pcm,int _buf_size){ - return op_read_native_filter(_of,_pcm,_buf_size,op_stereo_filter,NULL); + _of->state_channel_count=0; + return op_filter_read_native(_of,_pcm,_buf_size,op_stereo_filter,NULL); } #endif diff --git a/code/opusfile-0.5/src/stream.c b/code/opusfile-0.5/src/stream.c new file mode 100644 index 00000000..0238a6b3 --- /dev/null +++ b/code/opusfile-0.5/src/stream.c @@ -0,0 +1,366 @@ +/******************************************************************** + * * + * THIS FILE IS PART OF THE libopusfile SOFTWARE CODEC SOURCE CODE. * + * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS * + * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE * + * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. * + * * + * THE libopusfile SOURCE CODE IS (C) COPYRIGHT 1994-2012 * + * by the Xiph.Org Foundation and contributors http://www.xiph.org/ * + * * + ******************************************************************** + + function: stdio-based convenience library for opening/seeking/decoding + last mod: $Id: vorbisfile.c 17573 2010-10-27 14:53:59Z xiphmont $ + + ********************************************************************/ +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "internal.h" +#include +#include +#include +#include +#include +#if defined(_WIN32) +# include +#endif + +typedef struct OpusMemStream OpusMemStream; + +#define OP_MEM_SIZE_MAX (~(size_t)0>>1) +#define OP_MEM_DIFF_MAX ((ptrdiff_t)OP_MEM_SIZE_MAX) + +/*The context information needed to read from a block of memory as if it were a + file.*/ +struct OpusMemStream{ + /*The block of memory to read from.*/ + const unsigned char *data; + /*The total size of the block. + This must be at most OP_MEM_SIZE_MAX to prevent signed overflow while + seeking.*/ + ptrdiff_t size; + /*The current file position. + This is allowed to be set arbitrarily greater than size (i.e., past the end + of the block, though we will not read data past the end of the block), but + is not allowed to be negative (i.e., before the beginning of the block).*/ + ptrdiff_t pos; +}; + +static int op_fread(void *_stream,unsigned char *_ptr,int _buf_size){ + FILE *stream; + size_t ret; + /*Check for empty read.*/ + if(_buf_size<=0)return 0; + stream=(FILE *)_stream; + ret=fread(_ptr,1,_buf_size,stream); + OP_ASSERT(ret<=(size_t)_buf_size); + /*If ret==0 and !feof(stream), there was a read error.*/ + return ret>0||feof(stream)?(int)ret:OP_EREAD; +} + +static int op_fseek(void *_stream,opus_int64 _offset,int _whence){ +#if defined(_WIN32) + /*_fseeki64() is not exposed until MSCVCRT80. + This is the default starting with MSVC 2005 (_MSC_VER>=1400), but we want + to allow linking against older MSVCRT versions for compatibility back to + XP without installing extra runtime libraries. + i686-pc-mingw32 does not have fseeko() and requires + __MSVCRT_VERSION__>=0x800 for _fseeki64(), which screws up linking with + other libraries (that don't use MSVCRT80 from MSVC 2005 by default). + i686-w64-mingw32 does have fseeko() and respects _FILE_OFFSET_BITS, but I + don't know how to detect that at compile time. + We could just use fseeko64() (which is available in both), but its + implemented using fgetpos()/fsetpos() just like this code, except without + the overflow checking, so we prefer our version.*/ + opus_int64 pos; + /*We don't use fpos_t directly because it might be a struct if __STDC__ is + non-zero or _INTEGRAL_MAX_BITS < 64. + I'm not certain when the latter is true, but someone could in theory set + the former. + Either way, it should be binary compatible with a normal 64-bit int (this + assumption is not portable, but I believe it is true for MSVCRT).*/ + OP_ASSERT(sizeof(pos)==sizeof(fpos_t)); + /*Translate the seek to an absolute one.*/ + if(_whence==SEEK_CUR){ + int ret; + ret=fgetpos((FILE *)_stream,(fpos_t *)&pos); + if(ret)return ret; + } + else if(_whence==SEEK_END)pos=_filelengthi64(_fileno((FILE *)_stream)); + else if(_whence==SEEK_SET)pos=0; + else return -1; + /*Check for errors or overflow.*/ + if(pos<0||_offset<-pos||_offset>OP_INT64_MAX-pos)return -1; + pos+=_offset; + return fsetpos((FILE *)_stream,(fpos_t *)&pos); +#else + /*This function actually conforms to the SUSv2 and POSIX.1-2001, so we prefer + it except on Windows.*/ + return fseeko((FILE *)_stream,(off_t)_offset,_whence); +#endif +} + +static opus_int64 op_ftell(void *_stream){ +#if defined(_WIN32) + /*_ftelli64() is not exposed until MSCVCRT80, and ftello()/ftello64() have + the same problems as fseeko()/fseeko64() in MingW. + See above for a more detailed explanation.*/ + opus_int64 pos; + OP_ASSERT(sizeof(pos)==sizeof(fpos_t)); + return fgetpos((FILE *)_stream,(fpos_t *)&pos)?-1:pos; +#else + /*This function actually conforms to the SUSv2 and POSIX.1-2001, so we prefer + it except on Windows.*/ + return ftello((FILE *)_stream); +#endif +} + +static const OpusFileCallbacks OP_FILE_CALLBACKS={ + op_fread, + op_fseek, + op_ftell, + (op_close_func)fclose +}; + +#if defined(_WIN32) +# include +# include + +/*Windows doesn't accept UTF-8 by default, and we don't have a wchar_t API, + so if we just pass the path to fopen(), then there'd be no way for a user + of our API to open a Unicode filename. + Instead, we translate from UTF-8 to UTF-16 and use Windows' wchar_t API. + This makes this API more consistent with platforms where the character set + used by fopen is the same as used on disk, which is generally UTF-8, and + with our metadata API, which always uses UTF-8.*/ +static wchar_t *op_utf8_to_utf16(const char *_src){ + wchar_t *dst; + size_t len; + len=strlen(_src); + /*Worst-case output is 1 wide character per 1 input character.*/ + dst=(wchar_t *)_ogg_malloc(sizeof(*dst)*(len+1)); + if(dst!=NULL){ + size_t si; + size_t di; + for(di=si=0;si=0x80U){ + /*This is a 2-byte sequence that is not overlong.*/ + dst[di++]=w; + si++; + continue; + } + } + else{ + int c2; + /*This is safe, because c1 was not 0 and _src is NUL-terminated.*/ + c2=(unsigned char)_src[si+2]; + if((c2&0xC0)==0x80){ + /*Found at least two continuation bytes.*/ + if((c0&0xF0)==0xE0){ + wchar_t w; + /*Start byte says this is a 3-byte sequence.*/ + w=(c0&0xF)<<12|(c1&0x3F)<<6|c2&0x3F; + if(w>=0x800U&&(w<0xD800||w>=0xE000)&&w<0xFFFE){ + /*This is a 3-byte sequence that is not overlong, not a + UTF-16 surrogate pair value, and not a 'not a character' + value.*/ + dst[di++]=w; + si+=2; + continue; + } + } + else{ + int c3; + /*This is safe, because c2 was not 0 and _src is + NUL-terminated.*/ + c3=(unsigned char)_src[si+3]; + if((c3&0xC0)==0x80){ + /*Found at least three continuation bytes.*/ + if((c0&0xF8)==0xF0){ + opus_uint32 w; + /*Start byte says this is a 4-byte sequence.*/ + w=(c0&7)<<18|(c1&0x3F)<<12|(c2&0x3F)<<6&(c3&0x3F); + if(w>=0x10000U&&w<0x110000U){ + /*This is a 4-byte sequence that is not overlong and not + greater than the largest valid Unicode code point. + Convert it to a surrogate pair.*/ + w-=0x10000; + dst[di++]=(wchar_t)(0xD800+(w>>10)); + dst[di++]=(wchar_t)(0xDC00+(w&0x3FF)); + si+=3; + continue; + } + } + } + } + } + } + } + } + /*If we got here, we encountered an illegal UTF-8 sequence.*/ + _ogg_free(dst); + return NULL; + } + OP_ASSERT(di<=len); + dst[di]='\0'; + } + return dst; +} + +#endif + +void *op_fopen(OpusFileCallbacks *_cb,const char *_path,const char *_mode){ + FILE *fp; +#if !defined(_WIN32) + fp=fopen(_path,_mode); +#else + fp=NULL; + if(_path==NULL||_mode==NULL)errno=EINVAL; + else{ + wchar_t *wpath; + wchar_t *wmode; + wpath=op_utf8_to_utf16(_path); + wmode=op_utf8_to_utf16(_mode); + if(wmode==NULL)errno=EINVAL; + else if(wpath==NULL)errno=ENOENT; + else fp=_wfopen(wpath,wmode); + _ogg_free(wmode); + _ogg_free(wpath); + } +#endif + if(fp!=NULL)*_cb=*&OP_FILE_CALLBACKS; + return fp; +} + +void *op_fdopen(OpusFileCallbacks *_cb,int _fd,const char *_mode){ + FILE *fp; + fp=fdopen(_fd,_mode); + if(fp!=NULL)*_cb=*&OP_FILE_CALLBACKS; + return fp; +} + +void *op_freopen(OpusFileCallbacks *_cb,const char *_path,const char *_mode, + void *_stream){ + FILE *fp; +#if !defined(_WIN32) + fp=freopen(_path,_mode,(FILE *)_stream); +#else + fp=NULL; + if(_path==NULL||_mode==NULL)errno=EINVAL; + else{ + wchar_t *wpath; + wchar_t *wmode; + wpath=op_utf8_to_utf16(_path); + wmode=op_utf8_to_utf16(_mode); + if(wmode==NULL)errno=EINVAL; + else if(wpath==NULL)errno=ENOENT; + else fp=_wfreopen(wpath,wmode,(FILE *)_stream); + _ogg_free(wmode); + _ogg_free(wpath); + } +#endif + if(fp!=NULL)*_cb=*&OP_FILE_CALLBACKS; + return fp; +} + +static int op_mem_read(void *_stream,unsigned char *_ptr,int _buf_size){ + OpusMemStream *stream; + ptrdiff_t size; + ptrdiff_t pos; + stream=(OpusMemStream *)_stream; + /*Check for empty read.*/ + if(_buf_size<=0)return 0; + size=stream->size; + pos=stream->pos; + /*Check for EOF.*/ + if(pos>=size)return 0; + /*Check for a short read.*/ + _buf_size=(int)OP_MIN(size-pos,_buf_size); + memcpy(_ptr,stream->data+pos,_buf_size); + pos+=_buf_size; + stream->pos=pos; + return _buf_size; +} + +static int op_mem_seek(void *_stream,opus_int64 _offset,int _whence){ + OpusMemStream *stream; + ptrdiff_t pos; + stream=(OpusMemStream *)_stream; + pos=stream->pos; + OP_ASSERT(pos>=0); + switch(_whence){ + case SEEK_SET:{ + /*Check for overflow:*/ + if(_offset<0||_offset>OP_MEM_DIFF_MAX)return -1; + pos=(ptrdiff_t)_offset; + }break; + case SEEK_CUR:{ + /*Check for overflow:*/ + if(_offset<-pos||_offset>OP_MEM_DIFF_MAX-pos)return -1; + pos=(ptrdiff_t)(pos+_offset); + }break; + case SEEK_END:{ + ptrdiff_t size; + size=stream->size; + OP_ASSERT(size>=0); + /*Check for overflow:*/ + if(_offset>size||_offsetpos=pos; + return 0; +} + +static opus_int64 op_mem_tell(void *_stream){ + OpusMemStream *stream; + stream=(OpusMemStream *)_stream; + return (ogg_int64_t)stream->pos; +} + +static int op_mem_close(void *_stream){ + _ogg_free(_stream); + return 0; +} + +static const OpusFileCallbacks OP_MEM_CALLBACKS={ + op_mem_read, + op_mem_seek, + op_mem_tell, + op_mem_close +}; + +void *op_mem_stream_create(OpusFileCallbacks *_cb, + const unsigned char *_data,size_t _size){ + OpusMemStream *stream; + if(_size>OP_MEM_SIZE_MAX)return NULL; + stream=(OpusMemStream *)_ogg_malloc(sizeof(*stream)); + if(stream!=NULL){ + *_cb=*&OP_MEM_CALLBACKS; + stream->data=_data; + stream->size=_size; + stream->pos=0; + } + return stream; +} diff --git a/code/opusfile-0.5/src/wincerts.c b/code/opusfile-0.5/src/wincerts.c new file mode 100644 index 00000000..b0e35aa3 --- /dev/null +++ b/code/opusfile-0.5/src/wincerts.c @@ -0,0 +1,171 @@ +/******************************************************************** + * * + * THIS FILE IS PART OF THE libopusfile SOFTWARE CODEC SOURCE CODE. * + * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS * + * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE * + * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. * + * * + * THE libopusfile SOURCE CODE IS (C) COPYRIGHT 2013 * + * by the Xiph.Org Foundation and contributors http://www.xiph.org/ * + * * + ********************************************************************/ + +/*This should really be part of OpenSSL, but there's been a patch [1] sitting + in their bugtracker for over two years that implements this, without any + action, so I'm giving up and re-implementing it locally. + + [1] */ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "internal.h" +#if defined(OP_ENABLE_HTTP)&&defined(_WIN32) +/*You must include windows.h before wincrypt.h and x509.h.*/ +# define WIN32_LEAN_AND_MEAN +# define WIN32_EXTRA_LEAN +# include +/*You must include wincrypt.h before x509.h, too, or X509_NAME doesn't get + defined properly.*/ +# include +# include +# include +# include + +static int op_capi_new(X509_LOOKUP *_lu){ + HCERTSTORE h_store; + h_store=CertOpenStore(CERT_STORE_PROV_SYSTEM_A,0,0, + CERT_STORE_OPEN_EXISTING_FLAG|CERT_STORE_READONLY_FLAG| + CERT_SYSTEM_STORE_CURRENT_USER|CERT_STORE_SHARE_CONTEXT_FLAG,"ROOT"); + if(h_store!=NULL){ + _lu->method_data=(char *)h_store; + return 1; + } + return 0; +} + +static void op_capi_free(X509_LOOKUP *_lu){ + HCERTSTORE h_store; + h_store=(HCERTSTORE)_lu->method_data; +# if defined(OP_ENABLE_ASSERTIONS) + OP_ALWAYS_TRUE(CertCloseStore(h_store,CERT_CLOSE_STORE_CHECK_FLAG)); +# else + CertCloseStore(h_store,0); +# endif +} + +static int op_capi_retrieve_by_subject(X509_LOOKUP *_lu,int _type, + X509_NAME *_name,X509_OBJECT *_ret){ + X509_OBJECT *obj; + CRYPTO_w_lock(CRYPTO_LOCK_X509_STORE); + obj=X509_OBJECT_retrieve_by_subject(_lu->store_ctx->objs,_type,_name); + CRYPTO_w_unlock(CRYPTO_LOCK_X509_STORE); + if(obj!=NULL){ + _ret->type=obj->type; + memcpy(&_ret->data,&obj->data,sizeof(_ret->data)); + return 1; + } + return 0; +} + +static int op_capi_get_by_subject(X509_LOOKUP *_lu,int _type,X509_NAME *_name, + X509_OBJECT *_ret){ + HCERTSTORE h_store; + if(_name==NULL)return 0; + if(_name->bytes==NULL||_name->bytes->length<=0||_name->modified){ + if(i2d_X509_NAME(_name,NULL)<0)return 0; + OP_ASSERT(_name->bytes->length>0); + } + h_store=(HCERTSTORE)_lu->method_data; + switch(_type){ + case X509_LU_X509:{ + CERT_NAME_BLOB find_para; + PCCERT_CONTEXT cert; + X509 *x; + int ret; + /*Although X509_NAME contains a canon_enc field, that "canonical" [1] + encoding was just made up by OpenSSL. + It doesn't correspond to any actual standard, and since it drops the + initial sequence header, won't be recognized by the Crypto API. + The assumption here is that CertFindCertificateInStore() will allow any + appropriate variations in the encoding when it does its comparison. + This is, however, emphatically not true under Wine, which just compares + the encodings with memcmp(). + Most of the time things work anyway, though, and there isn't really + anything we can do to make the situation better. + + [1] A "canonical form" is defined as the one where, if you locked 10 + mathematicians in a room and asked them to come up with a + representation for something, it's the answer that 9 of them would + give you back. + I don't think OpenSSL's encoding qualifies.*/ + find_para.cbData=_name->bytes->length; + find_para.pbData=(unsigned char *)_name->bytes->data; + cert=CertFindCertificateInStore(h_store,X509_ASN_ENCODING,0, + CERT_FIND_SUBJECT_NAME,&find_para,NULL); + if(cert==NULL)return 0; + x=d2i_X509(NULL,(const unsigned char **)&cert->pbCertEncoded, + cert->cbCertEncoded); + CertFreeCertificateContext(cert); + if(x==NULL)return 0; + ret=X509_STORE_add_cert(_lu->store_ctx,x); + X509_free(x); + if(ret)return op_capi_retrieve_by_subject(_lu,_type,_name,_ret); + }break; + case X509_LU_CRL:{ + CERT_INFO cert_info; + CERT_CONTEXT find_para; + PCCRL_CONTEXT crl; + X509_CRL *x; + int ret; + ret=op_capi_retrieve_by_subject(_lu,_type,_name,_ret); + if(ret>0)return ret; + memset(&cert_info,0,sizeof(cert_info)); + cert_info.Issuer.cbData=_name->bytes->length; + cert_info.Issuer.pbData=(unsigned char *)_name->bytes->data; + memset(&find_para,0,sizeof(find_para)); + find_para.pCertInfo=&cert_info; + crl=CertFindCRLInStore(h_store,0,0,CRL_FIND_ISSUED_BY,&find_para,NULL); + if(crl==NULL)return 0; + x=d2i_X509_CRL(NULL,(const unsigned char **)&crl->pbCrlEncoded, + crl->cbCrlEncoded); + CertFreeCRLContext(crl); + if(x==NULL)return 0; + ret=X509_STORE_add_crl(_lu->store_ctx,x); + X509_CRL_free(x); + if(ret)return op_capi_retrieve_by_subject(_lu,_type,_name,_ret); + }break; + } + return 0; +} + +/*This is not const because OpenSSL doesn't allow it, even though it won't + write to it.*/ +static X509_LOOKUP_METHOD X509_LOOKUP_CAPI={ + "Load Crypto API store into cache", + op_capi_new, + op_capi_free, + NULL, + NULL, + NULL, + op_capi_get_by_subject, + NULL, + NULL, + NULL +}; + +int SSL_CTX_set_default_verify_paths_win32(SSL_CTX *_ssl_ctx){ + X509_STORE *store; + X509_LOOKUP *lu; + /*We intentionally do not add the normal default paths, as they are usually + wrong, and are just asking to be used as an exploit vector.*/ + store=SSL_CTX_get_cert_store(_ssl_ctx); + OP_ASSERT(store!=NULL); + lu=X509_STORE_add_lookup(store,&X509_LOOKUP_CAPI); + if(lu==NULL)return 0; + ERR_clear_error(); + return 1; +} + +#endif diff --git a/code/opusfile-0.5/src/winerrno.h b/code/opusfile-0.5/src/winerrno.h new file mode 100644 index 00000000..32a90b4e --- /dev/null +++ b/code/opusfile-0.5/src/winerrno.h @@ -0,0 +1,90 @@ +/******************************************************************** + * * + * THIS FILE IS PART OF THE libopusfile SOFTWARE CODEC SOURCE CODE. * + * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS * + * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE * + * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. * + * * + * THE libopusfile SOURCE CODE IS (C) COPYRIGHT 2012 * + * by the Xiph.Org Foundation and contributors http://www.xiph.org/ * + * * + ********************************************************************/ +#if !defined(_opusfile_winerrno_h) +# define _opusfile_winerrno_h (1) + +# include +# include + +/*These conflict with the MSVC errno.h definitions, but we don't need to use + the original ones in any file that deals with sockets. + We could map the WSA errors to the errno.h ones (most of which are only + available on sufficiently new versions of MSVC), but they aren't ordered the + same, and given how rarely we actually look at the values, I don't think + it's worth a lookup table.