Merge branch 'master' into sdl2

Makefile

@@ -240,9 +240,9 @@ UIDIR=$(MOUNT_DIR)/ui
 Q3UIDIR=$(MOUNT_DIR)/q3_ui
 JPDIR=$(MOUNT_DIR)/jpeg-8c
 SPEEXDIR=$(MOUNT_DIR)/libspeex
-OGGDIR=$(MOUNT_DIR)/libogg-1.3.0
+OGGDIR=$(MOUNT_DIR)/libogg-1.3.1
-OPUSDIR=$(MOUNT_DIR)/opus-1.0.2
+OPUSDIR=$(MOUNT_DIR)/opus-1.1
-OPUSFILEDIR=$(MOUNT_DIR)/opusfile-0.2
+OPUSFILEDIR=$(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
 

README.md

@@ -783,4 +783,4 @@ Significant contributions from
   * Aaron Gyes <floam@aaron.gy>
 
 
-[![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)

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;
 	}

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);

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);

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 );

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)
+	if ( s_muted->integer ) {
-		intVolume = 0;
+		intVolumeLeft = intVolumeRight = 0;
-	else
+	} else {
-		intVolume = 256 * volume * s_volume->value;
+		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].left = ((short *)data)[i*2] * intVolumeLeft;
-				rawsamples[dst].right = ((short *)data)[i*2+1] * intVolume;
+				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].left = ((short *)data)[src*2] * intVolumeLeft;
-				rawsamples[dst].right = ((short *)data)[src*2+1] * intVolume;
+				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].left = ((short *)data)[src] * intVolumeLeft;
-			rawsamples[dst].right = ((short *)data)[src] * intVolume;
+			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].left = ((char *)data)[src*2] * intVolumeLeft;
-			rawsamples[dst].right = ((char *)data)[src*2+1] * intVolume;
+			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].left = (((byte *)data)[src]-128) * intVolumeLeft;
-			rawsamples[dst].right = (((byte *)data)[src]-128) * intVolume;
+			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 ) {
+	Q_strncpyz( s_backgroundLoop, loop, sizeof( s_backgroundLoop ) );
-		s_backgroundLoop[0] = 0;
-	} else {
-		Q_strncpyz( s_backgroundLoop, loop, sizeof( s_backgroundLoop ) );
-	}
 
 	S_OpenBackgroundStream( intro );
 }

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;

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;
+	outcount = samples / stepscale;
-	sfx->soundLength = outcount;
 
 	samplefrac = 0;
-	fracstep = stepscale * 256;
+	fracstep = stepscale * 256 * channels;
 	chunk = sfx->soundData;
 
 	for (i=0 ; i<outcount ; i++)
 	{
 		srcsample = samplefrac >> 8;
 		samplefrac += fracstep;
-		if( inwidth == 2 ) {
+		for (j=0 ; j<channels ; j++)
-			sample = ( ((short *)data)[srcsample] );
+		{
-		} else {
+			if( inwidth == 2 ) {
-			sample = (int)( (unsigned char)(data[srcsample]) - 128) << 8;
+				sample = ( ((short *)data)[srcsample+j] );
-		}
-		part  = (i&(SND_CHUNK_SIZE-1));
-		if (part == 0) {
-			sndBuffer	*newchunk;
-			newchunk = SND_malloc();
-			if (chunk == NULL) {
-				sfx->soundData = newchunk;
 			} else {
-				chunk->next = 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<outcount ; i++)
 	{
 		srcsample = samplefrac >> 8;
 		samplefrac += fracstep;
-		if( inwidth == 2 ) {
+		for (j=0 ; j<channels ; j++)
-			sample = LittleShort ( ((short *)data)[srcsample] );
+		{
-		} else {
+			if( inwidth == 2 ) {
-			sample = (int)( (unsigned char)(data[srcsample]) - 128) << 8;
+				sample = LittleShort ( ((short *)data)[srcsample+j] );
+			} else {
+				sample = (int)( (unsigned char)(data[srcsample+j]) - 128) << 8;
+			}
+			sfx[i*channels+j] = sample;
 		}
-		sfx[i] = sample;
 	}
 	return outcount;
 }
@@ -221,7 +228,7 @@ qboolean S_LoadSound( sfx_t *sfx )
 		Com_DPrintf(S_COLOR_YELLOW "WARNING: %s is not a 22kHz audio file\n", sfx->soundName);
 	}
 
-	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);

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 ; i<count ; i++ ) {
 
 			aoff = ooff;
-			ooff = ooff + ch->dopplerScale;
+			ooff = ooff + ch->dopplerScale * sc->soundChannels;
 			boff = ooff;
-			fdata = 0;
+			fdata[0] = fdata[1] = 0;
-			for (j=aoff; j<boff; j++) {
+			for (j=aoff; j<boff; j += sc->soundChannels) {
 				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);
+			fdiv = 256 * (boff-aoff) / sc->soundChannels;
-			samp[i].left += (fdata * fleftvol)/fdiv;
+			samp[i].left += (fdata[0] * fleftvol)/fdiv;
-			samp[i].right += (fdata * frightvol)/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<count ; i++ ) {
 			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) {
@@ -447,10 +477,10 @@ static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int cou
 		for ( i=0 ; i<count ; i++ ) {
 
 			aoff = ooff;
-			ooff = ooff + ch->dopplerScale;
+			ooff = ooff + ch->dopplerScale * sc->soundChannels;
 			boff = ooff;
-			fdata = 0;
+			fdata[0] = fdata[1] = 0;
-			for (j=aoff; j<boff; j++) {
+			for (j=aoff; j<boff; j += sc->soundChannels) {
 				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);
+			fdiv = 256 * (boff-aoff) / sc->soundChannels;
-			samp[i].left += (fdata * fleftvol)/fdiv;
+			samp[i].left += (fdata[0] * fleftvol)/fdiv;
-			samp[i].right += (fdata * frightvol)/fdiv;
+			samp[i].right += (fdata[1] * frightvol)/fdiv;
 		}
 	}
 }

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);
+				while((curlen = strlen(inputdevicelist)))
-				Q_strcat(inputdevicenames, sizeof(inputdevicenames), "\n");
+				{
-				inputdevicelist += curlen + 1;
+					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 )
 			{

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 */

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 <stdlib.h>
+#include <limits.h>
 #include <string.h>
 #include <ogg/ogg.h>
 
@@ -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){
+static int _os_body_expand(ogg_stream_state *os,long needed){
-  if(os->body_storage<=os->body_fill+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)*
+    if(os->body_storage>LONG_MAX-needed){
-                     sizeof(*os->body_data));
+      ogg_stream_clear(os);
+      return -1;
+    }
+    body_storage=os->body_storage+needed;
+    if(body_storage<LONG_MAX-1024)body_storage+=1024;
+    ret=_ogg_realloc(os->body_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){
+static int _os_lacing_expand(ogg_stream_state *os,long needed){
-  if(os->lacing_storage<=os->lacing_fill+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)*
+    if(os->lacing_storage>LONG_MAX-needed){
-                     sizeof(*os->lacing_vals));
+      ogg_stream_clear(os);
+      return -1;
+    }
+    lacing_storage=os->lacing_storage+needed;
+    if(lacing_storage<LONG_MAX-32)lacing_storage+=32;
+    ret=_ogg_realloc(os->lacing_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){

code/opus-1.0.2/celt/celt_lpc.c

@@ -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<SHR32(ac[0],10))
-            break;
-#else
-         if (error<.001f*ac[0])
-            break;
-#endif
-      }
-   }
-#ifdef FIXED_POINT
-   for (i=0;i<p;i++)
-      _lpc[i] = ROUND16(lpc[i],16);
-#endif
-}
-
-void celt_fir(const opus_val16 *x,
-         const opus_val16 *num,
-         opus_val16 *y,
-         int N,
-         int ord,
-         opus_val16 *mem)
-{
-   int i,j;
-
-   for (i=0;i<N;i++)
-   {
-      opus_val32 sum = SHL32(EXTEND32(x[i]), SIG_SHIFT);
-      for (j=0;j<ord;j++)
-      {
-         sum += MULT16_16(num[j],mem[j]);
-      }
-      for (j=ord-1;j>=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<N;i++)
-   {
-      opus_val32 sum = x[i];
-      for (j=0;j<ord;j++)
-      {
-         sum -= MULT16_16(den[j],mem[j]);
-      }
-      for (j=ord-1;j>=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<n;i++)
-      xx[i] = x[i];
-   for (i=0;i<overlap;i++)
-   {
-      xx[i] = MULT16_16_Q15(x[i],window[i]);
-      xx[n-i-1] = MULT16_16_Q15(x[n-i-1],window[i]);
-   }
-#ifdef FIXED_POINT
-   {
-      opus_val32 ac0=0;
-      int shift;
-      for(i=0;i<n;i++)
-         ac0 += SHR32(MULT16_16(xx[i],xx[i]),9);
-      ac0 += 1+n;
-
-      shift = celt_ilog2(ac0)-30+10;
-      shift = (shift+1)/2;
-      for(i=0;i<n;i++)
-         xx[i] = VSHR32(xx[i], shift);
-   }
-#endif
-   while (lag>=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;
-}