*/ +# undef EWOULDBLOCK +# undef EINPROGRESS +# undef EALREADY +# undef ENOTSOCK +# undef EDESTADDRREQ +# undef EMSGSIZE +# undef EPROTOTYPE +# undef ENOPROTOOPT +# undef EPROTONOSUPPORT +# undef EOPNOTSUPP +# undef EAFNOSUPPORT +# undef EADDRINUSE +# undef EADDRNOTAVAIL +# undef ENETDOWN +# undef ENETUNREACH +# undef ENETRESET +# undef ECONNABORTED +# undef ECONNRESET +# undef ENOBUFS +# undef EISCONN +# undef ENOTCONN +# undef ETIMEDOUT +# undef ECONNREFUSED +# undef ELOOP +# undef ENAMETOOLONG +# undef EHOSTUNREACH +# undef ENOTEMPTY + +# define EWOULDBLOCK (WSAEWOULDBLOCK-WSABASEERR) +# define EINPROGRESS (WSAEINPROGRESS-WSABASEERR) +# define EALREADY (WSAEALREADY-WSABASEERR) +# define ENOTSOCK (WSAENOTSOCK-WSABASEERR) +# define EDESTADDRREQ (WSAEDESTADDRREQ-WSABASEERR) +# define EMSGSIZE (WSAEMSGSIZE-WSABASEERR) +# define EPROTOTYPE (WSAEPROTOTYPE-WSABASEERR) +# define ENOPROTOOPT (WSAENOPROTOOPT-WSABASEERR) +# define EPROTONOSUPPORT (WSAEPROTONOSUPPORT-WSABASEERR) +# define ESOCKTNOSUPPORT (WSAESOCKTNOSUPPORT-WSABASEERR) +# define EOPNOTSUPP (WSAEOPNOTSUPP-WSABASEERR) +# define EPFNOSUPPORT (WSAEPFNOSUPPORT-WSABASEERR) +# define EAFNOSUPPORT (WSAEAFNOSUPPORT-WSABASEERR) +# define EADDRINUSE (WSAEADDRINUSE-WSABASEERR) +# define EADDRNOTAVAIL (WSAEADDRNOTAVAIL-WSABASEERR) +# define ENETDOWN (WSAENETDOWN-WSABASEERR) +# define ENETUNREACH (WSAENETUNREACH-WSABASEERR) +# define ENETRESET (WSAENETRESET-WSABASEERR) +# define ECONNABORTED (WSAECONNABORTED-WSABASEERR) +# define ECONNRESET (WSAECONNRESET-WSABASEERR) +# define ENOBUFS (WSAENOBUFS-WSABASEERR) +# define EISCONN (WSAEISCONN-WSABASEERR) +# define ENOTCONN (WSAENOTCONN-WSABASEERR) +# define ESHUTDOWN (WSAESHUTDOWN-WSABASEERR) +# define ETOOMANYREFS (WSAETOOMANYREFS-WSABASEERR) +# define ETIMEDOUT (WSAETIMEDOUT-WSABASEERR) +# define ECONNREFUSED (WSAECONNREFUSED-WSABASEERR) +# define ELOOP (WSAELOOP-WSABASEERR) +# define ENAMETOOLONG (WSAENAMETOOLONG-WSABASEERR) +# define EHOSTDOWN (WSAEHOSTDOWN-WSABASEERR) +# define EHOSTUNREACH (WSAEHOSTUNREACH-WSABASEERR) +# define ENOTEMPTY (WSAENOTEMPTY-WSABASEERR) +# define EPROCLIM (WSAEPROCLIM-WSABASEERR) +# define EUSERS (WSAEUSERS-WSABASEERR) +# define EDQUOT (WSAEDQUOT-WSABASEERR) +# define ESTALE (WSAESTALE-WSABASEERR) +# define EREMOTE (WSAEREMOTE-WSABASEERR) + +#endif diff --git a/code/qcommon/qfiles.h b/code/qcommon/qfiles.h index 78b06da1..9f2b5fb2 100644 --- a/code/qcommon/qfiles.h +++ b/code/qcommon/qfiles.h @@ -178,94 +178,11 @@ typedef struct { /* ============================================================================== -MD4 file format +MDR file format ============================================================================== */ -#define MD4_IDENT (('4'<<24)+('P'<<16)+('D'<<8)+'I') -#define MD4_VERSION 1 -#define MD4_MAX_BONES 128 - -typedef struct { - int boneIndex; // these are indexes into the boneReferences, - float boneWeight; // not the global per-frame bone list - vec3_t offset; -} md4Weight_t; - -typedef struct { - vec3_t normal; - vec2_t texCoords; - int numWeights; - md4Weight_t weights[1]; // variable sized -} md4Vertex_t; - -typedef struct { - int indexes[3]; -} md4Triangle_t; - -typedef struct { - int ident; - - char name[MAX_QPATH]; // polyset name - char shader[MAX_QPATH]; - int shaderIndex; // for in-game use - - int ofsHeader; // this will be a negative number - - int numVerts; - int ofsVerts; - - int numTriangles; - int ofsTriangles; - - // Bone references are a set of ints representing all the bones - // present in any vertex weights for this surface. This is - // needed because a model may have surfaces that need to be - // drawn at different sort times, and we don't want to have - // to re-interpolate all the bones for each surface. - int numBoneReferences; - int ofsBoneReferences; - - int ofsEnd; // next surface follows -} md4Surface_t; - -typedef struct { - float matrix[3][4]; -} md4Bone_t; - -typedef struct { - vec3_t bounds[2]; // bounds of all surfaces of all LOD's for this frame - vec3_t localOrigin; // midpoint of bounds, used for sphere cull - float radius; // dist from localOrigin to corner - md4Bone_t bones[1]; // [numBones] -} md4Frame_t; - -typedef struct { - int numSurfaces; - int ofsSurfaces; // first surface, others follow - int ofsEnd; // next lod follows -} md4LOD_t; - -typedef struct { - int ident; - int version; - - char name[MAX_QPATH]; // model name - - // frames and bones are shared by all levels of detail - int numFrames; - int numBones; - int ofsBoneNames; // char name[ MAX_QPATH ] - int ofsFrames; // md4Frame_t[numFrames] - - // each level of detail has completely separate sets of surfaces - int numLODs; - int ofsLODs; - - int ofsEnd; // end of file -} md4Header_t; - /* * Here are the definitions for Ravensoft's model format of md4. Raven stores their * playermodels in .mdr files, in some games, which are pretty much like the md4 diff --git a/code/renderercommon/tr_image_jpg.c b/code/renderercommon/tr_image_jpg.c index bd81c048..3bf44bbb 100644 --- a/code/renderercommon/tr_image_jpg.c +++ b/code/renderercommon/tr_image_jpg.c @@ -37,8 +37,8 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA #include #ifndef USE_INTERNAL_JPEG -# if JPEG_LIB_VERSION < 80 -# error Need system libjpeg >= 80 +# if JPEG_LIB_VERSION < 80 && !defined(MEM_SRCDST_SUPPORTED) +# error Need system libjpeg >= 80 or jpeg_mem_ support # endif #endif diff --git a/code/renderercommon/tr_image_png.c b/code/renderercommon/tr_image_png.c index 532454c5..7b6fbada 100644 --- a/code/renderercommon/tr_image_png.c +++ b/code/renderercommon/tr_image_png.c @@ -2274,7 +2274,7 @@ void R_LoadPNG(const char *name, byte **pic, int *width, int *height) { case PNG_ColourType_Grey : { - if(!ChunkHeaderLength == 2) + if(ChunkHeaderLength != 2) { CloseBufferedFile(ThePNG); @@ -2296,7 +2296,7 @@ void R_LoadPNG(const char *name, byte **pic, int *width, int *height) case PNG_ColourType_True : { - if(!ChunkHeaderLength == 6) + if(ChunkHeaderLength != 6) { CloseBufferedFile(ThePNG); diff --git a/code/renderergl1/tr_animation.c b/code/renderergl1/tr_animation.c index ad54dd63..7bcc5bdf 100644 --- a/code/renderergl1/tr_animation.c +++ b/code/renderergl1/tr_animation.c @@ -33,140 +33,6 @@ frame. */ -/* -============== -R_AddAnimSurfaces -============== -*/ -void R_AddAnimSurfaces( trRefEntity_t *ent ) { - md4Header_t *header; - md4Surface_t *surface; - md4LOD_t *lod; - shader_t *shader; - int i; - - header = (md4Header_t *) tr.currentModel->modelData; - lod = (md4LOD_t *)( (byte *)header + header->ofsLODs ); - - surface = (md4Surface_t *)( (byte *)lod + lod->ofsSurfaces ); - for ( i = 0 ; i < lod->numSurfaces ; i++ ) { - shader = R_GetShaderByHandle( surface->shaderIndex ); - R_AddDrawSurf( (void *)surface, shader, 0 /*fogNum*/, qfalse ); - surface = (md4Surface_t *)( (byte *)surface + surface->ofsEnd ); - } -} - -/* -============== -RB_SurfaceAnim -============== -*/ -void RB_SurfaceAnim( md4Surface_t *surface ) { - int i, j, k; - float frontlerp, backlerp; - int *triangles; - int indexes; - int baseIndex, baseVertex; - int numVerts; - md4Vertex_t *v; - md4Bone_t bones[MD4_MAX_BONES]; - md4Bone_t *bonePtr, *bone; - md4Header_t *header; - md4Frame_t *frame; - md4Frame_t *oldFrame; - int frameSize; - - - if ( backEnd.currentEntity->e.oldframe == backEnd.currentEntity->e.frame ) { - backlerp = 0; - frontlerp = 1; - } else { - backlerp = backEnd.currentEntity->e.backlerp; - frontlerp = 1.0f - backlerp; - } - header = (md4Header_t *)((byte *)surface + surface->ofsHeader); - - frameSize = (size_t)( &((md4Frame_t *)0)->bones[ header->numBones ] ); - - frame = (md4Frame_t *)((byte *)header + header->ofsFrames + - backEnd.currentEntity->e.frame * frameSize ); - oldFrame = (md4Frame_t *)((byte *)header + header->ofsFrames + - backEnd.currentEntity->e.oldframe * frameSize ); - - RB_CheckOverflow( surface->numVerts, surface->numTriangles * 3 ); - - triangles = (int *) ((byte *)surface + surface->ofsTriangles); - indexes = surface->numTriangles * 3; - baseIndex = tess.numIndexes; - baseVertex = tess.numVertexes; - for (j = 0 ; j < indexes ; j++) { - tess.indexes[baseIndex + j] = baseIndex + triangles[j]; - } - tess.numIndexes += indexes; - - // - // lerp all the needed bones - // - if ( !backlerp ) { - // no lerping needed - bonePtr = frame->bones; - } else { - bonePtr = bones; - for ( i = 0 ; i < header->numBones*12 ; i++ ) { - ((float *)bonePtr)[i] = frontlerp * ((float *)frame->bones)[i] - + backlerp * ((float *)oldFrame->bones)[i]; - } - } - - // - // deform the vertexes by the lerped bones - // - numVerts = surface->numVerts; - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *) ((byte *)surface + surface->ofsVerts + 12); - v = (md4Vertex_t *) ((byte *)surface + surface->ofsVerts); - for ( j = 0; j < numVerts; j++ ) { - vec3_t tempVert, tempNormal; - md4Weight_t *w; - - VectorClear( tempVert ); - VectorClear( tempNormal ); - w = v->weights; - for ( k = 0 ; k < v->numWeights ; k++, w++ ) { - bone = bonePtr + w->boneIndex; - - tempVert[0] += w->boneWeight * ( DotProduct( bone->matrix[0], w->offset ) + bone->matrix[0][3] ); - tempVert[1] += w->boneWeight * ( DotProduct( bone->matrix[1], w->offset ) + bone->matrix[1][3] ); - tempVert[2] += w->boneWeight * ( DotProduct( bone->matrix[2], w->offset ) + bone->matrix[2][3] ); - - tempNormal[0] += w->boneWeight * DotProduct( bone->matrix[0], v->normal ); - tempNormal[1] += w->boneWeight * DotProduct( bone->matrix[1], v->normal ); - tempNormal[2] += w->boneWeight * DotProduct( bone->matrix[2], v->normal ); - } - - tess.xyz[baseVertex + j][0] = tempVert[0]; - tess.xyz[baseVertex + j][1] = tempVert[1]; - tess.xyz[baseVertex + j][2] = tempVert[2]; - - tess.normal[baseVertex + j][0] = tempNormal[0]; - tess.normal[baseVertex + j][1] = tempNormal[1]; - tess.normal[baseVertex + j][2] = tempNormal[2]; - - tess.texCoords[baseVertex + j][0][0] = v->texCoords[0]; - tess.texCoords[baseVertex + j][0][1] = v->texCoords[1]; - - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *)( ( byte * )&v->weights[v->numWeights] + 12 ); - v = (md4Vertex_t *)&v->weights[v->numWeights]; - } - - tess.numVertexes += surface->numVerts; -} - // copied and adapted from tr_mesh.c @@ -195,7 +61,7 @@ static int R_MDRCullModel( mdrHeader_t *header, trRefEntity_t *ent ) { switch ( R_CullLocalPointAndRadius( newFrame->localOrigin, newFrame->radius ) ) { // Ummm... yeah yeah I know we don't really have an md3 here.. but we pretend - // we do. After all, the purpose of md4s are not that different, are they? + // we do. After all, the purpose of mdrs are not that different, are they? case CULL_OUT: tr.pc.c_sphere_cull_md3_out++; @@ -442,7 +308,7 @@ void R_MDRAddAnimSurfaces( trRefEntity_t *ent ) { RB_MDRSurfaceAnim ============== */ -void RB_MDRSurfaceAnim( md4Surface_t *surface ) +void RB_MDRSurfaceAnim( mdrSurface_t *surface ) { int i, j, k; float frontlerp, backlerp; @@ -454,7 +320,7 @@ void RB_MDRSurfaceAnim( md4Surface_t *surface ) mdrHeader_t *header; mdrFrame_t *frame; mdrFrame_t *oldFrame; - mdrBone_t bones[MD4_MAX_BONES], *bonePtr, *bone; + mdrBone_t bones[MDR_MAX_BONES], *bonePtr, *bone; int frameSize; diff --git a/code/renderergl1/tr_image.c b/code/renderergl1/tr_image.c index 921bf7d0..d1127829 100644 --- a/code/renderergl1/tr_image.c +++ b/code/renderergl1/tr_image.c @@ -900,6 +900,8 @@ image_t *R_CreateImage( const char *name, byte *pic, int width, int height, qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, glWrapClampMode ); qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, glWrapClampMode ); + // FIXME: this stops fog from setting border color? + glState.currenttextures[glState.currenttmu] = 0; qglBindTexture( GL_TEXTURE_2D, 0 ); if ( image->TMU == 1 ) { diff --git a/code/renderergl1/tr_init.c b/code/renderergl1/tr_init.c index 98cbfc65..15e79729 100644 --- a/code/renderergl1/tr_init.c +++ b/code/renderergl1/tr_init.c @@ -1031,7 +1031,7 @@ void R_Register( void ) r_ignorehwgamma = ri.Cvar_Get( "r_ignorehwgamma", "0", CVAR_ARCHIVE | CVAR_LATCH); r_mode = ri.Cvar_Get( "r_mode", "3", CVAR_ARCHIVE | CVAR_LATCH ); r_fullscreen = ri.Cvar_Get( "r_fullscreen", "1", CVAR_ARCHIVE ); - r_noborder = ri.Cvar_Get("r_noborder", "0", CVAR_ARCHIVE); + r_noborder = ri.Cvar_Get("r_noborder", "0", CVAR_ARCHIVE | CVAR_LATCH); r_customwidth = ri.Cvar_Get( "r_customwidth", "1600", CVAR_ARCHIVE | CVAR_LATCH ); r_customheight = ri.Cvar_Get( "r_customheight", "1024", CVAR_ARCHIVE | CVAR_LATCH ); r_customPixelAspect = ri.Cvar_Get( "r_customPixelAspect", "1", CVAR_ARCHIVE | CVAR_LATCH ); diff --git a/code/renderergl1/tr_local.h b/code/renderergl1/tr_local.h index 3f5a975a..c0c62f03 100644 --- a/code/renderergl1/tr_local.h +++ b/code/renderergl1/tr_local.h @@ -473,7 +473,6 @@ typedef enum { SF_TRIANGLES, SF_POLY, SF_MD3, - SF_MD4, SF_MDR, SF_IQM, SF_FLARE, @@ -732,7 +731,6 @@ typedef enum { MOD_BAD, MOD_BRUSH, MOD_MESH, - MOD_MD4, MOD_MDR, MOD_IQM } modtype_t; @@ -745,7 +743,7 @@ typedef struct model_s { int dataSize; // just for listing purposes bmodel_t *bmodel; // only if type == MOD_BRUSH md3Header_t *md3[MD3_MAX_LODS]; // only if type == MOD_MESH - void *modelData; // only if type == (MOD_MD4 | MOD_MDR | MOD_IQM) + void *modelData; // only if type == (MOD_MDR | MOD_IQM) int numLods; } model_t; @@ -1390,11 +1388,8 @@ ANIMATED MODELS ============================================================= */ -// void R_MakeAnimModel( model_t *model ); haven't seen this one really, so not needed I guess. -void R_AddAnimSurfaces( trRefEntity_t *ent ); -void RB_SurfaceAnim( md4Surface_t *surfType ); void R_MDRAddAnimSurfaces( trRefEntity_t *ent ); -void RB_MDRSurfaceAnim( md4Surface_t *surface ); +void RB_MDRSurfaceAnim( mdrSurface_t *surface ); qboolean R_LoadIQM (model_t *mod, void *buffer, int filesize, const char *name ); void R_AddIQMSurfaces( trRefEntity_t *ent ); void RB_IQMSurfaceAnim( surfaceType_t *surface ); diff --git a/code/renderergl1/tr_main.c b/code/renderergl1/tr_main.c index 79d825f6..6a2118f1 100644 --- a/code/renderergl1/tr_main.c +++ b/code/renderergl1/tr_main.c @@ -1242,9 +1242,6 @@ void R_AddEntitySurfaces (void) { case MOD_MESH: R_AddMD3Surfaces( ent ); break; - case MOD_MD4: - R_AddAnimSurfaces( ent ); - break; case MOD_MDR: R_MDRAddAnimSurfaces( ent ); break; diff --git a/code/renderergl1/tr_model.c b/code/renderergl1/tr_model.c index f9e2c3af..a3f91aee 100644 --- a/code/renderergl1/tr_model.c +++ b/code/renderergl1/tr_model.c @@ -26,7 +26,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA #define LL(x) x=LittleLong(x) static qboolean R_LoadMD3(model_t *mod, int lod, void *buffer, const char *name ); -static qboolean R_LoadMD4(model_t *mod, void *buffer, const char *name ); static qboolean R_LoadMDR(model_t *mod, void *buffer, int filesize, const char *name ); /* @@ -72,15 +71,10 @@ qhandle_t R_RegisterMD3(const char *name, model_t *mod) continue; ident = LittleLong(* (unsigned *) buf.u); - if (ident == MD4_IDENT) - loaded = R_LoadMD4(mod, buf.u, name); + if (ident == MD3_IDENT) + loaded = R_LoadMD3(mod, lod, buf.u, name); else - { - if (ident == MD3_IDENT) - loaded = R_LoadMD3(mod, lod, buf.u, name); - else - ri.Printf(PRINT_WARNING,"R_RegisterMD3: unknown fileid for %s\n", name); - } + ri.Printf(PRINT_WARNING,"R_RegisterMD3: unknown fileid for %s\n", name); ri.FS_FreeFile(buf.v); @@ -200,7 +194,6 @@ static modelExtToLoaderMap_t modelLoaders[ ] = { { "iqm", R_RegisterIQM }, { "mdr", R_RegisterMDR }, - { "md4", R_RegisterMD3 }, { "md3", R_RegisterMD3 } }; @@ -576,7 +569,7 @@ static qboolean R_LoadMDR( model_t *mod, void *buffer, int filesize, const char LL(pinmodel->ofsFrames); // This is a model that uses some type of compressed Bones. We don't want to uncompress every bone for each rendered frame - // over and over again, we'll uncompress it in this function already, so we must adjust the size of the target md4. + // over and over again, we'll uncompress it in this function already, so we must adjust the size of the target mdr. if(pinmodel->ofsFrames < 0) { // mdrFrame_t is larger than mdrCompFrame_t: @@ -873,162 +866,6 @@ static qboolean R_LoadMDR( model_t *mod, void *buffer, int filesize, const char return qtrue; } -/* -================= -R_LoadMD4 -================= -*/ - -static qboolean R_LoadMD4( model_t *mod, void *buffer, const char *mod_name ) { - int i, j, k, lodindex; - md4Header_t *pinmodel, *md4; - md4Frame_t *frame; - md4LOD_t *lod; - md4Surface_t *surf; - md4Triangle_t *tri; - md4Vertex_t *v; - int version; - int size; - shader_t *sh; - int frameSize; - - pinmodel = (md4Header_t *)buffer; - - version = LittleLong (pinmodel->version); - if (version != MD4_VERSION) { - ri.Printf( PRINT_WARNING, "R_LoadMD4: %s has wrong version (%i should be %i)\n", - mod_name, version, MD4_VERSION); - return qfalse; - } - - mod->type = MOD_MD4; - size = LittleLong(pinmodel->ofsEnd); - mod->dataSize += size; - mod->modelData = md4 = ri.Hunk_Alloc( size, h_low ); - - Com_Memcpy(md4, buffer, size); - - LL(md4->ident); - LL(md4->version); - LL(md4->numFrames); - LL(md4->numBones); - LL(md4->numLODs); - LL(md4->ofsFrames); - LL(md4->ofsLODs); - md4->ofsEnd = size; - - if ( md4->numFrames < 1 ) { - ri.Printf( PRINT_WARNING, "R_LoadMD4: %s has no frames\n", mod_name ); - return qfalse; - } - - // we don't need to swap tags in the renderer, they aren't used - - // swap all the frames - frameSize = (size_t)( &((md4Frame_t *)0)->bones[ md4->numBones ] ); - for ( i = 0 ; i < md4->numFrames ; i++) { - frame = (md4Frame_t *) ( (byte *)md4 + md4->ofsFrames + i * frameSize ); - frame->radius = LittleFloat( frame->radius ); - for ( j = 0 ; j < 3 ; j++ ) { - frame->bounds[0][j] = LittleFloat( frame->bounds[0][j] ); - frame->bounds[1][j] = LittleFloat( frame->bounds[1][j] ); - frame->localOrigin[j] = LittleFloat( frame->localOrigin[j] ); - } - for ( j = 0 ; j < md4->numBones * sizeof( md4Bone_t ) / 4 ; j++ ) { - ((float *)frame->bones)[j] = LittleFloat( ((float *)frame->bones)[j] ); - } - } - - // swap all the LOD's - lod = (md4LOD_t *) ( (byte *)md4 + md4->ofsLODs ); - for ( lodindex = 0 ; lodindex < md4->numLODs ; lodindex++ ) { - - // swap all the surfaces - surf = (md4Surface_t *) ( (byte *)lod + lod->ofsSurfaces ); - for ( i = 0 ; i < lod->numSurfaces ; i++) { - LL(surf->ident); - LL(surf->numTriangles); - LL(surf->ofsTriangles); - LL(surf->numVerts); - LL(surf->ofsVerts); - LL(surf->ofsEnd); - - if ( surf->numVerts >= SHADER_MAX_VERTEXES ) { - ri.Printf(PRINT_WARNING, "R_LoadMD4: %s has more than %i verts on %s (%i).\n", - mod_name, SHADER_MAX_VERTEXES - 1, surf->name[0] ? surf->name : "a surface", - surf->numVerts ); - return qfalse; - } - if ( surf->numTriangles*3 >= SHADER_MAX_INDEXES ) { - ri.Printf(PRINT_WARNING, "R_LoadMD4: %s has more than %i triangles on %s (%i).\n", - mod_name, ( SHADER_MAX_INDEXES / 3 ) - 1, surf->name[0] ? surf->name : "a surface", - surf->numTriangles ); - return qfalse; - } - - // change to surface identifier - surf->ident = SF_MD4; - - // lowercase the surface name so skin compares are faster - Q_strlwr( surf->name ); - - // register the shaders - sh = R_FindShader( surf->shader, LIGHTMAP_NONE, qtrue ); - if ( sh->defaultShader ) { - surf->shaderIndex = 0; - } else { - surf->shaderIndex = sh->index; - } - - // swap all the triangles - tri = (md4Triangle_t *) ( (byte *)surf + surf->ofsTriangles ); - for ( j = 0 ; j < surf->numTriangles ; j++, tri++ ) { - LL(tri->indexes[0]); - LL(tri->indexes[1]); - LL(tri->indexes[2]); - } - - // swap all the vertexes - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *) ( (byte *)surf + surf->ofsVerts + 12); - v = (md4Vertex_t *) ( (byte *)surf + surf->ofsVerts); - for ( j = 0 ; j < surf->numVerts ; j++ ) { - v->normal[0] = LittleFloat( v->normal[0] ); - v->normal[1] = LittleFloat( v->normal[1] ); - v->normal[2] = LittleFloat( v->normal[2] ); - - v->texCoords[0] = LittleFloat( v->texCoords[0] ); - v->texCoords[1] = LittleFloat( v->texCoords[1] ); - - v->numWeights = LittleLong( v->numWeights ); - - for ( k = 0 ; k < v->numWeights ; k++ ) { - v->weights[k].boneIndex = LittleLong( v->weights[k].boneIndex ); - v->weights[k].boneWeight = LittleFloat( v->weights[k].boneWeight ); - v->weights[k].offset[0] = LittleFloat( v->weights[k].offset[0] ); - v->weights[k].offset[1] = LittleFloat( v->weights[k].offset[1] ); - v->weights[k].offset[2] = LittleFloat( v->weights[k].offset[2] ); - } - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *)( ( byte * )&v->weights[v->numWeights] + 12 ); - v = (md4Vertex_t *)( ( byte * )&v->weights[v->numWeights]); - } - - // find the next surface - surf = (md4Surface_t *)( (byte *)surf + surf->ofsEnd ); - } - - // find the next LOD - lod = (md4LOD_t *)( (byte *)lod + lod->ofsEnd ); - } - - return qtrue; -} - //============================================================================= @@ -1049,11 +886,6 @@ void RE_BeginRegistration( glconfig_t *glconfigOut ) { RE_ClearScene(); tr.registered = qtrue; - - // NOTE: this sucks, for some reason the first stretch pic is never drawn - // without this we'd see a white flash on a level load because the very - // first time the level shot would not be drawn -// RE_StretchPic(0, 0, 0, 0, 0, 0, 1, 1, 0); } //============================================================================= @@ -1265,17 +1097,6 @@ void R_ModelBounds( qhandle_t handle, vec3_t mins, vec3_t maxs ) { VectorCopy( frame->bounds[0], mins ); VectorCopy( frame->bounds[1], maxs ); - return; - } else if (model->type == MOD_MD4) { - md4Header_t *header; - md4Frame_t *frame; - - header = (md4Header_t *)model->modelData; - frame = (md4Frame_t *) ((byte *)header + header->ofsFrames); - - VectorCopy( frame->bounds[0], mins ); - VectorCopy( frame->bounds[1], maxs ); - return; } else if (model->type == MOD_MDR) { mdrHeader_t *header; diff --git a/code/renderergl1/tr_surface.c b/code/renderergl1/tr_surface.c index 3b4a943b..1879fe2c 100644 --- a/code/renderergl1/tr_surface.c +++ b/code/renderergl1/tr_surface.c @@ -344,6 +344,8 @@ static void DoRailCore( const vec3_t start, const vec3_t end, const vec3_t up, f int vbase; float t = len / 256.0f; + RB_CHECKOVERFLOW( 4, 6 ); + vbase = tess.numVertexes; spanWidth2 = -spanWidth; @@ -1233,7 +1235,6 @@ void (*rb_surfaceTable[SF_NUM_SURFACE_TYPES])( void *) = { (void(*)(void*))RB_SurfaceTriangles, // SF_TRIANGLES, (void(*)(void*))RB_SurfacePolychain, // SF_POLY, (void(*)(void*))RB_SurfaceMesh, // SF_MD3, - (void(*)(void*))RB_SurfaceAnim, // SF_MD4, (void(*)(void*))RB_MDRSurfaceAnim, // SF_MDR, (void(*)(void*))RB_IQMSurfaceAnim, // SF_IQM, (void(*)(void*))RB_SurfaceFlare, // SF_FLARE, diff --git a/code/renderergl2/glsl/bokeh_vp.glsl b/code/renderergl2/glsl/bokeh_vp.glsl index 5ca41600..bdaa74af 100644 --- a/code/renderergl2/glsl/bokeh_vp.glsl +++ b/code/renderergl2/glsl/bokeh_vp.glsl @@ -1,4 +1,4 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec4 attr_TexCoord0; uniform mat4 u_ModelViewProjectionMatrix; @@ -8,6 +8,6 @@ varying vec2 var_TexCoords; void main() { - gl_Position = u_ModelViewProjectionMatrix * attr_Position; + gl_Position = u_ModelViewProjectionMatrix * vec4(attr_Position, 1.0); var_TexCoords = attr_TexCoord0.st; } diff --git a/code/renderergl2/glsl/calclevels4x_vp.glsl b/code/renderergl2/glsl/calclevels4x_vp.glsl index 5ca41600..bdaa74af 100644 --- a/code/renderergl2/glsl/calclevels4x_vp.glsl +++ b/code/renderergl2/glsl/calclevels4x_vp.glsl @@ -1,4 +1,4 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec4 attr_TexCoord0; uniform mat4 u_ModelViewProjectionMatrix; @@ -8,6 +8,6 @@ varying vec2 var_TexCoords; void main() { - gl_Position = u_ModelViewProjectionMatrix * attr_Position; + gl_Position = u_ModelViewProjectionMatrix * vec4(attr_Position, 1.0); var_TexCoords = attr_TexCoord0.st; } diff --git a/code/renderergl2/glsl/dlight_vp.glsl b/code/renderergl2/glsl/dlight_vp.glsl index d9fd71d0..9566a04c 100644 --- a/code/renderergl2/glsl/dlight_vp.glsl +++ b/code/renderergl2/glsl/dlight_vp.glsl @@ -1,4 +1,4 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec4 attr_TexCoord0; attribute vec3 attr_Normal; @@ -32,7 +32,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) if (u_DeformGen == DGEN_BULGE) { - phase *= M_PI * 0.25 * st.x; + phase *= st.x; } else // if (u_DeformGen <= DGEN_WAVE_INVERSE_SAWTOOTH) { @@ -48,7 +48,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) } else if (u_DeformGen == DGEN_WAVE_SQUARE) { - func = sign(sin(value * 2.0 * M_PI)); + func = sign(0.5 - fract(value)); } else if (u_DeformGen == DGEN_WAVE_TRIANGLE) { @@ -62,7 +62,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) { func = (1.0 - fract(value)); } - else if (u_DeformGen == DGEN_BULGE) + else // if (u_DeformGen == DGEN_BULGE) { func = sin(value); } @@ -73,16 +73,16 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) void main() { - vec4 position = attr_Position; - vec3 normal = attr_Normal; + vec3 position = attr_Position; + vec3 normal = attr_Normal * 2.0 - vec3(1.0); #if defined(USE_DEFORM_VERTEXES) - position.xyz = DeformPosition(position.xyz, normal, attr_TexCoord0.st); + position = DeformPosition(position, normal, attr_TexCoord0.st); #endif - gl_Position = u_ModelViewProjectionMatrix * position; + gl_Position = u_ModelViewProjectionMatrix * vec4(position, 1.0); - vec3 dist = u_DlightInfo.xyz - position.xyz; + vec3 dist = u_DlightInfo.xyz - position; var_Tex1 = dist.xy * u_DlightInfo.a + vec2(0.5); float dlightmod = step(0.0, dot(dist, normal)); diff --git a/code/renderergl2/glsl/down4x_vp.glsl b/code/renderergl2/glsl/down4x_vp.glsl index 5ca41600..bdaa74af 100644 --- a/code/renderergl2/glsl/down4x_vp.glsl +++ b/code/renderergl2/glsl/down4x_vp.glsl @@ -1,4 +1,4 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec4 attr_TexCoord0; uniform mat4 u_ModelViewProjectionMatrix; @@ -8,6 +8,6 @@ varying vec2 var_TexCoords; void main() { - gl_Position = u_ModelViewProjectionMatrix * attr_Position; + gl_Position = u_ModelViewProjectionMatrix * vec4(attr_Position, 1.0); var_TexCoords = attr_TexCoord0.st; } diff --git a/code/renderergl2/glsl/fogpass_fp.glsl b/code/renderergl2/glsl/fogpass_fp.glsl index 91884304..e2ad465b 100644 --- a/code/renderergl2/glsl/fogpass_fp.glsl +++ b/code/renderergl2/glsl/fogpass_fp.glsl @@ -5,5 +5,5 @@ varying float var_Scale; void main() { gl_FragColor = u_Color; - gl_FragColor.a *= sqrt(clamp(var_Scale, 0.0, 1.0)); + gl_FragColor.a = sqrt(clamp(var_Scale, 0.0, 1.0)); } diff --git a/code/renderergl2/glsl/fogpass_vp.glsl b/code/renderergl2/glsl/fogpass_vp.glsl index ce0e015b..8f7bc728 100644 --- a/code/renderergl2/glsl/fogpass_vp.glsl +++ b/code/renderergl2/glsl/fogpass_vp.glsl @@ -1,27 +1,30 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec3 attr_Normal; + attribute vec4 attr_TexCoord0; -//#if defined(USE_VERTEX_ANIMATION) -attribute vec4 attr_Position2; +#if defined(USE_VERTEX_ANIMATION) +attribute vec3 attr_Position2; attribute vec3 attr_Normal2; -//#endif +#endif uniform vec4 u_FogDistance; uniform vec4 u_FogDepth; uniform float u_FogEyeT; -//#if defined(USE_DEFORM_VERTEXES) +#if defined(USE_DEFORM_VERTEXES) uniform int u_DeformGen; uniform float u_DeformParams[5]; -//#endif +#endif uniform float u_Time; uniform mat4 u_ModelViewProjectionMatrix; -//#if defined(USE_VERTEX_ANIMATION) +#if defined(USE_VERTEX_ANIMATION) uniform float u_VertexLerp; -//#endif +#endif + +uniform vec4 u_Color; varying float var_Scale; @@ -41,7 +44,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) if (u_DeformGen == DGEN_BULGE) { - phase *= M_PI * 0.25 * st.x; + phase *= st.x; } else // if (u_DeformGen <= DGEN_WAVE_INVERSE_SAWTOOTH) { @@ -57,7 +60,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) } else if (u_DeformGen == DGEN_WAVE_SQUARE) { - func = sign(sin(value * 2.0 * M_PI)); + func = sign(0.5 - fract(value)); } else if (u_DeformGen == DGEN_WAVE_TRIANGLE) { @@ -71,7 +74,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) { func = (1.0 - fract(value)); } - else if (u_DeformGen == DGEN_BULGE) + else // if (u_DeformGen == DGEN_BULGE) { func = sin(value); } @@ -80,15 +83,15 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) } #endif -float CalcFog(vec4 position) +float CalcFog(vec3 position) { - float s = dot(position, u_FogDistance) * 8.0; - float t = dot(position, u_FogDepth); + float s = dot(vec4(position, 1.0), u_FogDistance) * 8.0; + float t = dot(vec4(position, 1.0), u_FogDepth); - float eyeOutside = step(0.0, -u_FogEyeT); - float fogged = step(eyeOutside, t); - - t = max(t, 1e-6); + float eyeOutside = float(u_FogEyeT < 0.0); + float fogged = float(t >= eyeOutside); + + t += 1e-6; t *= fogged / (t - u_FogEyeT * eyeOutside); return s * t; @@ -97,18 +100,19 @@ float CalcFog(vec4 position) void main() { #if defined(USE_VERTEX_ANIMATION) - vec4 position = mix(attr_Position, attr_Position2, u_VertexLerp); - vec3 normal = normalize(mix(attr_Normal, attr_Normal2, u_VertexLerp)); + vec3 position = mix(attr_Position, attr_Position2, u_VertexLerp); + vec3 normal = mix(attr_Normal, attr_Normal2, u_VertexLerp); + normal = normalize(normal - vec3(0.5)); #else - vec4 position = attr_Position; - vec3 normal = attr_Normal; + vec3 position = attr_Position; + vec3 normal = attr_Normal * 2.0 - vec3(1.0); #endif #if defined(USE_DEFORM_VERTEXES) position.xyz = DeformPosition(position.xyz, normal, attr_TexCoord0.st); #endif - gl_Position = u_ModelViewProjectionMatrix * position; + gl_Position = u_ModelViewProjectionMatrix * vec4(position, 1.0); - var_Scale = CalcFog(position); + var_Scale = CalcFog(position) * u_Color.a * u_Color.a; } diff --git a/code/renderergl2/glsl/generic_fp.glsl b/code/renderergl2/glsl/generic_fp.glsl index 997d4daa..f485797f 100644 --- a/code/renderergl2/glsl/generic_fp.glsl +++ b/code/renderergl2/glsl/generic_fp.glsl @@ -37,6 +37,8 @@ void main() { color = color2; } + + //color = color * (u_Texture1Env.xxxx + color2 * u_Texture1Env.z) + color2 * u_Texture1Env.y; #endif gl_FragColor = color * var_Color; diff --git a/code/renderergl2/glsl/generic_vp.glsl b/code/renderergl2/glsl/generic_vp.glsl index baae6046..0e5b38b4 100644 --- a/code/renderergl2/glsl/generic_vp.glsl +++ b/code/renderergl2/glsl/generic_vp.glsl @@ -73,7 +73,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) if (u_DeformGen == DGEN_BULGE) { - phase *= M_PI * 0.25 * st.x; + phase *= st.x; } else // if (u_DeformGen <= DGEN_WAVE_INVERSE_SAWTOOTH) { @@ -89,11 +89,11 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) } else if (u_DeformGen == DGEN_WAVE_SQUARE) { - func = sign(sin(value * 2.0 * M_PI)); + func = sign(fract(0.5 - value)); } else if (u_DeformGen == DGEN_WAVE_TRIANGLE) { - func = 1.0 - abs(4.0 * fract(value + 0.25) - 2.0); + func = abs(fract(value + 0.75) - 0.5) * 4.0 - 1.0; } else if (u_DeformGen == DGEN_WAVE_SAWTOOTH) { @@ -103,7 +103,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) { func = (1.0 - fract(value)); } - else if (u_DeformGen == DGEN_BULGE) + else // if (u_DeformGen == DGEN_BULGE) { func = sin(value); } @@ -124,7 +124,9 @@ vec2 GenTexCoords(int TCGen, vec3 position, vec3 normal, vec3 TCGenVector0, vec3 else if (TCGen == TCGEN_ENVIRONMENT_MAPPED) { vec3 viewer = normalize(u_LocalViewOrigin - position); - tex = -reflect(viewer, normal).yz * vec2(0.5, -0.5) + 0.5; + vec2 ref = reflect(viewer, normal).yz; + tex.s = ref.x * -0.5 + 0.5; + tex.t = ref.y * 0.5 + 0.5; } else if (TCGen == TCGEN_VECTOR) { @@ -139,13 +141,14 @@ vec2 GenTexCoords(int TCGen, vec3 position, vec3 normal, vec3 TCGenVector0, vec3 vec2 ModTexCoords(vec2 st, vec3 position, vec4 texMatrix, vec4 offTurb) { float amplitude = offTurb.z; - float phase = offTurb.w; - vec2 st2 = vec2(dot(st, texMatrix.xz), dot(st, texMatrix.yw)) + offTurb.xy; + float phase = offTurb.w * 2.0 * M_PI; + vec2 st2; + st2.x = st.x * texMatrix.x + (st.y * texMatrix.z + offTurb.x); + st2.y = st.x * texMatrix.y + (st.y * texMatrix.w + offTurb.y); - vec3 offsetPos = position / 1024.0; - offsetPos.x += offsetPos.z; + vec2 offsetPos = vec2(position.x + position.z, position.y); - vec2 texOffset = sin((offsetPos.xy + vec2(phase)) * 2.0 * M_PI); + vec2 texOffset = sin(offsetPos * (2.0 * M_PI / 1024.0) + vec2(phase)); return st2 + texOffset * amplitude; } @@ -186,13 +189,13 @@ vec4 CalcColor(vec3 position, vec3 normal) #if defined(USE_FOG) float CalcFog(vec3 position) { - float s = (dot(position, u_FogDistance.xyz) + u_FogDistance.w) * 8.0; - float t = dot(position, u_FogDepth.xyz) + u_FogDepth.w; + float s = dot(vec4(position, 1.0), u_FogDistance) * 8.0; + float t = dot(vec4(position, 1.0), u_FogDepth); - float eyeOutside = step(0.0, -u_FogEyeT); - float fogged = step(eyeOutside, t); - - t = max(t, 1e-6); + float eyeOutside = float(u_FogEyeT < 0.0); + float fogged = float(t < eyeOutside); + + t += 1e-6; t *= fogged / (t - u_FogEyeT * eyeOutside); return s * t; @@ -202,8 +205,9 @@ float CalcFog(vec3 position) void main() { #if defined(USE_VERTEX_ANIMATION) - vec3 position = mix(attr_Position, attr_Position2, u_VertexLerp); - vec3 normal = normalize(mix(attr_Normal, attr_Normal2, u_VertexLerp) * 2.0 - vec3(1.0)); + vec3 position = mix(attr_Position, attr_Position2, u_VertexLerp); + vec3 normal = mix(attr_Normal, attr_Normal2, u_VertexLerp); + normal = normalize(normal - vec3(0.5)); #else vec3 position = attr_Position; vec3 normal = attr_Normal * 2.0 - vec3(1.0); diff --git a/code/renderergl2/glsl/lightall_fp.glsl b/code/renderergl2/glsl/lightall_fp.glsl index b937483e..d134e409 100644 --- a/code/renderergl2/glsl/lightall_fp.glsl +++ b/code/renderergl2/glsl/lightall_fp.glsl @@ -166,22 +166,41 @@ vec3 EnvironmentBRDF(float gloss, float NE, vec3 specular) float a1 = t.w; return clamp( a0 + specular * ( a1 - a0 ), 0.0, 1.0 ); #elif 0 - // from http://seblagarde.wordpress.com/2011/08/17/hello-world/ + // from http://seblagarde.wordpress.com/2011/08/17/hello-world/ return mix(specular.rgb, max(specular.rgb, vec3(gloss)), CalcFresnel(NE)); #else - // from http://advances.realtimerendering.com/s2011/Lazarov-Physically-Based-Lighting-in-Black-Ops%20%28Siggraph%202011%20Advances%20in%20Real-Time%20Rendering%20Course%29.pptx + // from http://advances.realtimerendering.com/s2011/Lazarov-Physically-Based-Lighting-in-Black-Ops%20%28Siggraph%202011%20Advances%20in%20Real-Time%20Rendering%20Course%29.pptx return mix(specular.rgb, vec3(1.0), CalcFresnel(NE) / (4.0 - 3.0 * gloss)); #endif } float CalcBlinn(float NH, float shininess) { -#if 0 - // from http://seblagarde.wordpress.com/2012/06/03/spherical-gaussien-approximation-for-blinn-phong-phong-and-fresnel/ - float a = shininess + 0.775; - return exp(a * NH - a); +#if defined(USE_BLINN) || defined(USE_BLINN_FRESNEL) + // Normalized Blinn-Phong + float norm = shininess * 0.125 + 1.0; +#elif defined(USE_MCAULEY) + // Cook-Torrance as done by Stephen McAuley + // http://blog.selfshadow.com/publications/s2012-shading-course/mcauley/s2012_pbs_farcry3_notes_v2.pdf + float norm = shininess * 0.25 + 0.125; +#elif defined(USE_GOTANDA) + // Neumann-Neumann as done by Yoshiharu Gotanda + // http://research.tri-ace.com/Data/s2012_beyond_CourseNotes.pdf + float norm = shininess * 0.124858 + 0.269182; +#elif defined(USE_LAZAROV) + // Cook-Torrance as done by Dimitar Lazarov + // http://blog.selfshadow.com/publications/s2013-shading-course/lazarov/s2013_pbs_black_ops_2_notes.pdf + float norm = shininess * 0.125 + 0.25; #else - return pow(NH, shininess); + float norm = 1.0; +#endif + +#if 0 + // from http://seblagarde.wordpress.com/2012/06/03/spherical-gaussien-approximation-for-blinn-phong-phong-and-fresnel/ + float a = shininess + 0.775; + return norm * exp(a * NH - a); +#else + return norm * pow(NH, shininess); #endif } @@ -215,145 +234,122 @@ float CalcFresnel(float EH) float CalcVisibility(float NH, float NL, float NE, float EH, float gloss) { -#if 1 - // From http://blog.selfshadow.com/publications/s2013-shading-course/lazarov/s2013_pbs_black_ops_2_notes.pdf +#if defined(USE_GOTANDA) + // Neumann-Neumann as done by Yoshiharu Gotanda + // http://research.tri-ace.com/Data/s2012_beyond_CourseNotes.pdf + return 1.0 / max(max(NL, NE), EPSILON); +#elif defined(USE_LAZAROV) + // Cook-Torrance as done by Dimitar Lazarov + // http://blog.selfshadow.com/publications/s2013-shading-course/lazarov/s2013_pbs_black_ops_2_notes.pdf float k = min(1.0, gloss + 