code/opus-1.0.2/celt/cwrs.c

@@ -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)<<frac;
-    /*Note that we always need one iteration, since the rounding up above means
-       that we might need to adjust the integer part of the logarithm.*/
-    do{
-      int b;
-      b=(int)(val>>16);
-      l+=b<<frac;
-      val=(val+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)<<frac;
-}
-#endif
-
-#ifndef SMALL_FOOTPRINT
-
-#define MASK32 (0xFFFFFFFF)
-
-/*INV_TABLE[i] holds the multiplicative inverse of (2*i+1) mod 2**32.*/
-static const opus_uint32 INV_TABLE[53]={
-  0x00000001,0xAAAAAAAB,0xCCCCCCCD,0xB6DB6DB7,
-  0x38E38E39,0xBA2E8BA3,0xC4EC4EC5,0xEEEEEEEF,
-  0xF0F0F0F1,0x286BCA1B,0x3CF3CF3D,0xE9BD37A7,
-  0xC28F5C29,0x684BDA13,0x4F72C235,0xBDEF7BDF,
-  0x3E0F83E1,0x8AF8AF8B,0x914C1BAD,0x96F96F97,
-  0xC18F9C19,0x2FA0BE83,0xA4FA4FA5,0x677D46CF,
-  0x1A1F58D1,0xFAFAFAFB,0x8C13521D,0x586FB587,
-  0xB823EE09,0xA08AD8F3,0xC10C9715,0xBEFBEFBF,
-  0xC0FC0FC1,0x07A44C6B,0xA33F128D,0xE327A977,
-  0xC7E3F1F9,0x962FC963,0x3F2B3885,0x613716AF,
-  0x781948B1,0x2B2E43DB,0xFCFCFCFD,0x6FD0EB67,
-  0xFA3F47E9,0xD2FD2FD3,0x3F4FD3F5,0xD4E25B9F,
-  0x5F02A3A1,0xBF5A814B,0x7C32B16D,0xD3431B57,
-  0xD8FD8FD9,
-};
-
-/*Computes (_a*_b-_c)/(2*_d+1) when the quotient is known to be exact.
-  _a, _b, _c, and _d may be arbitrary so long as the arbitrary precision result
-   fits in 32 bits, but currently the table for multiplicative inverses is only
-   valid for _d<=52.*/
-static inline opus_uint32 imusdiv32odd(opus_uint32 _a,opus_uint32 _b,
- opus_uint32 _c,int _d){
-  celt_assert(_d<=52);
-  return (_a*_b-_c)*INV_TABLE[_d]&MASK32;
-}
-
-/*Computes (_a*_b-_c)/_d when the quotient is known to be exact.
-  _d does not actually have to be even, but imusdiv32odd will be faster when
-   it's odd, so you should use that instead.
-  _a and _d are assumed to be small (e.g., _a*_d fits in 32 bits; currently the
-   table for multiplicative inverses is only valid for _d<=54).
-  _b and _c may be arbitrary so long as the arbitrary precision reuslt fits in
-   32 bits.*/
-static inline opus_uint32 imusdiv32even(opus_uint32 _a,opus_uint32 _b,
- opus_uint32 _c,int _d){
-  opus_uint32 inv;
-  int           mask;
-  int           shift;
-  int           one;
-  celt_assert(_d>0);
-  celt_assert(_d<=54);
-  shift=EC_ILOG(_d^(_d-1));
-  inv=INV_TABLE[(_d-1)>>shift];
-  shift--;
-  one=1<<shift;
-  mask=one-1;
-  return (_a*(_b>>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);
-    for(k=2;k<_n;k++)unext(_u+1,_k+1,1);
-  }
-#ifndef SMALL_FOOTPRINT
-  else{
-    opus_uint32 um1;
-    opus_uint32 n2m1;
-    _u[2]=n2m1=um1=(_n<<1)-1;
-    for(k=3;k<len;k++){
-      /*U(N,K) = ((2*N-1)*U(N,K-1)-U(N,K-2))/(K-1) + U(N,K-2)*/
-      _u[k]=um2=imusdiv32even(n2m1,um1,um2,k-1)+um2;
-      if(++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
-}

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) )

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
+#ifdef OPUS_ARM_INLINE_EDSP
-#include "fixed_arm5e.h"
+#include "arm/fixed_armv5e.h"
-#elif defined (ARM4_ASM)
+#elif defined (OPUS_ARM_INLINE_ASM)
-#include "fixed_arm4.h"
+#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))

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/@ $&/;
+    s/:OR:/|/g;
+    s/:SHL:/<</g;
+    s/:SHR:/>>/g;
+    s/:AND:/&/g;
+    s/:LAND:/&&/g;
+    s/CPSR/cpsr/;
+    s/SPSR/spsr/;
+    s/ALIGN$/.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");
+}

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

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 <windows.h>
+
+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