0.545); - return 1.0 / (k * EH * EH + (1.0 - k)); -#elif 0 + return 1.0 / (k * (EH * EH - 1.0) + 1.0); +#elif defined(USE_GGX) float roughness = exp2(gloss * -6.5); - #if defined(USE_GGX) - // From http://blog.selfshadow.com/publications/s2013-shading-course/karis/s2013_pbs_epic_notes_v2.pdf - float k = roughness + 1.0; - k *= k * 0.125; - #else - float k = roughness; - #endif // Modified from http://blog.selfshadow.com/publications/s2013-shading-course/karis/s2013_pbs_epic_notes_v2.pdf // NL, NE in numerator factored out from cook-torrance + float k = roughness + 1.0; + k *= k * 0.125; + float k2 = 1.0 - k; float invGeo1 = NL * k2 + k; float invGeo2 = NE * k2 + k; - + return 1.0 / (invGeo1 * invGeo2); #else - float geo = 2.0 * NH * min(NE, NL); - geo /= max(EH, geo); - - return geo; + return 1.0; #endif } vec3 CalcSpecular(vec3 specular, float NH, float NL, float NE, float EH, float gloss, float shininess) { - float blinn = CalcBlinn(NH, shininess); +#if defined(USE_GGX) + float distrib = CalcGGX(NH, gloss); +#else + float distrib = CalcBlinn(NH, shininess); +#endif + +#if defined(USE_BLINN) + vec3 fSpecular = specular; +#else vec3 fSpecular = mix(specular, vec3(1.0), CalcFresnel(EH)); +#endif + float vis = CalcVisibility(NH, NL, NE, EH, gloss); - #if defined(USE_BLINN) - // Normalized Blinn-Phong - return specular * blinn * (shininess * 0.125 + 1.0); - #elif defined(USE_BLINN_FRESNEL) - // Normalized Blinn-Phong with Fresnel - return fSpecular * blinn * (shininess * 0.125 + 1.0); - #elif defined(USE_MCAULEY) - // Cook-Torrance as done by Stephen McAuley - // http://blog.selfshadow.com/publications/s2012-shading-course/mcauley/s2012_pbs_farcry3_notes_v2.pdf - return fSpecular * blinn * (shininess * 0.25 + 0.125); - #elif defined(USE_GOTANDA) - // Neumann-Neumann as done by Yoshiharu Gotanda - // http://research.tri-ace.com/Data/s2012_beyond_CourseNotes.pdf - return fSpecular * blinn * (shininess * 0.124858 + 0.269182) / max(max(NL, NE), EPSILON); - #elif defined(USE_LAZAROV) - // Cook-Torrance as done by Dimitar Lazarov - // http://blog.selfshadow.com/publications/s2013-shading-course/lazarov/s2013_pbs_black_ops_2_notes.pdf - return fSpecular * blinn * (shininess * 0.125 + 0.25) * vis; - #endif - - return vec3(0.0); + return fSpecular * (distrib * vis); } -float CalcLightAttenuation(vec3 dir, float sqrRadius) +float CalcLightAttenuation(float point, float normDist) { - // point light at >0 radius, directional otherwise - float point = float(sqrRadius > 0.0); - - // inverse square light - float attenuation = sqrRadius / dot(dir, dir); - - // zero light at radius, approximating q3 style + // zero light at 1.0, approximating q3 style // also don't attenuate directional light - attenuation = (0.5 * attenuation - 1.5) * point + 1.0; - + float attenuation = (0.5 * normDist - 1.5) * point + 1.0; + // clamp attenuation #if defined(NO_LIGHT_CLAMP) attenuation = max(attenuation, 0.0); #else attenuation = clamp(attenuation, 0.0, 1.0); #endif - + return attenuation; } // from http://www.thetenthplanet.de/archives/1180 mat3 cotangent_frame( vec3 N, vec3 p, vec2 uv ) { - // get edge vectors of the pixel triangle - vec3 dp1 = dFdx( p ); - vec3 dp2 = dFdy( p ); - vec2 duv1 = dFdx( uv ); - vec2 duv2 = dFdy( uv ); - - // solve the linear system - vec3 dp2perp = cross( dp2, N ); - vec3 dp1perp = cross( N, dp1 ); - vec3 T = dp2perp * duv1.x + dp1perp * duv2.x; - vec3 B = dp2perp * duv1.y + dp1perp * duv2.y; - - // construct a scale-invariant frame - float invmax = inversesqrt( max( dot(T,T), dot(B,B) ) ); - return mat3( T * invmax, B * invmax, N ); + // get edge vectors of the pixel triangle + vec3 dp1 = dFdx( p ); + vec3 dp2 = dFdy( p ); + vec2 duv1 = dFdx( uv ); + vec2 duv2 = dFdy( uv ); + + // solve the linear system + vec3 dp2perp = cross( dp2, N ); + vec3 dp1perp = cross( N, dp1 ); + vec3 T = dp2perp * duv1.x + dp1perp * duv2.x; + vec3 B = dp2perp * duv1.y + dp1perp * duv2.y; + + // construct a scale-invariant frame + float invmax = inversesqrt( max( dot(T,T), dot(B,B) ) ); + return mat3( T * invmax, B * invmax, N ); } void main() { vec3 L, N, E, H; float NL, NH, NE, EH; - -#if (defined(USE_LIGHT) && !defined(USE_FAST_LIGHT)) || defined(USE_PARALLAXMAP) + +#if defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) #if defined(USE_VERT_TANGENT_SPACE) mat3 tangentToWorld = mat3(var_Tangent.xyz, var_Bitangent.xyz, var_Normal.xyz); + E = vec3(var_Normal.w, var_Tangent.w, var_Bitangent.w); #else mat3 tangentToWorld = cotangent_frame(var_Normal, -var_ViewDir, var_TexCoords.xy); + E = var_ViewDir; #endif -#endif -#if defined(USE_DELUXEMAP) - L = texture2D(u_DeluxeMap, var_TexCoords.zw).xyz - vec3(0.5); - L = L * u_EnableTextures.y + var_LightDir.xyz; -#elif defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) + E = normalize(E); + L = var_LightDir.xyz; -#endif - -#if (defined(USE_LIGHT) && !defined(USE_FAST_LIGHT)) || defined(USE_PARALLAXMAP) - #if defined(USE_VERT_TANGENT_SPACE) - E = normalize(vec3(var_Normal.w, var_Tangent.w, var_Bitangent.w)); - #else - E = normalize(var_ViewDir); + #if defined(USE_DELUXEMAP) + L += (texture2D(u_DeluxeMap, var_TexCoords.zw).xyz - vec3(0.5)) * u_EnableTextures.y; #endif + float sqrLightDist = dot(L, L); #endif #if defined(USE_LIGHTMAP) - vec4 lightSample = texture2D(u_LightMap, var_TexCoords.zw).rgba; - #if defined(RGBM_LIGHTMAP) - lightSample.rgb *= 32.0 * lightSample.a; - #endif + vec4 lightSample = texture2D(u_LightMap, var_TexCoords.zw); vec3 lightColor = lightSample.rgb; + #if defined(RGBM_LIGHTMAP) + lightColor *= 32.0 * lightSample.a; + #endif #elif defined(USE_LIGHT_VECTOR) && !defined(USE_FAST_LIGHT) - vec3 lightColor = u_DirectedLight * CalcLightAttenuation(L, var_LightDir.w); + vec3 lightColor = u_DirectedLight * CalcLightAttenuation(float(var_LightDir.w > 0.0), var_LightDir.w / sqrLightDist); vec3 ambientColor = u_AmbientLight; #elif defined(USE_LIGHT_VERTEX) && !defined(USE_FAST_LIGHT) vec3 lightColor = var_LightColor; @@ -383,21 +379,22 @@ void main() N.xy = texture2D(u_NormalMap, texCoords).rg - vec2(0.5); #endif N.xy *= u_EnableTextures.x; - N.z = sqrt(0.25 - dot(N.xy, N.xy)); - N = normalize(tangentToWorld * N); + N.z = sqrt((0.25 - N.x * N.x) - N.y * N.y); + N = tangentToWorld * N; #else - N = normalize(var_Normal.xyz); + N = var_Normal.xyz; #endif - - L = normalize(L); + + N = normalize(N); + L /= sqrt(sqrLightDist); #if defined(USE_SHADOWMAP) vec2 shadowTex = gl_FragCoord.xy * r_FBufScale; float shadowValue = texture2D(u_ShadowMap, shadowTex).r; // surfaces not facing the light are always shadowed - shadowValue *= step(0.0, dot(var_Normal.xyz, var_PrimaryLightDir.xyz)); - + shadowValue *= float(dot(var_Normal.xyz, var_PrimaryLightDir.xyz) > 0.0); + #if defined(SHADOWMAP_MODULATE) //vec3 shadowColor = min(u_PrimaryLightAmbient, lightColor); vec3 shadowColor = u_PrimaryLightAmbient * lightColor; @@ -428,44 +425,41 @@ void main() NL = clamp(dot(N, L), 0.0, 1.0); NE = clamp(dot(N, E), 0.0, 1.0); + vec4 specular = vec4(1.0); #if defined(USE_SPECULARMAP) - vec4 specular = texture2D(u_SpecularMap, texCoords); - specular = (specular - vec4(1.0)) * u_EnableTextures.z + vec4(1.0); + specular += texture2D(u_SpecularMap, texCoords) * u_EnableTextures.z - u_EnableTextures.zzzz; #if defined(USE_GAMMA2_TEXTURES) specular.rgb *= specular.rgb; #endif - #else - vec4 specular = vec4(1.0); #endif specular *= u_MaterialInfo.xxxy; - + float gloss = specular.a; float shininess = exp2(gloss * 13.0); #if defined(SPECULAR_IS_METALLIC) - // diffuse is actually base color, and red of specular is metallicness + // diffuse is actually base color, and red of specular is metallicness float metallic = specular.r; - - specular.rgb = vec3(0.04) + 0.96 * diffuse.rgb * metallic; + + specular.rgb = (0.96 * metallic) * diffuse.rgb + vec3(0.04); diffuse.rgb *= 1.0 - metallic; #else // adjust diffuse by specular reflectance, to maintain energy conservation diffuse.rgb *= vec3(1.0) - specular.rgb; #endif - - + reflectance = CalcDiffuse(diffuse.rgb, N, L, E, NE, NL, shininess); #if defined(r_deluxeSpecular) || defined(USE_LIGHT_VECTOR) float adjGloss = gloss; float adjShininess = shininess; - - #if !defined(USE_LIGHT_VECTOR) + + #if !defined(USE_LIGHT_VECTOR) adjGloss *= r_deluxeSpecular; adjShininess = exp2(adjGloss * 13.0); - #endif - + #endif + H = normalize(L + E); EH = clamp(dot(E, H), 0.0, 1.0); @@ -477,16 +471,16 @@ void main() reflectance += CalcSpecular(specular.rgb, NH, NL, NE, EH, adjGloss, adjShininess); #endif #endif - - gl_FragColor.rgb = lightColor * reflectance * NL; + + gl_FragColor.rgb = lightColor * reflectance * NL; gl_FragColor.rgb += ambientColor * (diffuse.rgb + specular.rgb); - + #if defined(USE_CUBEMAP) reflectance = EnvironmentBRDF(gloss, NE, specular.rgb); vec3 R = reflect(E, N); - vec3 cubeLightColor = textureCubeLod(u_CubeMap, R, 7.0 - gloss * 7.0).rgb * u_EnableTextures.w; + vec3 cubeLightColor = textureCubeLod(u_CubeMap, R, 7.0 - gloss * 7.0).rgb * u_EnableTextures.w; #if defined(USE_LIGHTMAP) cubeLightColor *= lightSample.rgb; @@ -495,29 +489,40 @@ void main() #else cubeLightColor *= lightColor * NL + ambientColor; #endif - + //gl_FragColor.rgb += diffuse.rgb * textureCubeLod(u_CubeMap, N, 7.0).rgb * u_EnableTextures.w; gl_FragColor.rgb += cubeLightColor * reflectance; #endif #if defined(USE_PRIMARY_LIGHT) - L = var_PrimaryLightDir.xyz; //normalize(var_PrimaryLightDir.xyz); - NL = clamp(dot(N, L), 0.0, 1.0); + vec3 L2, H2; + float NL2, EH2, NH2; - H = normalize(L + E); - EH = clamp(dot(E, H), 0.0, 1.0); - NH = clamp(dot(N, H), 0.0, 1.0); + L2 = var_PrimaryLightDir.xyz; - reflectance = CalcDiffuse(diffuse.rgb, N, L, E, NE, NL, shininess); - reflectance += CalcSpecular(specular.rgb, NH, NL, NE, EH, gloss, shininess); + // enable when point lights are supported as primary lights + //sqrLightDist = dot(L2, L2); + //L2 /= sqrt(sqrLightDist); + + NL2 = clamp(dot(N, L2), 0.0, 1.0); + + H2 = normalize(L2 + E); + EH2 = clamp(dot(E, H2), 0.0, 1.0); + NH2 = clamp(dot(N, H2), 0.0, 1.0); + + reflectance = CalcDiffuse(diffuse.rgb, N, L2, E, NE, NL2, shininess); + reflectance += CalcSpecular(specular.rgb, NH2, NL2, NE, EH2, gloss, shininess); + + lightColor = u_PrimaryLightColor; + + // enable when point lights are supported as primary lights + //lightColor *= CalcLightAttenuation(float(u_PrimaryLightDir.w > 0.0), u_PrimaryLightDir.w / sqrLightDist); - lightColor = u_PrimaryLightColor; // * CalcLightAttenuation(L, u_PrimaryLightDir.w); - #if defined(USE_SHADOWMAP) lightColor *= shadowValue; #endif - gl_FragColor.rgb += lightColor * reflectance * NL; + gl_FragColor.rgb += lightColor * reflectance * NL2; #endif gl_FragColor.a = diffuse.a; diff --git a/code/renderergl2/glsl/lightall_vp.glsl b/code/renderergl2/glsl/lightall_vp.glsl index f6164957..2e1c899c 100644 --- a/code/renderergl2/glsl/lightall_vp.glsl +++ b/code/renderergl2/glsl/lightall_vp.glsl @@ -14,8 +14,7 @@ attribute vec4 attr_Tangent; attribute vec3 attr_Position2; attribute vec3 attr_Normal2; #if defined(USE_VERT_TANGENT_SPACE) -attribute vec3 attr_Tangent2; -attribute vec3 attr_Bitangent2; +attribute vec4 attr_Tangent2; #endif #endif @@ -108,13 +107,15 @@ vec2 GenTexCoords(int TCGen, vec3 position, vec3 normal, vec3 TCGenVector0, vec3 else if (TCGen == TCGEN_ENVIRONMENT_MAPPED) { vec3 viewer = normalize(u_LocalViewOrigin - position); - tex = -reflect(viewer, normal).yz * vec2(0.5, -0.5) + 0.5; + vec2 ref = reflect(viewer, normal).yz; + tex.s = ref.x * -0.5 + 0.5; + tex.t = ref.y * 0.5 + 0.5; } else if (TCGen == TCGEN_VECTOR) { tex = vec2(dot(position, TCGenVector0), dot(position, TCGenVector1)); } - + return tex; } #endif @@ -123,38 +124,33 @@ vec2 GenTexCoords(int TCGen, vec3 position, vec3 normal, vec3 TCGenVector0, vec3 vec2 ModTexCoords(vec2 st, vec3 position, vec4 texMatrix, vec4 offTurb) { float amplitude = offTurb.z; - float phase = offTurb.w; - vec2 st2 = vec2(dot(st, texMatrix.xz), dot(st, texMatrix.yw)) + offTurb.xy; + float phase = offTurb.w * 2.0 * M_PI; + vec2 st2; + st2.x = st.x * texMatrix.x + (st.y * texMatrix.z + offTurb.x); + st2.y = st.x * texMatrix.y + (st.y * texMatrix.w + offTurb.y); + + vec2 offsetPos = vec2(position.x + position.z, position.y); + + vec2 texOffset = sin(offsetPos * (2.0 * M_PI / 1024.0) + vec2(phase)); - vec3 offsetPos = position / 1024.0; - offsetPos.x += offsetPos.z; - - vec2 texOffset = sin((offsetPos.xy + vec2(phase)) * 2.0 * M_PI); - return st2 + texOffset * amplitude; } #endif -float CalcLightAttenuation(vec3 dir, float sqrRadius) +float CalcLightAttenuation(float point, float normDist) { - // point light at >0 radius, directional otherwise - float point = float(sqrRadius > 0.0); - - // inverse square light - float attenuation = sqrRadius / dot(dir, dir); - - // zero light at radius, approximating q3 style + // zero light at 1.0, approximating q3 style // also don't attenuate directional light - attenuation = (0.5 * attenuation - 1.5) * point + 1.0; - + float attenuation = (0.5 * normDist - 1.5) * point + 1.0; + // clamp attenuation #if defined(NO_LIGHT_CLAMP) attenuation = max(attenuation, 0.0); #else attenuation = clamp(attenuation, 0.0, 1.0); #endif - + return attenuation; } @@ -162,22 +158,22 @@ float CalcLightAttenuation(vec3 dir, float sqrRadius) void main() { #if defined(USE_VERTEX_ANIMATION) - vec3 position = mix(attr_Position, attr_Position2, u_VertexLerp); - vec3 normal = normalize(mix(attr_Normal, attr_Normal2, u_VertexLerp) * 2.0 - vec3(1.0)); + vec3 position = mix(attr_Position, attr_Position2, u_VertexLerp); + vec3 normal = mix(attr_Normal, attr_Normal2, u_VertexLerp); #if defined(USE_VERT_TANGENT_SPACE) && defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) - vec3 tangent = normalize(mix(attr_Tangent.xyz, attr_Tangent2.xyz, u_VertexLerp) * 2.0 - vec3(1.0)); + vec3 tangent = mix(attr_Tangent.xyz, attr_Tangent2.xyz, u_VertexLerp); #endif #else vec3 position = attr_Position; - vec3 normal = attr_Normal * 2.0 - vec3(1.0); + vec3 normal = attr_Normal; #if defined(USE_VERT_TANGENT_SPACE) && defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) - vec3 tangent = attr_Tangent.xyz * 2.0 - vec3(1.0); + vec3 tangent = attr_Tangent.xyz; #endif #endif + normal = normal * 2.0 - vec3(1.0); #if defined(USE_VERT_TANGENT_SPACE) && defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) - vec3 bitangent = cross(normal, tangent); - bitangent *= attr_Tangent.w * 2.0 - 1.0; + tangent = tangent * 2.0 - vec3(1.0); #endif #if defined(USE_TCGEN) @@ -195,18 +191,21 @@ void main() gl_Position = u_ModelViewProjectionMatrix * vec4(position, 1.0); #if defined(USE_MODELMATRIX) - position = (u_ModelMatrix * vec4(position, 1.0)).xyz; - normal = (u_ModelMatrix * vec4(normal, 0.0)).xyz; - #if defined(USE_VERT_TANGENT_SPACE) - tangent = (u_ModelMatrix * vec4(tangent, 0.0)).xyz; - bitangent = (u_ModelMatrix * vec4(bitangent, 0.0)).xyz; + position = (u_ModelMatrix * vec4(position, 1.0)).xyz; + normal = (u_ModelMatrix * vec4(normal, 0.0)).xyz; + #if defined(USE_VERT_TANGENT_SPACE) && defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) + tangent = (u_ModelMatrix * vec4(tangent, 0.0)).xyz; #endif #endif +#if defined(USE_VERT_TANGENT_SPACE) && defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) + vec3 bitangent = cross(normal, tangent) * (attr_Tangent.w * 2.0 - 1.0); +#endif + #if defined(USE_LIGHT_VECTOR) vec3 L = u_LightOrigin.xyz - (position * u_LightOrigin.w); -#elif defined(USE_LIGHT) - vec3 L = attr_LightDirection; +#elif defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) + vec3 L = attr_LightDirection * 2.0 - vec3(1.0); #if defined(USE_MODELMATRIX) L = (u_ModelMatrix * vec4(L, 0.0)).xyz; #endif @@ -215,7 +214,7 @@ void main() #if defined(USE_LIGHTMAP) var_TexCoords.zw = attr_TexCoord1.st; #endif - + var_Color = u_VertColor * attr_Color + u_BaseColor; #if defined(USE_LIGHT_VERTEX) && !defined(USE_FAST_LIGHT) var_LightColor = var_Color.rgb; @@ -223,10 +222,11 @@ void main() #endif #if defined(USE_LIGHT_VECTOR) && defined(USE_FAST_LIGHT) - float attenuation = CalcLightAttenuation(L, u_LightRadius * u_LightRadius); - float NL = clamp(dot(normal, normalize(L)), 0.0, 1.0); + float sqrLightDist = dot(L, L); + float attenuation = CalcLightAttenuation(u_LightOrigin.w, u_LightRadius * u_LightRadius / sqrLightDist); + float NL = clamp(dot(normalize(normal), L) / sqrt(sqrLightDist), 0.0, 1.0); - var_Color.rgb *= u_DirectedLight * attenuation * NL + u_AmbientLight; + var_Color.rgb *= u_DirectedLight * (attenuation * NL) + u_AmbientLight; #endif #if defined(USE_PRIMARY_LIGHT) || defined(USE_SHADOWMAP) @@ -241,15 +241,12 @@ void main() var_LightDir = vec4(L, 0.0); #endif #if defined(USE_DELUXEMAP) - var_LightDir *= 1.0 - u_EnableTextures.y; + var_LightDir -= u_EnableTextures.y * var_LightDir; #endif #endif #if defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) vec3 viewDir = u_ViewOrigin - position; -#endif - -#if defined(USE_LIGHT) && !defined(USE_FAST_LIGHT) #if defined(USE_VERT_TANGENT_SPACE) // store view direction in tangent space to save on varyings var_Normal = vec4(normal, viewDir.x); diff --git a/code/renderergl2/glsl/pshadow_vp.glsl b/code/renderergl2/glsl/pshadow_vp.glsl index 0e0e3b3d..0f9940cd 100644 --- a/code/renderergl2/glsl/pshadow_vp.glsl +++ b/code/renderergl2/glsl/pshadow_vp.glsl @@ -1,4 +1,4 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec3 attr_Normal; uniform mat4 u_ModelViewProjectionMatrix; @@ -8,10 +8,8 @@ varying vec3 var_Normal; void main() { - vec4 position = attr_Position; + gl_Position = u_ModelViewProjectionMatrix * vec4(attr_Position, 1.0); - gl_Position = u_ModelViewProjectionMatrix * position; - - var_Position = position.xyz; - var_Normal = attr_Normal; + var_Position = attr_Position; + var_Normal = attr_Normal * 2.0 - vec3(1.0); } diff --git a/code/renderergl2/glsl/shadowfill_vp.glsl b/code/renderergl2/glsl/shadowfill_vp.glsl index 10802eca..7a5cc571 100644 --- a/code/renderergl2/glsl/shadowfill_vp.glsl +++ b/code/renderergl2/glsl/shadowfill_vp.glsl @@ -1,9 +1,9 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec3 attr_Normal; attribute vec4 attr_TexCoord0; //#if defined(USE_VERTEX_ANIMATION) -attribute vec4 attr_Position2; +attribute vec3 attr_Position2; attribute vec3 attr_Normal2; //#endif @@ -38,7 +38,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) if (u_DeformGen == DGEN_BULGE) { - phase *= M_PI * 0.25 * st.x; + phase *= st.x; } else // if (u_DeformGen <= DGEN_WAVE_INVERSE_SAWTOOTH) { @@ -54,7 +54,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) } else if (u_DeformGen == DGEN_WAVE_SQUARE) { - func = sign(sin(value * 2.0 * M_PI)); + func = sign(0.5 - fract(value)); } else if (u_DeformGen == DGEN_WAVE_TRIANGLE) { @@ -68,7 +68,7 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) { func = (1.0 - fract(value)); } - else if (u_DeformGen == DGEN_BULGE) + else // if (u_DeformGen == DGEN_BULGE) { func = sin(value); } @@ -78,12 +78,13 @@ vec3 DeformPosition(const vec3 pos, const vec3 normal, const vec2 st) void main() { - vec4 position = mix(attr_Position, attr_Position2, u_VertexLerp); - vec3 normal = normalize(mix(attr_Normal, attr_Normal2, u_VertexLerp)); + vec3 position = mix(attr_Position, attr_Position2, u_VertexLerp); + vec3 normal = mix(attr_Normal, attr_Normal2, u_VertexLerp); + normal = normalize(normal - vec3(0.5)); - position.xyz = DeformPosition(position.xyz, normal, attr_TexCoord0.st); + position = DeformPosition(position, normal, attr_TexCoord0.st); - gl_Position = u_ModelViewProjectionMatrix * position; + gl_Position = u_ModelViewProjectionMatrix * vec4(position, 1.0); - var_Position = (u_ModelMatrix * position).xyz; + var_Position = (u_ModelMatrix * vec4(position, 1.0)).xyz; } diff --git a/code/renderergl2/glsl/texturecolor_fp.glsl b/code/renderergl2/glsl/texturecolor_fp.glsl index 5646b511..e077e7da 100644 --- a/code/renderergl2/glsl/texturecolor_fp.glsl +++ b/code/renderergl2/glsl/texturecolor_fp.glsl @@ -3,7 +3,7 @@ uniform sampler2D u_DiffuseMap; uniform vec4 u_Color; -varying vec2 var_Tex1; +varying vec2 var_Tex1; void main() diff --git a/code/renderergl2/glsl/texturecolor_vp.glsl b/code/renderergl2/glsl/texturecolor_vp.glsl index ae26a18e..7a5750a5 100644 --- a/code/renderergl2/glsl/texturecolor_vp.glsl +++ b/code/renderergl2/glsl/texturecolor_vp.glsl @@ -1,6 +1,6 @@ #version 120 -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec4 attr_TexCoord0; uniform mat4 u_ModelViewProjectionMatrix; @@ -10,6 +10,6 @@ varying vec2 var_Tex1; void main() { - gl_Position = u_ModelViewProjectionMatrix * attr_Position; + gl_Position = u_ModelViewProjectionMatrix * vec4(attr_Position, 1.0); var_Tex1 = attr_TexCoord0.st; } diff --git a/code/renderergl2/glsl/tonemap_vp.glsl b/code/renderergl2/glsl/tonemap_vp.glsl index 5ca41600..bdaa74af 100644 --- a/code/renderergl2/glsl/tonemap_vp.glsl +++ b/code/renderergl2/glsl/tonemap_vp.glsl @@ -1,4 +1,4 @@ -attribute vec4 attr_Position; +attribute vec3 attr_Position; attribute vec4 attr_TexCoord0; uniform mat4 u_ModelViewProjectionMatrix; @@ -8,6 +8,6 @@ varying vec2 var_TexCoords; void main() { - gl_Position = u_ModelViewProjectionMatrix * attr_Position; + gl_Position = u_ModelViewProjectionMatrix * vec4(attr_Position, 1.0); var_TexCoords = attr_TexCoord0.st; } diff --git a/code/renderergl2/tr_animation.c b/code/renderergl2/tr_animation.c index f3329ee4..a5f08986 100644 --- a/code/renderergl2/tr_animation.c +++ b/code/renderergl2/tr_animation.c @@ -33,143 +33,6 @@ frame. */ -/* -============== -R_AddAnimSurfaces -============== -*/ -void R_AddAnimSurfaces( trRefEntity_t *ent ) { - md4Header_t *header; - md4Surface_t *surface; - md4LOD_t *lod; - shader_t *shader; - int cubemapIndex; - int i; - - header = (md4Header_t *) tr.currentModel->modelData; - lod = (md4LOD_t *)( (byte *)header + header->ofsLODs ); - cubemapIndex = R_CubemapForPoint(ent->e.origin); - - surface = (md4Surface_t *)( (byte *)lod + lod->ofsSurfaces ); - for ( i = 0 ; i < lod->numSurfaces ; i++ ) { - shader = R_GetShaderByHandle( surface->shaderIndex ); - R_AddDrawSurf( (void *)surface, shader, 0 /*fogNum*/, qfalse, qfalse, cubemapIndex ); - surface = (md4Surface_t *)( (byte *)surface + surface->ofsEnd ); - } -} - -/* -============== -RB_SurfaceAnim -============== -*/ -void RB_SurfaceAnim( md4Surface_t *surface ) { - int i, j, k; - float frontlerp, backlerp; - int *triangles; - int indexes; - int baseIndex, baseVertex; - int numVerts; - md4Vertex_t *v; - md4Bone_t bones[MD4_MAX_BONES]; - md4Bone_t *bonePtr, *bone; - md4Header_t *header; - md4Frame_t *frame; - md4Frame_t *oldFrame; - int frameSize; - - - if ( backEnd.currentEntity->e.oldframe == backEnd.currentEntity->e.frame ) { - backlerp = 0; - frontlerp = 1; - } else { - backlerp = backEnd.currentEntity->e.backlerp; - frontlerp = 1.0f - backlerp; - } - header = (md4Header_t *)((byte *)surface + surface->ofsHeader); - - frameSize = (size_t)( &((md4Frame_t *)0)->bones[ header->numBones ] ); - - frame = (md4Frame_t *)((byte *)header + header->ofsFrames + - backEnd.currentEntity->e.frame * frameSize ); - oldFrame = (md4Frame_t *)((byte *)header + header->ofsFrames + - backEnd.currentEntity->e.oldframe * frameSize ); - - RB_CheckOverflow( surface->numVerts, surface->numTriangles * 3 ); - - triangles = (int *) ((byte *)surface + surface->ofsTriangles); - indexes = surface->numTriangles * 3; - baseIndex = tess.numIndexes; - baseVertex = tess.numVertexes; - for (j = 0 ; j < indexes ; j++) { - tess.indexes[baseIndex + j] = baseIndex + triangles[j]; - } - tess.numIndexes += indexes; - - // - // lerp all the needed bones - // - if ( !backlerp ) { - // no lerping needed - bonePtr = frame->bones; - } else { - bonePtr = bones; - for ( i = 0 ; i < header->numBones*12 ; i++ ) { - ((float *)bonePtr)[i] = frontlerp * ((float *)frame->bones)[i] - + backlerp * ((float *)oldFrame->bones)[i]; - } - } - - // - // deform the vertexes by the lerped bones - // - numVerts = surface->numVerts; - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *) ((byte *)surface + surface->ofsVerts + 12); - v = (md4Vertex_t *) ((byte *)surface + surface->ofsVerts); - for ( j = 0; j < numVerts; j++ ) { - vec3_t tempVert, tempNormal; - md4Weight_t *w; - - VectorClear( tempVert ); - VectorClear( tempNormal ); - w = v->weights; - for ( k = 0 ; k < v->numWeights ; k++, w++ ) { - bone = bonePtr + w->boneIndex; - - tempVert[0] += w->boneWeight * ( DotProduct( bone->matrix[0], w->offset ) + bone->matrix[0][3] ); - tempVert[1] += w->boneWeight * ( DotProduct( bone->matrix[1], w->offset ) + bone->matrix[1][3] ); - tempVert[2] += w->boneWeight * ( DotProduct( bone->matrix[2], w->offset ) + bone->matrix[2][3] ); - - tempNormal[0] += w->boneWeight * DotProduct( bone->matrix[0], v->normal ); - tempNormal[1] += w->boneWeight * DotProduct( bone->matrix[1], v->normal ); - tempNormal[2] += w->boneWeight * DotProduct( bone->matrix[2], v->normal ); - } - - tess.xyz[baseVertex + j][0] = tempVert[0]; - tess.xyz[baseVertex + j][1] = tempVert[1]; - tess.xyz[baseVertex + j][2] = tempVert[2]; - - tess.normal[baseVertex + j][0] = (uint8_t)(tempNormal[0] * 127.5f + 128.0f); - tess.normal[baseVertex + j][1] = (uint8_t)(tempNormal[1] * 127.5f + 128.0f); - tess.normal[baseVertex + j][2] = (uint8_t)(tempNormal[2] * 127.5f + 128.0f); - - tess.texCoords[baseVertex + j][0][0] = v->texCoords[0]; - tess.texCoords[baseVertex + j][0][1] = v->texCoords[1]; - - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *)( ( byte * )&v->weights[v->numWeights] + 12 ); - v = (md4Vertex_t *)&v->weights[v->numWeights]; - } - - tess.numVertexes += surface->numVerts; -} - - // copied and adapted from tr_mesh.c @@ -198,7 +61,7 @@ static int R_MDRCullModel( mdrHeader_t *header, trRefEntity_t *ent ) { switch ( R_CullLocalPointAndRadius( newFrame->localOrigin, newFrame->radius ) ) { // Ummm... yeah yeah I know we don't really have an md3 here.. but we pretend - // we do. After all, the purpose of md4s are not that different, are they? + // we do. After all, the purpose of mdrs are not that different, are they? case CULL_OUT: tr.pc.c_sphere_cull_md3_out++; @@ -448,7 +311,7 @@ void R_MDRAddAnimSurfaces( trRefEntity_t *ent ) { RB_MDRSurfaceAnim ============== */ -void RB_MDRSurfaceAnim( md4Surface_t *surface ) +void RB_MDRSurfaceAnim( mdrSurface_t *surface ) { int i, j, k; float frontlerp, backlerp; @@ -460,7 +323,7 @@ void RB_MDRSurfaceAnim( md4Surface_t *surface ) mdrHeader_t *header; mdrFrame_t *frame; mdrFrame_t *oldFrame; - mdrBone_t bones[MD4_MAX_BONES], *bonePtr, *bone; + mdrBone_t bones[MDR_MAX_BONES], *bonePtr, *bone; int frameSize; @@ -548,9 +411,7 @@ void RB_MDRSurfaceAnim( md4Surface_t *surface ) tess.xyz[baseVertex + j][1] = tempVert[1]; tess.xyz[baseVertex + j][2] = tempVert[2]; - tess.normal[baseVertex + j][0] = (uint8_t)(tempNormal[0] * 127.5f + 128.0f); - tess.normal[baseVertex + j][1] = (uint8_t)(tempNormal[1] * 127.5f + 128.0f); - tess.normal[baseVertex + j][2] = (uint8_t)(tempNormal[2] * 127.5f + 128.0f); + tess.normal[baseVertex + j] = R_VboPackNormal(tempNormal); tess.texCoords[baseVertex + j][0][0] = v->texCoords[0]; tess.texCoords[baseVertex + j][0][1] = v->texCoords[1]; diff --git a/code/renderergl2/tr_backend.c b/code/renderergl2/tr_backend.c index a050379a..e4431cf5 100644 --- a/code/renderergl2/tr_backend.c +++ b/code/renderergl2/tr_backend.c @@ -461,7 +461,7 @@ void RB_BeginDrawingView (void) { FBO_Bind(backEnd.viewParms.targetFbo); // FIXME: hack for cubemap testing - if (backEnd.viewParms.targetFbo == tr.renderCubeFbo) + if (tr.renderCubeFbo && backEnd.viewParms.targetFbo == tr.renderCubeFbo) { //qglFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT, GL_TEXTURE_CUBE_MAP_POSITIVE_X + backEnd.viewParms.targetFboLayer, backEnd.viewParms.targetFbo->colorImage[0]->texnum, 0); qglFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT, GL_TEXTURE_CUBE_MAP_POSITIVE_X + backEnd.viewParms.targetFboLayer, tr.cubemaps[backEnd.viewParms.targetFboCubemapIndex]->texnum, 0); @@ -501,7 +501,7 @@ void RB_BeginDrawingView (void) { } // clear to black for cube maps - if (backEnd.viewParms.targetFbo == tr.renderCubeFbo) + if (tr.renderCubeFbo && backEnd.viewParms.targetFbo == tr.renderCubeFbo) { clearBits |= GL_COLOR_BUFFER_BIT; qglClearColor( 0.0f, 0.0f, 0.0f, 1.0f ); @@ -1081,7 +1081,7 @@ const void *RB_DrawSurfs( const void *data ) { FBO_BlitFromTexture(tr.renderDepthImage, NULL, NULL, tr.hdrDepthFbo, NULL, NULL, NULL, 0); } - if (backEnd.viewParms.flags & VPF_USESUNLIGHT) + if (r_sunlightMode->integer && backEnd.viewParms.flags & VPF_USESUNLIGHT) { vec4_t quadVerts[4]; vec2_t texCoords[4]; @@ -1296,7 +1296,7 @@ const void *RB_DrawSurfs( const void *data ) { RB_RenderFlares(); } - if (glRefConfig.framebufferObject && backEnd.viewParms.targetFbo == tr.renderCubeFbo) + if (glRefConfig.framebufferObject && tr.renderCubeFbo && backEnd.viewParms.targetFbo == tr.renderCubeFbo) { FBO_Bind(NULL); GL_SelectTexture(TB_CUBEMAP); @@ -1558,13 +1558,19 @@ const void *RB_CapShadowMap(const void *data) GL_SelectTexture(0); if (cmd->cubeSide != -1) { - GL_Bind(tr.shadowCubemaps[cmd->map]); - qglCopyTexImage2D(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cmd->cubeSide, 0, GL_RGBA8, backEnd.refdef.x, glConfig.vidHeight - ( backEnd.refdef.y + PSHADOW_MAP_SIZE ), PSHADOW_MAP_SIZE, PSHADOW_MAP_SIZE, 0); + if (tr.shadowCubemaps[cmd->map]) + { + GL_Bind(tr.shadowCubemaps[cmd->map]); + qglCopyTexImage2D(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cmd->cubeSide, 0, GL_RGBA8, backEnd.refdef.x, glConfig.vidHeight - ( backEnd.refdef.y + PSHADOW_MAP_SIZE ), PSHADOW_MAP_SIZE, PSHADOW_MAP_SIZE, 0); + } } else { - GL_Bind(tr.pshadowMaps[cmd->map]); - qglCopyTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, backEnd.refdef.x, glConfig.vidHeight - ( backEnd.refdef.y + PSHADOW_MAP_SIZE ), PSHADOW_MAP_SIZE, PSHADOW_MAP_SIZE, 0); + if (tr.pshadowMaps[cmd->map]) + { + GL_Bind(tr.pshadowMaps[cmd->map]); + qglCopyTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, backEnd.refdef.x, glConfig.vidHeight - ( backEnd.refdef.y + PSHADOW_MAP_SIZE ), PSHADOW_MAP_SIZE, PSHADOW_MAP_SIZE, 0); + } } } @@ -1666,7 +1672,7 @@ const void *RB_PostProcess(const void *data) else RB_GaussianBlur(backEnd.refdef.blurFactor); - if (0) + if (0 && r_sunlightMode->integer) { ivec4_t dstBox; VectorSet4(dstBox, 0, 0, 128, 128); diff --git a/code/renderergl2/tr_bsp.c b/code/renderergl2/tr_bsp.c index 2947e8ff..3413ff2d 100644 --- a/code/renderergl2/tr_bsp.c +++ b/code/renderergl2/tr_bsp.c @@ -128,17 +128,34 @@ static void R_ColorShiftLightingBytes( byte in[4], byte out[4] ) { /* =============== -R_ColorShiftLightingBytes +R_ColorShiftLightingFloats =============== */ static void R_ColorShiftLightingFloats(float in[4], float out[4], float scale ) { + float r, g, b; + scale *= pow(2.0f, r_mapOverBrightBits->integer - tr.overbrightBits); - out[0] = in[0] * scale; - out[1] = in[1] * scale; - out[2] = in[2] * scale; + r = in[0] * scale; + g = in[1] * scale; + b = in[2] * scale; + + // normalize by color instead of saturating to white + if ( !r_hdr->integer && ( r > 1 || g > 1 || b > 1 ) ) { + float max; + + max = r > g ? r : g; + max = max > b ? max : b; + r = r / max; + g = g / max; + b = b / max; + } + + out[0] = r; + out[1] = g; + out[2] = b; out[3] = in[3]; } @@ -471,7 +488,7 @@ static void R_LoadLightmaps( lump_t *l, lump_t *surfs ) { image[j*4+2] = buf_p[j*3+2]; // make 0,0,0 into 127,127,127 - if ((image[j*4+0] == 0) && (image[j*4+0] == 0) && (image[j*4+2] == 0)) + if ((image[j*4+0] == 0) && (image[j*4+1] == 0) && (image[j*4+2] == 0)) { image[j*4+0] = image[j*4+1] = diff --git a/code/renderergl2/tr_extensions.c b/code/renderergl2/tr_extensions.c index 5cd5c2a5..d7e5f1ba 100644 --- a/code/renderergl2/tr_extensions.c +++ b/code/renderergl2/tr_extensions.c @@ -710,4 +710,19 @@ void GLimp_InitExtraExtensions() { ri.Printf(PRINT_ALL, result[2], extension); } + + // GL_ARB_vertex_type_2_10_10_10_rev + extension = "GL_ARB_vertex_type_2_10_10_10_rev"; + glRefConfig.packedNormalDataType = GL_UNSIGNED_BYTE; + if( GLimp_HaveExtension( extension ) ) + { + if (r_arb_vertex_type_2_10_10_10_rev->integer) + glRefConfig.packedNormalDataType = GL_UNSIGNED_INT_2_10_10_10_REV; + + ri.Printf(PRINT_ALL, result[r_arb_vertex_type_2_10_10_10_rev->integer ? 1 : 0], extension); + } + else + { + ri.Printf(PRINT_ALL, result[2], extension); + } } diff --git a/code/renderergl2/tr_extramath.h b/code/renderergl2/tr_extramath.h index 08fb080a..0223c640 100644 --- a/code/renderergl2/tr_extramath.h +++ b/code/renderergl2/tr_extramath.h @@ -52,7 +52,7 @@ void Mat4SimpleInverse( const mat4_t in, mat4_t out); #define VectorCopy5(a,b) ((b)[0]=(a)[0],(b)[1]=(a)[1],(b)[2]=(a)[2],(b)[3]=(a)[3],(b)[4]=(a)[4]) #define OffsetByteToFloat(a) ((float)(a) * 1.0f/127.5f - 1.0f) -#define FloatToOffsetByte(a) (byte)(((a) + 1.0f) * 127.5f) +#define FloatToOffsetByte(a) (byte)((a) * 127.5f + 128.0f) #define ByteToFloat(a) ((float)(a) * 1.0f/255.0f) #define FloatToByte(a) (byte)((a) * 255.0f) diff --git a/code/renderergl2/tr_fbo.c b/code/renderergl2/tr_fbo.c index c76e7db1..c642b732 100644 --- a/code/renderergl2/tr_fbo.c +++ b/code/renderergl2/tr_fbo.c @@ -465,34 +465,48 @@ void FBO_Init(void) } // FIXME: Don't use separate color/depth buffers for a shadow buffer - for( i = 0; i < MAX_DRAWN_PSHADOWS; i++) + if (MAX_DRAWN_PSHADOWS && tr.pshadowMaps[0]) { - tr.pshadowFbos[i] = FBO_Create(va("_shadowmap%d", i), tr.pshadowMaps[i]->width, tr.pshadowMaps[i]->height); - FBO_Bind(tr.pshadowFbos[i]); + for( i = 0; i < MAX_DRAWN_PSHADOWS; i++) + { + tr.pshadowFbos[i] = FBO_Create(va("_shadowmap%d", i), tr.pshadowMaps[i]->width, tr.pshadowMaps[i]->height); + FBO_Bind(tr.pshadowFbos[i]); - //FBO_CreateBuffer(tr.pshadowFbos[i], GL_RGBA8, 0, 0); - FBO_AttachTextureImage(tr.pshadowMaps[i], 0); + //FBO_CreateBuffer(tr.pshadowFbos[i], GL_RGBA8, 0, 0); + FBO_AttachTextureImage(tr.pshadowMaps[i], 0); - FBO_CreateBuffer(tr.pshadowFbos[i], GL_DEPTH_COMPONENT24_ARB, 0, 0); - //R_AttachFBOTextureDepth(tr.textureDepthImage->texnum); + FBO_CreateBuffer(tr.pshadowFbos[i], GL_DEPTH_COMPONENT24_ARB, 0, 0); + //R_AttachFBOTextureDepth(tr.textureDepthImage->texnum); - R_CheckFBO(tr.pshadowFbos[i]); + R_CheckFBO(tr.pshadowFbos[i]); + } } - for ( i = 0; i < 3; i++) + if (tr.sunShadowDepthImage[0]) { - tr.sunShadowFbo[i] = FBO_Create("_sunshadowmap", tr.sunShadowDepthImage[i]->width, tr.sunShadowDepthImage[i]->height); - FBO_Bind(tr.sunShadowFbo[i]); + for ( i = 0; i < 3; i++) + { + tr.sunShadowFbo[i] = FBO_Create("_sunshadowmap", tr.sunShadowDepthImage[i]->width, tr.sunShadowDepthImage[i]->height); + FBO_Bind(tr.sunShadowFbo[i]); - //FBO_CreateBuffer(tr.sunShadowFbo[i], GL_RGBA8, 0, 0); - //FBO_AttachTextureImage(tr.sunShadowImage, 0); - qglDrawBuffer(GL_NONE); - qglReadBuffer(GL_NONE); + //FBO_CreateBuffer(tr.sunShadowFbo[i], GL_RGBA8, 0, 0); + //FBO_AttachTextureImage(tr.sunShadowImage, 0); + qglDrawBuffer(GL_NONE); + qglReadBuffer(GL_NONE); - //FBO_CreateBuffer(tr.sunShadowFbo, GL_DEPTH_COMPONENT24_ARB, 0, 0); - R_AttachFBOTextureDepth(tr.sunShadowDepthImage[i]->texnum); + //FBO_CreateBuffer(tr.sunShadowFbo, GL_DEPTH_COMPONENT24_ARB, 0, 0); + R_AttachFBOTextureDepth(tr.sunShadowDepthImage[i]->texnum); - R_CheckFBO(tr.sunShadowFbo[i]); + R_CheckFBO(tr.sunShadowFbo[i]); + + } + + tr.screenShadowFbo = FBO_Create("_screenshadow", tr.screenShadowImage->width, tr.screenShadowImage->height); + FBO_Bind(tr.screenShadowFbo); + + FBO_AttachTextureImage(tr.screenShadowImage, 0); + + R_CheckFBO(tr.screenShadowFbo); } for (i = 0; i < 2; i++) @@ -537,15 +551,6 @@ void FBO_Init(void) R_CheckFBO(tr.quarterFbo[i]); } - { - tr.screenShadowFbo = FBO_Create("_screenshadow", tr.screenShadowImage->width, tr.screenShadowImage->height); - FBO_Bind(tr.screenShadowFbo); - - FBO_AttachTextureImage(tr.screenShadowImage, 0); - - R_CheckFBO(tr.screenShadowFbo); - } - if (r_ssao->integer) { tr.hdrDepthFbo = FBO_Create("_hdrDepth", tr.hdrDepthImage->width, tr.hdrDepthImage->height); @@ -563,6 +568,7 @@ void FBO_Init(void) R_CheckFBO(tr.screenSsaoFbo); } + if (tr.renderCubeImage) { tr.renderCubeFbo = FBO_Create("_renderCubeFbo", tr.renderCubeImage->width, tr.renderCubeImage->height); FBO_Bind(tr.renderCubeFbo); diff --git a/code/renderergl2/tr_glsl.c b/code/renderergl2/tr_glsl.c index 42ce225e..2212c4fe 100644 --- a/code/renderergl2/tr_glsl.c +++ b/code/renderergl2/tr_glsl.c @@ -1647,7 +1647,10 @@ void GLSL_VertexAttribPointers(uint32_t attribBits) } // don't just call LogComment, or we will get a call to va() every frame! - GLimp_LogComment(va("--- GL_VertexAttribPointers( %s ) ---\n", vbo->name)); + if(r_logFile->integer) + { + GLimp_LogComment(va("--- GL_VertexAttribPointers( %s ) ---\n", vbo->name)); + } // position/normal/tangent are always set in case of animation oldFrame = glState.vertexAttribsOldFrame; @@ -1682,7 +1685,7 @@ void GLSL_VertexAttribPointers(uint32_t attribBits) { GLimp_LogComment("qglVertexAttribPointerARB( ATTR_INDEX_NORMAL )\n"); - qglVertexAttribPointerARB(ATTR_INDEX_NORMAL, 3, GL_UNSIGNED_BYTE, GL_TRUE, vbo->stride_normal, BUFFER_OFFSET(vbo->ofs_normal + newFrame * vbo->size_normal)); + qglVertexAttribPointerARB(ATTR_INDEX_NORMAL, 4, glRefConfig.packedNormalDataType, GL_TRUE, vbo->stride_normal, BUFFER_OFFSET(vbo->ofs_normal + newFrame * vbo->size_normal)); glState.vertexAttribPointersSet |= ATTR_NORMAL; } @@ -1691,7 +1694,7 @@ void GLSL_VertexAttribPointers(uint32_t attribBits) { GLimp_LogComment("qglVertexAttribPointerARB( ATTR_INDEX_TANGENT )\n"); - qglVertexAttribPointerARB(ATTR_INDEX_TANGENT, 4, GL_UNSIGNED_BYTE, GL_TRUE, vbo->stride_tangent, BUFFER_OFFSET(vbo->ofs_tangent + newFrame * vbo->size_normal)); // FIXME + qglVertexAttribPointerARB(ATTR_INDEX_TANGENT, 4, glRefConfig.packedNormalDataType, GL_TRUE, vbo->stride_tangent, BUFFER_OFFSET(vbo->ofs_tangent + newFrame * vbo->size_normal)); // FIXME glState.vertexAttribPointersSet |= ATTR_TANGENT; } #endif @@ -1708,7 +1711,7 @@ void GLSL_VertexAttribPointers(uint32_t attribBits) { GLimp_LogComment("qglVertexAttribPointerARB( ATTR_INDEX_LIGHTDIRECTION )\n"); - qglVertexAttribPointerARB(ATTR_INDEX_LIGHTDIRECTION, 3, GL_FLOAT, 0, vbo->stride_lightdir, BUFFER_OFFSET(vbo->ofs_lightdir)); + qglVertexAttribPointerARB(ATTR_INDEX_LIGHTDIRECTION, 4, glRefConfig.packedNormalDataType, GL_TRUE, vbo->stride_lightdir, BUFFER_OFFSET(vbo->ofs_lightdir)); glState.vertexAttribPointersSet |= ATTR_LIGHTDIRECTION; } @@ -1724,7 +1727,7 @@ void GLSL_VertexAttribPointers(uint32_t attribBits) { GLimp_LogComment("qglVertexAttribPointerARB( ATTR_INDEX_NORMAL2 )\n"); - qglVertexAttribPointerARB(ATTR_INDEX_NORMAL2, 3, GL_UNSIGNED_BYTE, GL_TRUE, vbo->stride_normal, BUFFER_OFFSET(vbo->ofs_normal + oldFrame * vbo->size_normal)); + qglVertexAttribPointerARB(ATTR_INDEX_NORMAL2, 4, glRefConfig.packedNormalDataType, GL_TRUE, vbo->stride_normal, BUFFER_OFFSET(vbo->ofs_normal + oldFrame * vbo->size_normal)); glState.vertexAttribPointersSet |= ATTR_NORMAL2; } @@ -1733,7 +1736,7 @@ void GLSL_VertexAttribPointers(uint32_t attribBits) { GLimp_LogComment("qglVertexAttribPointerARB( ATTR_INDEX_TANGENT2 )\n"); - qglVertexAttribPointerARB(ATTR_INDEX_TANGENT2, 4, GL_UNSIGNED_BYTE, GL_TRUE, vbo->stride_tangent, BUFFER_OFFSET(vbo->ofs_tangent + oldFrame * vbo->size_normal)); // FIXME + qglVertexAttribPointerARB(ATTR_INDEX_TANGENT2, 4, glRefConfig.packedNormalDataType, GL_TRUE, vbo->stride_tangent, BUFFER_OFFSET(vbo->ofs_tangent + oldFrame * vbo->size_normal)); // FIXME glState.vertexAttribPointersSet |= ATTR_TANGENT2; } #endif diff --git a/code/renderergl2/tr_image.c b/code/renderergl2/tr_image.c index 670b4499..243d27df 100644 --- a/code/renderergl2/tr_image.c +++ b/code/renderergl2/tr_image.c @@ -2972,25 +2972,34 @@ void R_CreateBuiltinImages( void ) { tr.quarterImage[x] = R_CreateImage(va("*quarter%d", x), NULL, width / 2, height / 2, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_RGBA8); } - tr.screenShadowImage = R_CreateImage("*screenShadow", NULL, width, height, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_RGBA8); - if (r_ssao->integer) { tr.screenSsaoImage = R_CreateImage("*screenSsao", NULL, width / 2, height / 2, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_RGBA8); tr.hdrDepthImage = R_CreateImage("*hdrDepth", NULL, width, height, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_INTENSITY32F_ARB); } - for( x = 0; x < MAX_DRAWN_PSHADOWS; x++) + if (r_shadows->integer == 4) { - tr.pshadowMaps[x] = R_CreateImage(va("*shadowmap%i", x), NULL, PSHADOW_MAP_SIZE, PSHADOW_MAP_SIZE, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_RGBA8); + for( x = 0; x < MAX_DRAWN_PSHADOWS; x++) + { + tr.pshadowMaps[x] = R_CreateImage(va("*shadowmap%i", x), NULL, PSHADOW_MAP_SIZE, PSHADOW_MAP_SIZE, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_RGBA8); + } } - for ( x = 0; x < 3; x++) + if (r_sunlightMode->integer) { - tr.sunShadowDepthImage[x] = R_CreateImage(va("*sunshadowdepth%i", x), NULL, r_shadowMapSize->integer, r_shadowMapSize->integer, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_DEPTH_COMPONENT24_ARB); + for ( x = 0; x < 3; x++) + { + tr.sunShadowDepthImage[x] = R_CreateImage(va("*sunshadowdepth%i", x), NULL, r_shadowMapSize->integer, r_shadowMapSize->integer, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_DEPTH_COMPONENT24_ARB); + } + + tr.screenShadowImage = R_CreateImage("*screenShadow", NULL, width, height, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE, GL_RGBA8); } - tr.renderCubeImage = R_CreateImage("*renderCube", NULL, CUBE_MAP_SIZE, CUBE_MAP_SIZE, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE | IMGFLAG_MIPMAP | IMGFLAG_CUBEMAP, rgbFormat); + if (r_cubeMapping->integer) + { + tr.renderCubeImage = R_CreateImage("*renderCube", NULL, CUBE_MAP_SIZE, CUBE_MAP_SIZE, IMGTYPE_COLORALPHA, IMGFLAG_NO_COMPRESSION | IMGFLAG_CLAMPTOEDGE | IMGFLAG_MIPMAP | IMGFLAG_CUBEMAP, rgbFormat); + } } } diff --git a/code/renderergl2/tr_init.c b/code/renderergl2/tr_init.c index dc67e3ee..f650c307 100644 --- a/code/renderergl2/tr_init.c +++ b/code/renderergl2/tr_init.c @@ -103,6 +103,7 @@ cvar_t *r_ext_texture_float; cvar_t *r_arb_half_float_pixel; cvar_t *r_ext_framebuffer_multisample; cvar_t *r_arb_seamless_cube_map; +cvar_t *r_arb_vertex_type_2_10_10_10_rev; cvar_t *r_mergeMultidraws; cvar_t *r_mergeLeafSurfaces; @@ -160,6 +161,7 @@ cvar_t *r_shadowMapSize; cvar_t *r_shadowCascadeZNear; cvar_t *r_shadowCascadeZFar; cvar_t *r_shadowCascadeZBias; +cvar_t *r_ignoreDstAlpha; cvar_t *r_ignoreGLErrors; cvar_t *r_logFile; @@ -1128,6 +1130,7 @@ void R_Register( void ) r_arb_half_float_pixel = ri.Cvar_Get( "r_arb_half_float_pixel", "1", CVAR_ARCHIVE | CVAR_LATCH); r_ext_framebuffer_multisample = ri.Cvar_Get( "r_ext_framebuffer_multisample", "0", CVAR_ARCHIVE | CVAR_LATCH); r_arb_seamless_cube_map = ri.Cvar_Get( "r_arb_seamless_cube_map", "0", CVAR_ARCHIVE | CVAR_LATCH); + r_arb_vertex_type_2_10_10_10_rev = ri.Cvar_Get( "r_arb_vertex_type_2_10_10_10_rev", "1", CVAR_ARCHIVE | CVAR_LATCH); r_ext_texture_filter_anisotropic = ri.Cvar_Get( "r_ext_texture_filter_anisotropic", "0", CVAR_ARCHIVE | CVAR_LATCH ); @@ -1148,7 +1151,7 @@ void R_Register( void ) r_ignorehwgamma = ri.Cvar_Get( "r_ignorehwgamma", "0", CVAR_ARCHIVE | CVAR_LATCH); r_mode = ri.Cvar_Get( "r_mode", "-2", CVAR_ARCHIVE | CVAR_LATCH ); r_fullscreen = ri.Cvar_Get( "r_fullscreen", "1", CVAR_ARCHIVE ); - r_noborder = ri.Cvar_Get("r_noborder", "0", CVAR_ARCHIVE); + r_noborder = ri.Cvar_Get("r_noborder", "0", CVAR_ARCHIVE | CVAR_LATCH); r_customwidth = ri.Cvar_Get( "r_customwidth", "1600", CVAR_ARCHIVE | CVAR_LATCH ); r_customheight = ri.Cvar_Get( "r_customheight", "1024", CVAR_ARCHIVE | CVAR_LATCH ); r_customPixelAspect = ri.Cvar_Get( "r_customPixelAspect", "1", CVAR_ARCHIVE | CVAR_LATCH ); @@ -1215,6 +1218,7 @@ void R_Register( void ) r_shadowCascadeZNear = ri.Cvar_Get( "r_shadowCascadeZNear", "4", CVAR_ARCHIVE | CVAR_LATCH ); r_shadowCascadeZFar = ri.Cvar_Get( "r_shadowCascadeZFar", "3072", CVAR_ARCHIVE | CVAR_LATCH ); r_shadowCascadeZBias = ri.Cvar_Get( "r_shadowCascadeZBias", "-320", CVAR_ARCHIVE | CVAR_LATCH ); + r_ignoreDstAlpha = ri.Cvar_Get( "r_ignoreDstAlpha", "1", CVAR_ARCHIVE | CVAR_LATCH ); // // temporary latched variables that can only change over a restart diff --git a/code/renderergl2/tr_local.h b/code/renderergl2/tr_local.h index 6268549e..bcbaa1c5 100644 --- a/code/renderergl2/tr_local.h +++ b/code/renderergl2/tr_local.h @@ -526,83 +526,6 @@ enum ATTR_INDEX_NORMAL2 = 12 }; -enum -{ - GLS_SRCBLEND_ZERO = (1 << 0), - GLS_SRCBLEND_ONE = (1 << 1), - GLS_SRCBLEND_DST_COLOR = (1 << 2), - GLS_SRCBLEND_ONE_MINUS_DST_COLOR = (1 << 3), - GLS_SRCBLEND_SRC_ALPHA = (1 << 4), - GLS_SRCBLEND_ONE_MINUS_SRC_ALPHA = (1 << 5), - GLS_SRCBLEND_DST_ALPHA = (1 << 6), - GLS_SRCBLEND_ONE_MINUS_DST_ALPHA = (1 << 7), - GLS_SRCBLEND_ALPHA_SATURATE = (1 << 8), - - GLS_SRCBLEND_BITS = GLS_SRCBLEND_ZERO - | GLS_SRCBLEND_ONE - | GLS_SRCBLEND_DST_COLOR - | GLS_SRCBLEND_ONE_MINUS_DST_COLOR - | GLS_SRCBLEND_SRC_ALPHA - | GLS_SRCBLEND_ONE_MINUS_SRC_ALPHA - | GLS_SRCBLEND_DST_ALPHA - | GLS_SRCBLEND_ONE_MINUS_DST_ALPHA - | GLS_SRCBLEND_ALPHA_SATURATE, - - GLS_DSTBLEND_ZERO = (1 << 9), - GLS_DSTBLEND_ONE = (1 << 10), - GLS_DSTBLEND_SRC_COLOR = (1 << 11), - GLS_DSTBLEND_ONE_MINUS_SRC_COLOR = (1 << 12), - GLS_DSTBLEND_SRC_ALPHA = (1 << 13), - GLS_DSTBLEND_ONE_MINUS_SRC_ALPHA = (1 << 14), - GLS_DSTBLEND_DST_ALPHA = (1 << 15), - GLS_DSTBLEND_ONE_MINUS_DST_ALPHA = (1 << 16), - - GLS_DSTBLEND_BITS = GLS_DSTBLEND_ZERO - | GLS_DSTBLEND_ONE - | GLS_DSTBLEND_SRC_COLOR - | GLS_DSTBLEND_ONE_MINUS_SRC_COLOR - | GLS_DSTBLEND_SRC_ALPHA - | GLS_DSTBLEND_ONE_MINUS_SRC_ALPHA - | GLS_DSTBLEND_DST_ALPHA - | GLS_DSTBLEND_ONE_MINUS_DST_ALPHA, - - GLS_DEPTHMASK_TRUE = (1 << 17), - - GLS_POLYMODE_LINE = (1 << 18), - - GLS_DEPTHTEST_DISABLE = (1 << 19), - - GLS_DEPTHFUNC_LESS = (1 << 20), - GLS_DEPTHFUNC_EQUAL = (1 << 21), - - GLS_DEPTHFUNC_BITS = GLS_DEPTHFUNC_LESS - | GLS_DEPTHFUNC_EQUAL, - - GLS_ATEST_GT_0 = (1 << 22), - GLS_ATEST_LT_128 = (1 << 23), - GLS_ATEST_GE_128 = (1 << 24), -// GLS_ATEST_GE_CUSTOM = (1 << 25), - - GLS_ATEST_BITS = GLS_ATEST_GT_0 - | GLS_ATEST_LT_128 - | GLS_ATEST_GE_128, -// | GLS_ATEST_GT_CUSTOM, - - GLS_REDMASK_FALSE = (1 << 26), - GLS_GREENMASK_FALSE = (1 << 27), - GLS_BLUEMASK_FALSE = (1 << 28), - GLS_ALPHAMASK_FALSE = (1 << 29), - - GLS_COLORMASK_BITS = GLS_REDMASK_FALSE - | GLS_GREENMASK_FALSE - | GLS_BLUEMASK_FALSE - | GLS_ALPHAMASK_FALSE, - - GLS_STENCILTEST_ENABLE = (1 << 30), - - GLS_DEFAULT = GLS_DEPTHMASK_TRUE -}; - enum { ATTR_POSITION = 0x0001, @@ -924,7 +847,6 @@ typedef enum { SF_TRIANGLES, SF_POLY, SF_MDV, - SF_MD4, SF_MDR, SF_IQM, SF_FLARE, @@ -1342,7 +1264,6 @@ typedef enum { MOD_BAD, MOD_BRUSH, MOD_MESH, - MOD_MD4, MOD_MDR, MOD_IQM } modtype_t; @@ -1355,7 +1276,7 @@ typedef struct model_s { int dataSize; // just for listing purposes bmodel_t *bmodel; // only if type == MOD_BRUSH mdvModel_t *mdv[MD3_MAX_LODS]; // only if type == MOD_MESH - void *modelData; // only if type == (MOD_MD4 | MOD_MDR | MOD_IQM) + void *modelData; // only if type == (MOD_MDR | MOD_IQM) int numLods; } model_t; @@ -1502,6 +1423,8 @@ typedef struct { qboolean depthClamp; qboolean seamlessCubeMap; + + GLenum packedNormalDataType; } glRefConfig_t; @@ -1800,6 +1723,7 @@ extern cvar_t *r_ext_texture_float; extern cvar_t *r_arb_half_float_pixel; extern cvar_t *r_ext_framebuffer_multisample; extern cvar_t *r_arb_seamless_cube_map; +extern cvar_t *r_arb_vertex_type_2_10_10_10_rev; extern cvar_t *r_nobind; // turns off binding to appropriate textures extern cvar_t *r_singleShader; // make most world faces use default shader @@ -1893,6 +1817,7 @@ extern cvar_t *r_shadowMapSize; extern cvar_t *r_shadowCascadeZNear; extern cvar_t *r_shadowCascadeZFar; extern cvar_t *r_shadowCascadeZBias; +extern cvar_t *r_ignoreDstAlpha; extern cvar_t *r_greyscale; @@ -2086,13 +2011,13 @@ typedef struct shaderCommands_s { glIndex_t indexes[SHADER_MAX_INDEXES] QALIGN(16); vec4_t xyz[SHADER_MAX_VERTEXES] QALIGN(16); - uint8_t normal[SHADER_MAX_VERTEXES][4] QALIGN(16); + uint32_t normal[SHADER_MAX_VERTEXES] QALIGN(16); #ifdef USE_VERT_TANGENT_SPACE - uint8_t tangent[SHADER_MAX_VERTEXES][4] QALIGN(16); + uint32_t tangent[SHADER_MAX_VERTEXES] QALIGN(16); #endif vec2_t texCoords[SHADER_MAX_VERTEXES][2] QALIGN(16); vec4_t vertexColors[SHADER_MAX_VERTEXES] QALIGN(16); - vec4_t lightdir[SHADER_MAX_VERTEXES] QALIGN(16); + uint32_t lightdir[SHADER_MAX_VERTEXES] QALIGN(16); //int vertexDlightBits[SHADER_MAX_VERTEXES] QALIGN(16); VBO_t *vbo; @@ -2256,6 +2181,12 @@ VERTEX BUFFER OBJECTS ============================================================ */ + +uint32_t R_VboPackTangent(vec4_t v); +uint32_t R_VboPackNormal(vec3_t v); +void R_VboUnpackTangent(vec4_t v, uint32_t b); +void R_VboUnpackNormal(vec3_t v, uint32_t b); + VBO_t *R_CreateVBO(const char *name, byte * vertexes, int vertexesSize, vboUsage_t usage); VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vertexes, uint32_t stateBits, vboUsage_t usage); @@ -2346,11 +2277,8 @@ ANIMATED MODELS ============================================================= */ -// void R_MakeAnimModel( model_t *model ); haven't seen this one really, so not needed I guess. -void R_AddAnimSurfaces( trRefEntity_t *ent ); -void RB_SurfaceAnim( md4Surface_t *surfType ); void R_MDRAddAnimSurfaces( trRefEntity_t *ent ); -void RB_MDRSurfaceAnim( md4Surface_t *surface ); +void RB_MDRSurfaceAnim( mdrSurface_t *surface ); qboolean R_LoadIQM (model_t *mod, void *buffer, int filesize, const char *name ); void R_AddIQMSurfaces( trRefEntity_t *ent ); void RB_IQMSurfaceAnim( surfaceType_t *surface ); diff --git a/code/renderergl2/tr_main.c b/code/renderergl2/tr_main.c index 81bd09f3..df1f8e60 100644 --- a/code/renderergl2/tr_main.c +++ b/code/renderergl2/tr_main.c @@ -1574,9 +1574,7 @@ static qboolean SurfIsOffscreen( const drawSurf_t *drawSurf, vec4_t clipDest[128 shortest = len; } - tNormal[0] = tess.normal[tess.indexes[i]][0] / 127.5f - 1.0f; - tNormal[1] = tess.normal[tess.indexes[i]][1] / 127.5f - 1.0f; - tNormal[2] = tess.normal[tess.indexes[i]][2] / 127.5f - 1.0f; + R_VboUnpackNormal(tNormal, tess.normal[tess.indexes[i]]); if ( DotProduct( normal, tNormal ) >= 0 ) { @@ -1915,9 +1913,6 @@ static void R_AddEntitySurface (int entityNum) case MOD_MESH: R_AddMD3Surfaces( ent ); break; - case MOD_MD4: - R_AddAnimSurfaces( ent ); - break; case MOD_MDR: R_MDRAddAnimSurfaces( ent ); break; @@ -2204,12 +2199,6 @@ void R_RenderPshadowMaps(const refdef_t *fd) } break; - case MOD_MD4: - { - // FIXME: actually calculate the radius and bounds, this is a horrible hack - radius = r_pshadowDist->value / 2.0f; - } - break; case MOD_MDR: { // FIXME: never actually tested this @@ -2862,7 +2851,7 @@ void R_RenderCubemapSide( int cubemapIndex, int cubemapSide, qboolean subscene ) // FIXME: sun shadows aren't rendered correctly in cubemaps // fix involves changing r_FBufScale to fit smaller cubemap image size, or rendering cubemap to framebuffer first - if(0) //(glRefConfig.framebufferObject && (r_forceSun->integer || tr.sunShadows)) + if(0) //(glRefConfig.framebufferObject && r_sunlightMode->integer && (r_forceSun->integer || tr.sunShadows)) { R_RenderSunShadowMaps(&refdef, 0); R_RenderSunShadowMaps(&refdef, 1); diff --git a/code/renderergl2/tr_model.c b/code/renderergl2/tr_model.c index 0b1500d6..c0dbeaa2 100644 --- a/code/renderergl2/tr_model.c +++ b/code/renderergl2/tr_model.c @@ -26,7 +26,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA #define LL(x) x=LittleLong(x) static qboolean R_LoadMD3(model_t *mod, int lod, void *buffer, int bufferSize, const char *modName); -static qboolean R_LoadMD4(model_t *mod, void *buffer, const char *name ); static qboolean R_LoadMDR(model_t *mod, void *buffer, int filesize, const char *name ); /* @@ -73,15 +72,10 @@ qhandle_t R_RegisterMD3(const char *name, model_t *mod) continue; ident = LittleLong(* (unsigned *) buf.u); - if (ident == MD4_IDENT) - loaded = R_LoadMD4(mod, buf.u, name); + if (ident == MD3_IDENT) + loaded = R_LoadMD3(mod, lod, buf.u, size, name); else - { - if (ident == MD3_IDENT) - loaded = R_LoadMD3(mod, lod, buf.u, size, name); - else - ri.Printf(PRINT_WARNING,"R_RegisterMD3: unknown fileid for %s\n", name); - } + ri.Printf(PRINT_WARNING,"R_RegisterMD3: unknown fileid for %s\n", name); ri.FS_FreeFile(buf.v); @@ -201,7 +195,6 @@ static modelExtToLoaderMap_t modelLoaders[ ] = { { "iqm", R_RegisterIQM }, { "mdr", R_RegisterMDR }, - { "md4", R_RegisterMD3 }, { "md3", R_RegisterMD3 } }; @@ -690,9 +683,9 @@ static qboolean R_LoadMD3(model_t * mod, int lod, void *buffer, int bufferSize, { vec3_t *verts; vec2_t *texcoords; - uint8_t *normals; + uint32_t *normals; #ifdef USE_VERT_TANGENT_SPACE - uint8_t *tangents; + uint32_t *tangents; #endif byte *data; @@ -709,11 +702,11 @@ static qboolean R_LoadMD3(model_t * mod, int lod, void *buffer, int bufferSize, dataSize += surf->numVerts * mdvModel->numFrames * sizeof(*verts); ofs_normal = dataSize; - dataSize += surf->numVerts * mdvModel->numFrames * sizeof(*normals) * 4; + dataSize += surf->numVerts * mdvModel->numFrames * sizeof(*normals); #ifdef USE_VERT_TANGENT_SPACE ofs_tangent = dataSize; - dataSize += surf->numVerts * mdvModel->numFrames * sizeof(*tangents) * 4; + dataSize += surf->numVerts * mdvModel->numFrames * sizeof(*tangents); #endif ofs_st = dataSize; @@ -732,18 +725,17 @@ static qboolean R_LoadMD3(model_t * mod, int lod, void *buffer, int bufferSize, for ( j = 0; j < surf->numVerts * mdvModel->numFrames ; j++, v++ ) { vec3_t nxt; + vec4_t tangent; - CrossProduct(v->normal, v->tangent, nxt); VectorCopy(v->xyz, verts[j]); - normals[j*4+0] = (uint8_t)(v->normal[0] * 127.5f + 128.0f); - normals[j*4+1] = (uint8_t)(v->normal[1] * 127.5f + 128.0f); - normals[j*4+2] = (uint8_t)(v->normal[2] * 127.5f + 128.0f); - normals[j*4+3] = 0; + + normals[j] = R_VboPackNormal(v->normal); #ifdef USE_VERT_TANGENT_SPACE - tangents[j*4+0] = (uint8_t)(v->tangent[0] * 127.5f + 128.0f); - tangents[j*4+1] = (uint8_t)(v->tangent[1] * 127.5f + 128.0f); - tangents[j*4+2] = (uint8_t)(v->tangent[2] * 127.5f + 128.0f); - tangents[j*4+3] = (DotProduct(nxt, v->bitangent) < 0.0f) ? 0 : 255; + CrossProduct(v->normal, v->tangent, nxt); + VectorCopy(v->tangent, tangent); + tangent[3] = (DotProduct(nxt, v->bitangent) < 0.0f) ? -1.0f : 1.0f; + + tangents[j] = R_VboPackTangent(tangent); #endif } @@ -772,14 +764,14 @@ static qboolean R_LoadMD3(model_t * mod, int lod, void *buffer, int bufferSize, vboSurf->vbo->ofs_st = ofs_st; vboSurf->vbo->stride_xyz = sizeof(*verts); - vboSurf->vbo->stride_normal = sizeof(*normals) * 4; + vboSurf->vbo->stride_normal = sizeof(*normals); #ifdef USE_VERT_TANGENT_SPACE - vboSurf->vbo->stride_tangent = sizeof(*tangents) * 4; + vboSurf->vbo->stride_tangent = sizeof(*tangents); #endif vboSurf->vbo->stride_st = sizeof(*st); vboSurf->vbo->size_xyz = sizeof(*verts) * surf->numVerts; - vboSurf->vbo->size_normal = sizeof(*normals) * surf->numVerts * 4; + vboSurf->vbo->size_normal = sizeof(*normals) * surf->numVerts; ri.Free(data); @@ -835,7 +827,7 @@ static qboolean R_LoadMDR( model_t *mod, void *buffer, int filesize, const char LL(pinmodel->ofsFrames); // This is a model that uses some type of compressed Bones. We don't want to uncompress every bone for each rendered frame - // over and over again, we'll uncompress it in this function already, so we must adjust the size of the target md4. + // over and over again, we'll uncompress it in this function already, so we must adjust the size of the target mdr. if(pinmodel->ofsFrames < 0) { // mdrFrame_t is larger than mdrCompFrame_t: @@ -1132,162 +1124,6 @@ static qboolean R_LoadMDR( model_t *mod, void *buffer, int filesize, const char return qtrue; } -/* -================= -R_LoadMD4 -================= -*/ - -static qboolean R_LoadMD4( model_t *mod, void *buffer, const char *mod_name ) { - int i, j, k, lodindex; - md4Header_t *pinmodel, *md4; - md4Frame_t *frame; - md4LOD_t *lod; - md4Surface_t *surf; - md4Triangle_t *tri; - md4Vertex_t *v; - int version; - int size; - shader_t *sh; - int frameSize; - - pinmodel = (md4Header_t *)buffer; - - version = LittleLong (pinmodel->version); - if (version != MD4_VERSION) { - ri.Printf( PRINT_WARNING, "R_LoadMD4: %s has wrong version (%i should be %i)\n", - mod_name, version, MD4_VERSION); - return qfalse; - } - - mod->type = MOD_MD4; - size = LittleLong(pinmodel->ofsEnd); - mod->dataSize += size; - mod->modelData = md4 = ri.Hunk_Alloc( size, h_low ); - - Com_Memcpy(md4, buffer, size); - - LL(md4->ident); - LL(md4->version); - LL(md4->numFrames); - LL(md4->numBones); - LL(md4->numLODs); - LL(md4->ofsFrames); - LL(md4->ofsLODs); - md4->ofsEnd = size; - - if ( md4->numFrames < 1 ) { - ri.Printf( PRINT_WARNING, "R_LoadMD4: %s has no frames\n", mod_name ); - return qfalse; - } - - // we don't need to swap tags in the renderer, they aren't used - - // swap all the frames - frameSize = (size_t)( &((md4Frame_t *)0)->bones[ md4->numBones ] ); - for ( i = 0 ; i < md4->numFrames ; i++) { - frame = (md4Frame_t *) ( (byte *)md4 + md4->ofsFrames + i * frameSize ); - frame->radius = LittleFloat( frame->radius ); - for ( j = 0 ; j < 3 ; j++ ) { - frame->bounds[0][j] = LittleFloat( frame->bounds[0][j] ); - frame->bounds[1][j] = LittleFloat( frame->bounds[1][j] ); - frame->localOrigin[j] = LittleFloat( frame->localOrigin[j] ); - } - for ( j = 0 ; j < md4->numBones * sizeof( md4Bone_t ) / 4 ; j++ ) { - ((float *)frame->bones)[j] = LittleFloat( ((float *)frame->bones)[j] ); - } - } - - // swap all the LOD's - lod = (md4LOD_t *) ( (byte *)md4 + md4->ofsLODs ); - for ( lodindex = 0 ; lodindex < md4->numLODs ; lodindex++ ) { - - // swap all the surfaces - surf = (md4Surface_t *) ( (byte *)lod + lod->ofsSurfaces ); - for ( i = 0 ; i < lod->numSurfaces ; i++) { - LL(surf->ident); - LL(surf->numTriangles); - LL(surf->ofsTriangles); - LL(surf->numVerts); - LL(surf->ofsVerts); - LL(surf->ofsEnd); - - if ( surf->numVerts >= SHADER_MAX_VERTEXES ) { - ri.Printf(PRINT_WARNING, "R_LoadMD4: %s has more than %i verts on %s (%i).\n", - mod_name, SHADER_MAX_VERTEXES - 1, surf->name[0] ? surf->name : "a surface", - surf->numVerts ); - return qfalse; - } - if ( surf->numTriangles*3 >= SHADER_MAX_INDEXES ) { - ri.Printf(PRINT_WARNING, "R_LoadMD4: %s has more than %i triangles on %s (%i).\n", - mod_name, ( SHADER_MAX_INDEXES / 3 ) - 1, surf->name[0] ? surf->name : "a surface", - surf->numTriangles ); - return qfalse; - } - - // change to surface identifier - surf->ident = SF_MD4; - - // lowercase the surface name so skin compares are faster - Q_strlwr( surf->name ); - - // register the shaders - sh = R_FindShader( surf->shader, LIGHTMAP_NONE, qtrue ); - if ( sh->defaultShader ) { - surf->shaderIndex = 0; - } else { - surf->shaderIndex = sh->index; - } - - // swap all the triangles - tri = (md4Triangle_t *) ( (byte *)surf + surf->ofsTriangles ); - for ( j = 0 ; j < surf->numTriangles ; j++, tri++ ) { - LL(tri->indexes[0]); - LL(tri->indexes[1]); - LL(tri->indexes[2]); - } - - // swap all the vertexes - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *) ( (byte *)surf + surf->ofsVerts + 12); - v = (md4Vertex_t *) ( (byte *)surf + surf->ofsVerts); - for ( j = 0 ; j < surf->numVerts ; j++ ) { - v->normal[0] = LittleFloat( v->normal[0] ); - v->normal[1] = LittleFloat( v->normal[1] ); - v->normal[2] = LittleFloat( v->normal[2] ); - - v->texCoords[0] = LittleFloat( v->texCoords[0] ); - v->texCoords[1] = LittleFloat( v->texCoords[1] ); - - v->numWeights = LittleLong( v->numWeights ); - - for ( k = 0 ; k < v->numWeights ; k++ ) { - v->weights[k].boneIndex = LittleLong( v->weights[k].boneIndex ); - v->weights[k].boneWeight = LittleFloat( v->weights[k].boneWeight ); - v->weights[k].offset[0] = LittleFloat( v->weights[k].offset[0] ); - v->weights[k].offset[1] = LittleFloat( v->weights[k].offset[1] ); - v->weights[k].offset[2] = LittleFloat( v->weights[k].offset[2] ); - } - // FIXME - // This makes TFC's skeletons work. Shouldn't be necessary anymore, but left - // in for reference. - //v = (md4Vertex_t *)( ( byte * )&v->weights[v->numWeights] + 12 ); - v = (md4Vertex_t *)( ( byte * )&v->weights[v->numWeights]); - } - - // find the next surface - surf = (md4Surface_t *)( (byte *)surf + surf->ofsEnd ); - } - - // find the next LOD - lod = (md4LOD_t *)( (byte *)lod + lod->ofsEnd ); - } - - return qtrue; -} - //============================================================================= @@ -1310,11 +1146,6 @@ void RE_BeginRegistration( glconfig_t *glconfigOut ) { RE_ClearScene(); tr.registered = qtrue; - - // NOTE: this sucks, for some reason the first stretch pic is never drawn - // without this we'd see a white flash on a level load because the very - // first time the level shot would not be drawn -// RE_StretchPic(0, 0, 0, 0, 0, 0, 1, 1, 0); } //============================================================================= @@ -1527,17 +1358,6 @@ void R_ModelBounds( qhandle_t handle, vec3_t mins, vec3_t maxs ) { VectorCopy( frame->bounds[0], mins ); VectorCopy( frame->bounds[1], maxs ); - return; - } else if (model->type == MOD_MD4) { - md4Header_t *header; - md4Frame_t *frame; - - header = (md4Header_t *)model->modelData; - frame = (md4Frame_t *) ((byte *)header + header->ofsFrames); - - VectorCopy( frame->bounds[0], mins ); - VectorCopy( frame->bounds[1], maxs ); - return; } else if (model->type == MOD_MDR) { mdrHeader_t *header; diff --git a/code/renderergl2/tr_model_iqm.c b/code/renderergl2/tr_model_iqm.c index d41a8f52..b3843c5d 100644 --- a/code/renderergl2/tr_model_iqm.c +++ b/code/renderergl2/tr_model_iqm.c @@ -1024,7 +1024,10 @@ void RB_IQMSurfaceAnim( surfaceType_t *surface ) { int i; vec4_t *outXYZ; - uint8_t *outNormal; + uint32_t *outNormal; +#ifdef USE_VERT_TANGENT_SPACE + uint32_t *outTangent; +#endif vec2_t (*outTexCoord)[2]; vec4_t *outColor; @@ -1039,7 +1042,10 @@ void RB_IQMSurfaceAnim( surfaceType_t *surface ) { RB_CHECKOVERFLOW( surf->num_vertexes, surf->num_triangles * 3 ); outXYZ = &tess.xyz[tess.numVertexes]; - outNormal = &tess.normal[tess.numVertexes][0]; + outNormal = &tess.normal[tess.numVertexes]; +#ifdef USE_VERT_TANGENT_SPACE + outTangent = &tess.tangent[tess.numVertexes]; +#endif outTexCoord = &tess.texCoords[tess.numVertexes]; outColor = &tess.vertexColors[tess.numVertexes]; @@ -1050,7 +1056,7 @@ void RB_IQMSurfaceAnim( surfaceType_t *surface ) { // transform vertexes and fill other data for( i = 0; i < surf->num_vertexes; - i++, outXYZ++, outNormal+=4, outTexCoord++, outColor++ ) { + i++, outXYZ++, outNormal++, outTexCoord++, outColor++ ) { int j, k; float vtxMat[12]; float nrmMat[9]; @@ -1116,22 +1122,25 @@ void RB_IQMSurfaceAnim( surfaceType_t *surface ) { vtxMat[11]; (*outXYZ)[3] = 1.0f; - (outNormal)[0] = (uint8_t)(( - nrmMat[ 0] * data->normals[3*vtx+0] + - nrmMat[ 1] * data->normals[3*vtx+1] + - nrmMat[ 2] * data->normals[3*vtx+2] - )* 127.5f + 128.0f); - (outNormal)[1] = (uint8_t)(( - nrmMat[ 3] * data->normals[3*vtx+0] + - nrmMat[ 4] * data->normals[3*vtx+1] + - nrmMat[ 5] * data->normals[3*vtx+2] - )* 127.5f + 128.0f); - (outNormal)[2] = (uint8_t)(( - nrmMat[ 6] * data->normals[3*vtx+0] + - nrmMat[ 7] * data->normals[3*vtx+1] + - nrmMat[ 8] * data->normals[3*vtx+2] - )* 127.5f + 128.0f); - (outNormal)[3] = 0; + { + vec3_t normal; + vec4_t tangent; + + normal[0] = DotProduct(&nrmMat[0], &data->normals[3*vtx]); + normal[1] = DotProduct(&nrmMat[3], &data->normals[3*vtx]); + normal[2] = DotProduct(&nrmMat[6], &data->normals[3*vtx]); + + *outNormal = R_VboPackNormal(normal); + +#ifdef USE_VERT_TANGENT_SPACE + tangent[0] = DotProduct(&nrmMat[0], &data->tangents[4*vtx]); + tangent[1] = DotProduct(&nrmMat[3], &data->tangents[4*vtx]); + tangent[2] = DotProduct(&nrmMat[6], &data->tangents[4*vtx]); + tangent[3] = data->tangents[4*vtx+3]; + + *outTangent++ = R_VboPackTangent(tangent); +#endif + } (*outColor)[0] = data->colors[4*vtx+0] / 255.0f; (*outColor)[1] = data->colors[4*vtx+1] / 255.0f; diff --git a/code/renderergl2/tr_scene.c b/code/renderergl2/tr_scene.c index e3f52958..db2d1386 100644 --- a/code/renderergl2/tr_scene.c +++ b/code/renderergl2/tr_scene.c @@ -496,7 +496,7 @@ void RE_RenderScene( const refdef_t *fd ) { } // playing with even more shadows - if(glRefConfig.framebufferObject && !( fd->rdflags & RDF_NOWORLDMODEL ) && (r_forceSun->integer || tr.sunShadows)) + if(glRefConfig.framebufferObject && r_sunlightMode->integer && !( fd->rdflags & RDF_NOWORLDMODEL ) && (r_forceSun->integer || tr.sunShadows)) { R_RenderSunShadowMaps(fd, 0); R_RenderSunShadowMaps(fd, 1); diff --git a/code/renderergl2/tr_shade.c b/code/renderergl2/tr_shade.c index 9e53e73b..856b8748 100644 --- a/code/renderergl2/tr_shade.c +++ b/code/renderergl2/tr_shade.c @@ -1241,7 +1241,7 @@ static void RB_IterateStagesGeneric( shaderCommands_t *input ) int i; vec4_t enableTextures; - if ((backEnd.viewParms.flags & VPF_USESUNLIGHT) && (pStage->glslShaderIndex & LIGHTDEF_LIGHTTYPE_MASK)) + if (r_sunlightMode->integer && (backEnd.viewParms.flags & VPF_USESUNLIGHT) && (pStage->glslShaderIndex & LIGHTDEF_LIGHTTYPE_MASK)) { GL_BindToTMU(tr.screenShadowImage, TB_SHADOWMAP); GLSL_SetUniformVec3(sp, UNIFORM_PRIMARYLIGHTAMBIENT, backEnd.refdef.sunAmbCol); @@ -1552,8 +1552,8 @@ void RB_StageIteratorGeneric( void ) // // pshadows! // - if (glRefConfig.framebufferObject && tess.pshadowBits && tess.shader->sort <= SS_OPAQUE - && !(tess.shader->surfaceFlags & (SURF_NODLIGHT | SURF_SKY) ) ) { + if (glRefConfig.framebufferObject && r_shadows->integer == 4 && tess.pshadowBits + && tess.shader->sort <= SS_OPAQUE && !(tess.shader->surfaceFlags & (SURF_NODLIGHT | SURF_SKY) ) ) { ProjectPshadowVBOGLSL(); } diff --git a/code/renderergl2/tr_shade_calc.c b/code/renderergl2/tr_shade_calc.c index d77136e7..c9816fa8 100644 --- a/code/renderergl2/tr_shade_calc.c +++ b/code/renderergl2/tr_shade_calc.c @@ -116,29 +116,27 @@ void RB_CalcDeformVertexes( deformStage_t *ds ) vec3_t offset; float scale; float *xyz = ( float * ) tess.xyz; - uint8_t *normal = ( uint8_t * ) tess.normal; + uint32_t *normal = tess.normal; float *table; if ( ds->deformationWave.frequency == 0 ) { scale = EvalWaveForm( &ds->deformationWave ); - for ( i = 0; i < tess.numVertexes; i++, xyz += 4, normal += 4 ) + for ( i = 0; i < tess.numVertexes; i++, xyz += 4, normal++ ) { - offset[0] = (normal[0] / 127.5f - 1.0f) * scale; - offset[1] = (normal[1] / 127.5f - 1.0f) * scale; - offset[2] = (normal[2] / 127.5f - 1.0f) * scale; + R_VboUnpackNormal(offset, *normal); - xyz[0] += offset[0]; - xyz[1] += offset[1]; - xyz[2] += offset[2]; + xyz[0] += offset[0] * scale; + xyz[1] += offset[1] * scale; + xyz[2] += offset[2] * scale; } } else { table = TableForFunc( ds->deformationWave.func ); - for ( i = 0; i < tess.numVertexes; i++, xyz += 4, normal += 4 ) + for ( i = 0; i < tess.numVertexes; i++, xyz += 4, normal++ ) { float off = ( xyz[0] + xyz[1] + xyz[2] ) * ds->deformationSpread; @@ -147,13 +145,11 @@ void RB_CalcDeformVertexes( deformStage_t *ds ) ds->deformationWave.phase + off, ds->deformationWave.frequency ); - offset[0] = (normal[0] / 127.5f - 1.0f) * scale; - offset[1] = (normal[1] / 127.5f - 1.0f) * scale; - offset[2] = (normal[2] / 127.5f - 1.0f) * scale; - - xyz[0] += offset[0]; - xyz[1] += offset[1]; - xyz[2] += offset[2]; + R_VboUnpackNormal(offset, *normal); + + xyz[0] += offset[0] * scale; + xyz[1] += offset[1] * scale; + xyz[2] += offset[2] * scale; } } } @@ -169,14 +165,12 @@ void RB_CalcDeformNormals( deformStage_t *ds ) { int i; float scale; float *xyz = ( float * ) tess.xyz; - uint8_t *normal = ( uint8_t * ) tess.normal; + uint32_t *normal = tess.normal; - for ( i = 0; i < tess.numVertexes; i++, xyz += 4, normal += 4 ) { + for ( i = 0; i < tess.numVertexes; i++, xyz += 4, normal++ ) { vec3_t fNormal; - fNormal[0] = normal[0] / 127.5f - 1.0f; - fNormal[1] = normal[1] / 127.5f - 1.0f; - fNormal[2] = normal[2] / 127.5f - 1.0f; + R_VboUnpackNormal(fNormal, *normal); scale = 0.98f; scale = R_NoiseGet4f( xyz[0] * scale, xyz[1] * scale, xyz[2] * scale, @@ -195,9 +189,7 @@ void RB_CalcDeformNormals( deformStage_t *ds ) { VectorNormalizeFast( fNormal ); - normal[0] = (uint8_t)(fNormal[0] * 127.5f + 128.0f); - normal[1] = (uint8_t)(fNormal[0] * 127.5f + 128.0f); - normal[2] = (uint8_t)(fNormal[0] * 127.5f + 128.0f); + *normal = R_VboPackNormal(fNormal); } } @@ -211,22 +203,25 @@ void RB_CalcBulgeVertexes( deformStage_t *ds ) { int i; const float *st = ( const float * ) tess.texCoords[0]; float *xyz = ( float * ) tess.xyz; - uint8_t *normal = ( uint8_t * ) tess.normal; + uint32_t *normal = tess.normal; float now; now = backEnd.refdef.time * ds->bulgeSpeed * 0.001f; - for ( i = 0; i < tess.numVertexes; i++, xyz += 4, st += 4, normal += 4 ) { + for ( i = 0; i < tess.numVertexes; i++, xyz += 4, st += 4, normal++ ) { int off; float scale; + vec3_t fNormal; + + R_VboUnpackNormal(fNormal, *normal); off = (float)( FUNCTABLE_SIZE / (M_PI*2) ) * ( st[0] * ds->bulgeWidth + now ); scale = tr.sinTable[ off & FUNCTABLE_MASK ] * ds->bulgeHeight; - xyz[0] += (normal[0] / 127.5f - 1.0f) * scale; - xyz[1] += (normal[1] / 127.5f - 1.0f) * scale; - xyz[2] += (normal[2] / 127.5f - 1.0f) * scale; + xyz[0] += fNormal[0] * scale; + xyz[1] += fNormal[1] * scale; + xyz[2] += fNormal[2] * scale; } } @@ -281,11 +276,8 @@ void DeformText( const char *text ) { height[0] = 0; height[1] = 0; height[2] = -1; - - fNormal[0] = tess.normal[0][0] / 127.5f - 1.0f; - fNormal[1] = tess.normal[0][1] / 127.5f - 1.0f; - fNormal[2] = tess.normal[0][2] / 127.5f - 1.0f; + R_VboUnpackNormal(fNormal, tess.normal[0]); CrossProduct( fNormal, height, width ); // find the midpoint of the box diff --git a/code/renderergl2/tr_shader.c b/code/renderergl2/tr_shader.c index 308b0f14..6dbac55a 100644 --- a/code/renderergl2/tr_shader.c +++ b/code/renderergl2/tr_shader.c @@ -203,10 +203,16 @@ static int NameToSrcBlendMode( const char *name ) } else if ( !Q_stricmp( name, "GL_DST_ALPHA" ) ) { + if (r_ignoreDstAlpha->integer) + return GLS_SRCBLEND_ONE; + return GLS_SRCBLEND_DST_ALPHA; } else if ( !Q_stricmp( name, "GL_ONE_MINUS_DST_ALPHA" ) ) { + if (r_ignoreDstAlpha->integer) + return GLS_SRCBLEND_ZERO; + return GLS_SRCBLEND_ONE_MINUS_DST_ALPHA; } else if ( !Q_stricmp( name, "GL_SRC_ALPHA_SATURATE" ) ) @@ -243,10 +249,16 @@ static int NameToDstBlendMode( const char *name ) } else if ( !Q_stricmp( name, "GL_DST_ALPHA" ) ) { + if (r_ignoreDstAlpha->integer) + return GLS_DSTBLEND_ONE; + return GLS_DSTBLEND_DST_ALPHA; } else if ( !Q_stricmp( name, "GL_ONE_MINUS_DST_ALPHA" ) ) { + if (r_ignoreDstAlpha->integer) + return GLS_DSTBLEND_ZERO; + return GLS_DSTBLEND_ONE_MINUS_DST_ALPHA; } else if ( !Q_stricmp( name, "GL_SRC_COLOR" ) ) diff --git a/code/renderergl2/tr_surface.c b/code/renderergl2/tr_surface.c index ee368cbd..825582bd 100644 --- a/code/renderergl2/tr_surface.c +++ b/code/renderergl2/tr_surface.c @@ -124,26 +124,11 @@ void RB_AddQuadStampExt( vec3_t origin, vec3_t left, vec3_t up, float color[4], // constant normal all the way around VectorSubtract( vec3_origin, backEnd.viewParms.or.axis[0], normal ); - tess.normal[ndx][0] = (uint8_t)(normal[0] * 127.5f + 128.0f); - tess.normal[ndx][1] = (uint8_t)(normal[1] * 127.5f + 128.0f); - tess.normal[ndx][2] = (uint8_t)(normal[2] * 127.5f + 128.0f); - tess.normal[ndx][3] = 0; + tess.normal[ndx] = + tess.normal[ndx+1] = + tess.normal[ndx+2] = + tess.normal[ndx+3] = R_VboPackNormal(normal); - tess.normal[ndx+1][0] = (uint8_t)(normal[0] * 127.5f + 128.0f); - tess.normal[ndx+1][1] = (uint8_t)(normal[1] * 127.5f + 128.0f); - tess.normal[ndx+1][2] = (uint8_t)(normal[2] * 127.5f + 128.0f); - tess.normal[ndx+1][3] = 0; - - tess.normal[ndx+2][0] = (uint8_t)(normal[0] * 127.5f + 128.0f); - tess.normal[ndx+2][1] = (uint8_t)(normal[1] * 127.5f + 128.0f); - tess.normal[ndx+2][2] = (uint8_t)(normal[2] * 127.5f + 128.0f); - tess.normal[ndx+2][3] = 0; - - tess.normal[ndx+3][0] = (uint8_t)(normal[0] * 127.5f + 128.0f); - tess.normal[ndx+3][1] = (uint8_t)(normal[1] * 127.5f + 128.0f); - tess.normal[ndx+3][2] = (uint8_t)(normal[2] * 127.5f + 128.0f); - tess.normal[ndx+3][3] = 0; - // standard square texture coordinates VectorSet2(tess.texCoords[ndx ][0], s1, t1); VectorSet2(tess.texCoords[ndx ][1], s1, t1); @@ -331,10 +316,11 @@ static void RB_SurfaceVertsAndIndexes( int numVerts, srfVert_t *verts, int numIn int i; glIndex_t *inIndex; srfVert_t *dv; - float *xyz, *texCoords, *lightCoords, *lightdir; - uint8_t *normal; + float *xyz, *texCoords, *lightCoords; + uint32_t *lightdir; + uint32_t *normal; #ifdef USE_VERT_TANGENT_SPACE - uint8_t *tangent; + uint32_t *tangent; #endif glIndex_t *outIndex; float *color; @@ -361,28 +347,18 @@ static void RB_SurfaceVertsAndIndexes( int numVerts, srfVert_t *verts, int numIn if ( tess.shader->vertexAttribs & ATTR_NORMAL ) { dv = verts; - normal = tess.normal[ tess.numVertexes ]; - for ( i = 0 ; i < numVerts ; i++, dv++, normal+=4 ) - { - normal[0] = (uint8_t)(dv->normal[0] * 127.5f + 128.0f); - normal[1] = (uint8_t)(dv->normal[1] * 127.5f + 128.0f); - normal[2] = (uint8_t)(dv->normal[2] * 127.5f + 128.0f); - normal[3] = 0; - } + normal = &tess.normal[ tess.numVertexes ]; + for ( i = 0 ; i < numVerts ; i++, dv++, normal++ ) + *normal = R_VboPackNormal(dv->normal); } #ifdef USE_VERT_TANGENT_SPACE if ( tess.shader->vertexAttribs & ATTR_TANGENT ) { dv = verts; - tangent = tess.tangent[ tess.numVertexes ]; - for ( i = 0 ; i < numVerts ; i++, dv++, tangent+=4 ) - { - tangent[0] = (uint8_t)(dv->tangent[0] * 127.5f + 128.0f); - tangent[1] = (uint8_t)(dv->tangent[1] * 127.5f + 128.0f); - tangent[2] = (uint8_t)(dv->tangent[2] * 127.5f + 128.0f); - tangent[3] = (uint8_t)(dv->tangent[3] * 127.5f + 128.0f); - } + tangent = &tess.tangent[ tess.numVertexes ]; + for ( i = 0 ; i < numVerts ; i++, dv++, tangent++ ) + *tangent = R_VboPackTangent(dv->tangent); } #endif @@ -413,9 +389,9 @@ static void RB_SurfaceVertsAndIndexes( int numVerts, srfVert_t *verts, int numIn if ( tess.shader->vertexAttribs & ATTR_LIGHTDIRECTION ) { dv = verts; - lightdir = tess.lightdir[ tess.numVertexes ]; - for ( i = 0 ; i < numVerts ; i++, dv++, lightdir+=4 ) - VectorCopy(dv->lightdir, lightdir); + lightdir = &tess.lightdir[ tess.numVertexes ]; + for ( i = 0 ; i < numVerts ; i++, dv++, lightdir++ ) + *lightdir = R_VboPackNormal(dv->lightdir); } #if 0 // nothing even uses vertex dlightbits @@ -651,6 +627,8 @@ static void DoRailCore( const vec3_t start, const vec3_t end, const vec3_t up, f int vbase; float t = len / 256.0f; + RB_CHECKOVERFLOW( 4, 6 ); + vbase = tess.numVertexes; spanWidth2 = -spanWidth; @@ -1150,14 +1128,14 @@ static void LerpMeshVertexes_scalar(mdvSurface_t *surf, float backlerp) } #endif float *outXyz; - uint8_t *outNormal; + uint32_t *outNormal; mdvVertex_t *newVerts; int vertNum; newVerts = surf->verts + backEnd.currentEntity->e.frame * surf->numVerts; outXyz = tess.xyz[tess.numVertexes]; - outNormal = tess.normal[tess.numVertexes]; + outNormal = &tess.normal[tess.numVertexes]; if (backlerp == 0) { @@ -1172,14 +1150,11 @@ static void LerpMeshVertexes_scalar(mdvSurface_t *surf, float backlerp) VectorCopy(newVerts->xyz, outXyz); VectorCopy(newVerts->normal, normal); - outNormal[0] = (uint8_t)(normal[0] * 127.5f + 128.0f); - outNormal[1] = (uint8_t)(normal[1] * 127.5f + 128.0f); - outNormal[2] = (uint8_t)(normal[2] * 127.5f + 128.0f); - outNormal[3] = 0; + *outNormal = R_VboPackNormal(normal); newVerts++; outXyz += 4; - outNormal += 4; + outNormal++; } } else @@ -1200,15 +1175,12 @@ static void LerpMeshVertexes_scalar(mdvSurface_t *surf, float backlerp) VectorLerp(newVerts->normal, oldVerts->normal, backlerp, normal); VectorNormalize(normal); - outNormal[0] = (uint8_t)(normal[0] * 127.5f + 128.0f); - outNormal[1] = (uint8_t)(normal[1] * 127.5f + 128.0f); - outNormal[2] = (uint8_t)(normal[2] * 127.5f + 128.0f); - outNormal[3] = 0; + *outNormal = R_VboPackNormal(normal); newVerts++; oldVerts++; outXyz += 4; - outNormal += 4; + outNormal++; } } @@ -1330,11 +1302,12 @@ static void RB_SurfaceGrid( srfBspSurface_t *srf ) { int i, j; float *xyz; float *texCoords, *lightCoords; - uint8_t *normal; + uint32_t *normal; #ifdef USE_VERT_TANGENT_SPACE - uint8_t *tangent; + uint32_t *tangent; #endif - float *color, *lightdir; + float *color; + uint32_t *lightdir; srfVert_t *dv; int rows, irows, vrows; int used; @@ -1417,14 +1390,14 @@ static void RB_SurfaceGrid( srfBspSurface_t *srf ) { numVertexes = tess.numVertexes; xyz = tess.xyz[numVertexes]; - normal = tess.normal[numVertexes]; + normal = &tess.normal[numVertexes]; #ifdef USE_VERT_TANGENT_SPACE - tangent = tess.tangent[numVertexes]; + tangent = &tess.tangent[numVertexes]; #endif texCoords = tess.texCoords[numVertexes][0]; lightCoords = tess.texCoords[numVertexes][1]; color = tess.vertexColors[numVertexes]; - lightdir = tess.lightdir[numVertexes]; + lightdir = &tess.lightdir[numVertexes]; //vDlightBits = &tess.vertexDlightBits[numVertexes]; for ( i = 0 ; i < rows ; i++ ) { @@ -1440,21 +1413,13 @@ static void RB_SurfaceGrid( srfBspSurface_t *srf ) { if ( tess.shader->vertexAttribs & ATTR_NORMAL ) { - normal[0] = (uint8_t)(dv->normal[0] * 127.5f + 128.0f); - normal[1] = (uint8_t)(dv->normal[1] * 127.5f + 128.0f); - normal[2] = (uint8_t)(dv->normal[2] * 127.5f + 128.0f); - normal[3] = 0; - normal += 4; + *normal++ = R_VboPackNormal(dv->normal); } #ifdef USE_VERT_TANGENT_SPACE if ( tess.shader->vertexAttribs & ATTR_TANGENT ) { - tangent[0] = (uint8_t)(dv->tangent[0] * 127.5f + 128.0f); - tangent[1] = (uint8_t)(dv->tangent[1] * 127.5f + 128.0f); - tangent[2] = (uint8_t)(dv->tangent[2] * 127.5f + 128.0f); - tangent[3] = (uint8_t)(dv->tangent[3] * 127.5f + 128.0f); - tangent += 4; + *tangent++ = R_VboPackTangent(dv->tangent); } #endif if ( tess.shader->vertexAttribs & ATTR_TEXCOORD ) @@ -1477,8 +1442,7 @@ static void RB_SurfaceGrid( srfBspSurface_t *srf ) { if ( tess.shader->vertexAttribs & ATTR_LIGHTDIRECTION ) { - VectorCopy(dv->lightdir, lightdir); - lightdir += 4; + *lightdir++ = R_VboPackNormal(dv->lightdir); } //*vDlightBits++ = dlightBits; @@ -1676,7 +1640,6 @@ void (*rb_surfaceTable[SF_NUM_SURFACE_TYPES])( void *) = { (void(*)(void*))RB_SurfaceTriangles, // SF_TRIANGLES, (void(*)(void*))RB_SurfacePolychain, // SF_POLY, (void(*)(void*))RB_SurfaceMesh, // SF_MDV, - (void(*)(void*))RB_SurfaceAnim, // SF_MD4, (void(*)(void*))RB_MDRSurfaceAnim, // SF_MDR, (void(*)(void*))RB_IQMSurfaceAnim, // SF_IQM, (void(*)(void*))RB_SurfaceFlare, // SF_FLARE, diff --git a/code/renderergl2/tr_vbo.c b/code/renderergl2/tr_vbo.c index 8c1a6779..dc0f33be 100644 --- a/code/renderergl2/tr_vbo.c +++ b/code/renderergl2/tr_vbo.c @@ -22,6 +22,75 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA // tr_vbo.c #include "tr_local.h" + +uint32_t R_VboPackTangent(vec4_t v) +{ + if (glRefConfig.packedNormalDataType == GL_UNSIGNED_INT_2_10_10_10_REV) + { + return (((uint32_t)(v[3] * 1.5f + 2.0f )) << 30) + | (((uint32_t)(v[2] * 511.5f + 512.0f)) << 20) + | (((uint32_t)(v[1] * 511.5f + 512.0f)) << 10) + | (((uint32_t)(v[0] * 511.5f + 512.0f))); + } + else + { + return (((uint32_t)(v[3] * 127.5f + 128.0f)) << 24) + | (((uint32_t)(v[2] * 127.5f + 128.0f)) << 16) + | (((uint32_t)(v[1] * 127.5f + 128.0f)) << 8) + | (((uint32_t)(v[0] * 127.5f + 128.0f))); + } +} + +uint32_t R_VboPackNormal(vec3_t v) +{ + if (glRefConfig.packedNormalDataType == GL_UNSIGNED_INT_2_10_10_10_REV) + { + return (((uint32_t)(v[2] * 511.5f + 512.0f)) << 20) + | (((uint32_t)(v[1] * 511.5f + 512.0f)) << 10) + | (((uint32_t)(v[0] * 511.5f + 512.0f))); + } + else + { + return (((uint32_t)(v[2] * 127.5f + 128.0f)) << 16) + | (((uint32_t)(v[1] * 127.5f + 128.0f)) << 8) + | (((uint32_t)(v[0] * 127.5f + 128.0f))); + } +} + +void R_VboUnpackTangent(vec4_t v, uint32_t b) +{ + if (glRefConfig.packedNormalDataType == GL_UNSIGNED_INT_2_10_10_10_REV) + { + v[0] = ((b) & 0x3ff) * 1.0f/511.5f - 1.0f; + v[1] = ((b >> 10) & 0x3ff) * 1.0f/511.5f - 1.0f; + v[2] = ((b >> 20) & 0x3ff) * 1.0f/511.5f - 1.0f; + v[3] = ((b >> 30) & 0x3) * 1.0f/1.5f - 1.0f; + } + else + { + v[0] = ((b) & 0xff) * 1.0f/127.5f - 1.0f; + v[1] = ((b >> 8) & 0xff) * 1.0f/127.5f - 1.0f; + v[2] = ((b >> 16) & 0xff) * 1.0f/127.5f - 1.0f; + v[3] = ((b >> 24) & 0xff) * 1.0f/127.5f - 1.0f; + } +} + +void R_VboUnpackNormal(vec3_t v, uint32_t b) +{ + if (glRefConfig.packedNormalDataType == GL_UNSIGNED_INT_2_10_10_10_REV) + { + v[0] = ((b) & 0x3ff) * 1.0f/511.5f - 1.0f; + v[1] = ((b >> 10) & 0x3ff) * 1.0f/511.5f - 1.0f; + v[2] = ((b >> 20) & 0x3ff) * 1.0f/511.5f - 1.0f; + } + else + { + v[0] = ((b) & 0xff) * 1.0f/127.5f - 1.0f; + v[1] = ((b >> 8) & 0xff) * 1.0f/127.5f - 1.0f; + v[2] = ((b >> 16) & 0xff) * 1.0f/127.5f - 1.0f; + } +} + /* ============ R_CreateVBO @@ -142,14 +211,14 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert if(stateBits & ATTR_NORMAL) { vbo->ofs_normal = dataSize; - dataSize += sizeof(uint8_t) * 4; + dataSize += sizeof(uint32_t); } #ifdef USE_VERT_TANGENT_SPACE if(stateBits & ATTR_TANGENT) { vbo->ofs_tangent = dataSize; - dataSize += sizeof(uint8_t) * 4; + dataSize += sizeof(uint32_t); } #endif @@ -174,7 +243,7 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert if(stateBits & ATTR_LIGHTDIRECTION) { vbo->ofs_lightdir = dataSize; - dataSize += sizeof(verts[0].lightdir); + dataSize += sizeof(uint32_t); } vbo->stride_xyz = dataSize; @@ -204,31 +273,22 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert // normal if(stateBits & ATTR_NORMAL) { - uint8_t *p = data + dataOfs; + uint32_t *p = (uint32_t *)(data + dataOfs); - p[0] = (uint8_t)(verts[i].normal[0] * 127.5f + 128.0f); - p[1] = (uint8_t)(verts[i].normal[1] * 127.5f + 128.0f); - p[2] = (uint8_t)(verts[i].normal[2] * 127.5f + 128.0f); - p[3] = 0; + *p = R_VboPackNormal(verts[i].normal); - dataOfs += sizeof(uint8_t) * 4; + dataOfs += sizeof(uint32_t); } #ifdef USE_VERT_TANGENT_SPACE // tangent if(stateBits & ATTR_TANGENT) { - vec3_t nxt; - uint8_t *p = data + dataOfs; + uint32_t *p = (uint32_t *)(data + dataOfs); - CrossProduct(verts[i].normal, verts[i].tangent, nxt); + *p = R_VboPackTangent(verts[i].tangent); - p[0] = (uint8_t)(verts[i].tangent[0] * 127.5f + 128.0f); - p[1] = (uint8_t)(verts[i].tangent[1] * 127.5f + 128.0f); - p[2] = (uint8_t)(verts[i].tangent[2] * 127.5f + 128.0f); - p[3] = (uint8_t)(verts[i].tangent[3] * 127.5f + 128.0f); - - dataOfs += sizeof(uint8_t) * 4; + dataOfs += sizeof(uint32_t); } #endif @@ -256,8 +316,11 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert // feed vertex light directions if(stateBits & ATTR_LIGHTDIRECTION) { - memcpy(data + dataOfs, &verts[i].lightdir, sizeof(verts[i].lightdir)); - dataOfs += sizeof(verts[i].lightdir); + uint32_t *p = (uint32_t *)(data + dataOfs); + + *p = R_VboPackNormal(verts[i].lightdir); + + dataOfs += sizeof(uint32_t); } } } @@ -268,13 +331,13 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert if(stateBits & ATTR_NORMAL) { - dataSize += sizeof(uint8_t) * 4; + dataSize += sizeof(uint32_t); } #ifdef USE_VERT_TANGENT_SPACE if(stateBits & ATTR_TANGENT) { - dataSize += sizeof(uint8_t) * 4; + dataSize += sizeof(uint32_t); } #endif @@ -295,7 +358,7 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert if(stateBits & ATTR_LIGHTDIRECTION) { - dataSize += sizeof(verts[0].lightdir); + dataSize += sizeof(uint32_t); } // create VBO @@ -314,14 +377,14 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert vbo->ofs_lightdir = 0; vbo->stride_xyz = sizeof(verts[0].xyz); - vbo->stride_normal = sizeof(uint8_t) * 4; + vbo->stride_normal = sizeof(uint32_t); #ifdef USE_VERT_TANGENT_SPACE - vbo->stride_tangent = sizeof(uint8_t) * 4; + vbo->stride_tangent = sizeof(uint32_t); #endif vbo->stride_vertexcolor = sizeof(verts[0].vertexColors); vbo->stride_st = sizeof(verts[0].st); vbo->stride_lightmap = sizeof(verts[0].lightmap); - vbo->stride_lightdir = sizeof(verts[0].lightdir); + vbo->stride_lightdir = sizeof(uint32_t); //ri.Printf(PRINT_ALL, "2CreateVBO: %d, %d %d %d %d %d, %d %d %d %d %d\n", dataSize, vbo->ofs_xyz, vbo->ofs_normal, vbo->ofs_st, vbo->ofs_lightmap, vbo->ofs_vertexcolor, //vbo->stride_xyz, vbo->stride_normal, vbo->stride_st, vbo->stride_lightmap, vbo->stride_vertexcolor); @@ -339,14 +402,11 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert vbo->ofs_normal = dataOfs; for (i = 0; i < numVertexes; i++) { - uint8_t *p = data + dataOfs; + uint32_t *p = (uint32_t *)(data + dataOfs); - p[0] = (uint8_t)(verts[i].normal[0] * 127.5f + 128.0f); - p[1] = (uint8_t)(verts[i].normal[1] * 127.5f + 128.0f); - p[2] = (uint8_t)(verts[i].normal[2] * 127.5f + 128.0f); - p[3] = 0; + *p = R_VboPackNormal(verts[i].normal); - dataOfs += sizeof(uint8_t) * 4; + dataOfs += sizeof(uint32_t); } } @@ -357,17 +417,11 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert vbo->ofs_tangent = dataOfs; for (i = 0; i < numVertexes; i++) { - vec3_t nxt; - uint8_t *p = data + dataOfs; + uint32_t *p = (uint32_t *)(data + dataOfs); - CrossProduct(verts[i].normal, verts[i].tangent, nxt); + *p = R_VboPackTangent(verts[i].tangent); - p[0] = (uint8_t)(verts[i].tangent[0] * 127.5f + 128.0f); - p[1] = (uint8_t)(verts[i].tangent[1] * 127.5f + 128.0f); - p[2] = (uint8_t)(verts[i].tangent[2] * 127.5f + 128.0f); - p[3] = (uint8_t)(verts[i].tangent[3] * 127.5f + 128.0f); - - dataOfs += sizeof(uint8_t) * 4; + dataOfs += sizeof(uint32_t); } } #endif @@ -411,8 +465,11 @@ VBO_t *R_CreateVBO2(const char *name, int numVertexes, srfVert_t * vert vbo->ofs_lightdir = dataOfs; for (i = 0; i < numVertexes; i++) { - memcpy(data + dataOfs, &verts[i].lightdir, sizeof(verts[i].lightdir)); - dataOfs += sizeof(verts[i].lightdir); + uint32_t *p = (uint32_t *)(data + dataOfs); + + *p = R_VboPackNormal(verts[i].lightdir); + + dataOfs += sizeof(uint32_t); } } } diff --git a/code/sdl/sdl_glimp.c b/code/sdl/sdl_glimp.c index 166ce52d..cd9581bf 100644 --- a/code/sdl/sdl_glimp.c +++ b/code/sdl/sdl_glimp.c @@ -712,8 +712,8 @@ void GLimp_Init( void ) r_allowSoftwareGL = ri.Cvar_Get( "r_allowSoftwareGL", "0", CVAR_LATCH ); r_sdlDriver = ri.Cvar_Get( "r_sdlDriver", "", CVAR_ROM ); - r_allowResize = ri.Cvar_Get( "r_allowResize", "0", CVAR_ARCHIVE ); - r_centerWindow = ri.Cvar_Get( "r_centerWindow", "0", CVAR_ARCHIVE ); + r_allowResize = ri.Cvar_Get( "r_allowResize", "0", CVAR_ARCHIVE | CVAR_LATCH ); + r_centerWindow = ri.Cvar_Get( "r_centerWindow", "0", CVAR_ARCHIVE | CVAR_LATCH ); if( ri.Cvar_VariableIntegerValue( "com_abnormalExit" ) ) { diff --git a/code/server/server.h b/code/server/server.h index b577591f..407d6ab6 100644 --- a/code/server/server.h +++ b/code/server/server.h @@ -194,7 +194,7 @@ typedef struct client_s { #endif int oldServerTime; - qboolean csUpdated[MAX_CONFIGSTRINGS+1]; + qboolean csUpdated[MAX_CONFIGSTRINGS]; #ifdef LEGACY_PROTOCOL qboolean compat; diff --git a/code/server/sv_init.c b/code/server/sv_init.c index 71c789f6..7332180b 100644 --- a/code/server/sv_init.c +++ b/code/server/sv_init.c @@ -82,7 +82,7 @@ void SV_UpdateConfigstrings(client_t *client) { int index; - for( index = 0; index <= MAX_CONFIGSTRINGS; index++ ) { + for( index = 0; index < MAX_CONFIGSTRINGS; index++ ) { // if the CS hasn't changed since we went to CS_PRIMED, ignore if(!client->csUpdated[index]) continue; @@ -422,11 +422,6 @@ void SV_SpawnServer( char *server, qboolean killBots ) { // clear the whole hunk because we're (re)loading the server Hunk_Clear(); -#ifndef DEDICATED - // Restart renderer - CL_StartHunkUsers( qtrue ); -#endif - // clear collision map data CM_ClearMap(); @@ -614,6 +609,14 @@ void SV_SpawnServer( char *server, qboolean killBots ) { Hunk_SetMark(); +#ifndef DEDICATED + if ( com_dedicated->integer ) { + // restart renderer in order to show console for dedicated servers + // launched through the regular binary + CL_StartHunkUsers( qtrue ); + } +#endif + Com_Printf ("-----------------------------------\n"); } diff --git a/code/ui/ui_atoms.c b/code/ui/ui_atoms.c index 16ce6f2f..cdc4aeb0 100644 --- a/code/ui/ui_atoms.c +++ b/code/ui/ui_atoms.c @@ -351,6 +351,7 @@ qboolean UI_ConsoleCommand( int realTime ) { if ( Q_stricmp (cmd, "ui_test") == 0 ) { UI_ShowPostGame(qtrue); + return qtrue; } if ( Q_stricmp (cmd, "ui_report") == 0 ) { diff --git a/make-macosx-app.sh b/make-macosx-app.sh index 7768964a..7e484dd4 100755 --- a/make-macosx-app.sh +++ b/make-macosx-app.sh @@ -221,13 +221,13 @@ for ARCH in $SEARCH_ARCHS; do IOQ3_UI_ARCHS="${BUILT_PRODUCTS_DIR}/${BASEDIR}/${IOQ3_UI} ${IOQ3_UI_ARCHS}" fi # missionpack - if [ -e ${BUILT_PRODUCTS_DIR}/${BASEDIR}/${IOQ3_MP_CGAME} ]; then + if [ -e ${BUILT_PRODUCTS_DIR}/${MISSIONPACKDIR}/${IOQ3_CGAME} ]; then IOQ3_MP_CGAME_ARCHS="${BUILT_PRODUCTS_DIR}/${MISSIONPACKDIR}/${IOQ3_CGAME} ${IOQ3_MP_CGAME_ARCHS}" fi - if [ -e ${BUILT_PRODUCTS_DIR}/${BASEDIR}/${IOQ3_MP_GAME} ]; then + if [ -e ${BUILT_PRODUCTS_DIR}/${MISSIONPACKDIR}/${IOQ3_GAME} ]; then IOQ3_MP_GAME_ARCHS="${BUILT_PRODUCTS_DIR}/${MISSIONPACKDIR}/${IOQ3_GAME} ${IOQ3_MP_GAME_ARCHS}" fi - if [ -e ${BUILT_PRODUCTS_DIR}/${BASEDIR}/${IOQ3_MP_UI} ]; then + if [ -e ${BUILT_PRODUCTS_DIR}/${MISSIONPACKDIR}/${IOQ3_UI} ]; then IOQ3_MP_UI_ARCHS="${BUILT_PRODUCTS_DIR}/${MISSIONPACKDIR}/${IOQ3_UI} ${IOQ3_MP_UI_ARCHS}" fi diff --git a/misc/msvc/opengl1.vcproj b/misc/msvc/opengl1.vcproj index e23eddc0..6dc6973f 100644 --- a/misc/msvc/opengl1.vcproj +++ b/misc/msvc/opengl1.vcproj @@ -609,23 +609,23 @@ >