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

code/opus-1.1/celt/arm/armopts.s.in

@@ -1,5 +1,4 @@
-/* Copyright (c) 2011 Xiph.Org Foundation, Skype Limited
+/* Copyright (C) 2013 Mozilla Corporation */
-   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
@@ -25,23 +24,14 @@
    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */
 
-#ifdef HAVE_CONFIG_H
+; Set the following to 1 if we have EDSP instructions
-#include "config.h"
+;  (LDRD/STRD, etc., ARMv5E and later).
-#endif
+OPUS_ARM_MAY_HAVE_EDSP  * @OPUS_ARM_MAY_HAVE_EDSP@
 
-#include "opus.h"
+; Set the following to 1 if we have ARMv6 media instructions.
-#include "opus_private.h"
+OPUS_ARM_MAY_HAVE_MEDIA * @OPUS_ARM_MAY_HAVE_MEDIA@
 
-int encode_size(int size, unsigned char *data)
+; Set the following to 1 if we have NEON (some ARMv7)
-{
+OPUS_ARM_MAY_HAVE_NEON  * @OPUS_ARM_MAY_HAVE_NEON@
-   if (size < 252)
-   {
-      data[0] = size;
-      return 1;
-   } else {
-      data[0] = 252+(size&0x3);
-      data[1] = (size-(int)data[0])>>2;
-      return 2;
-   }
-}
 
+END

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

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

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

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 */

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 */

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

code/opus-1.1/celt/bands.h

@@ -39,7 +39,7 @@
 /** Compute the amplitude (sqrt energy) in each of the bands
  * @param m Mode data
  * @param X Spectrum
- * @param bands Square root of the energy for each band (returned)
+ * @param bandE Square root of the energy for each band (returned)
  */
 void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int M);
 
@@ -49,16 +49,17 @@ void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *band
     equal to 1
  * @param m Mode data
  * @param X Spectrum (returned normalised)
- * @param bands Square root of the energy for each band
+ * @param bandE Square root of the energy for each band
  */
 void normalise_bands(const CELTMode *m, const celt_sig * OPUS_RESTRICT freq, celt_norm * OPUS_RESTRICT X, const celt_ener *bandE, int end, int C, int M);
 
 /** Denormalise each band of X to restore full amplitude
  * @param m Mode data
  * @param X Spectrum (returned de-normalised)
- * @param bands Square root of the energy for each band
+ * @param bandE Square root of the energy for each band
  */
-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 *bandE, int start, int end, int C, int M);
 
 #define SPREAD_NONE       (0)
 #define SPREAD_LIGHT      (1)
@@ -76,14 +77,30 @@ void measure_norm_mse(const CELTMode *m, float *X, float *X0, float *bandE, floa
 void haar1(celt_norm *X, int N0, int stride);
 
 /** Quantisation/encoding of the residual spectrum
+ * @param encode flag that indicates whether we're encoding (1) or decoding (0)
  * @param m Mode data
+ * @param start First band to process
+ * @param end Last band to process + 1
  * @param X Residual (normalised)
+ * @param Y Residual (normalised) for second channel (or NULL for mono)
+ * @param collapse_masks Anti-collapse tracking mask
+ * @param bandE Square root of the energy for each band
+ * @param pulses Bit allocation (per band) for PVQ
+ * @param shortBlocks Zero for long blocks, non-zero for short blocks
+ * @param spread Amount of spreading to use
+ * @param dual_stereo Zero for MS stereo, non-zero for dual stereo
+ * @param intensity First band to use intensity stereo
+ * @param tf_res Time-frequency resolution change
  * @param total_bits Total number of bits that can be used for the frame (including the ones already spent)
- * @param enc Entropy encoder
+ * @param balance Number of unallocated bits
+ * @param en Entropy coder state
+ * @param LM log2() of the number of 2.5 subframes in the frame
+ * @param codedBands Last band to receive bits + 1
+ * @param seed Random generator seed
  */
 void quant_all_bands(int encode, const CELTMode *m, int start, int end,
       celt_norm * X, celt_norm * Y, unsigned char *collapse_masks, const celt_ener *bandE, int *pulses,
-      int time_domain, int fold, int dual_stereo, int intensity, int *tf_res,
+      int shortBlocks, int spread, int dual_stereo, int intensity, int *tf_res,
       opus_int32 total_bits, opus_int32 balance, ec_ctx *ec, int M, int codedBands, opus_uint32 *seed);
 
 void anti_collapse(const CELTMode *m, celt_norm *X_, unsigned char *collapse_masks, int LM, int C, int size,
@@ -92,4 +109,6 @@ void anti_collapse(const CELTMode *m, celt_norm *X_, unsigned char *collapse_mas
 
 opus_uint32 celt_lcg_rand(opus_uint32 seed);
 
+int hysteresis_decision(opus_val16 val, const opus_val16 *thresholds, const opus_val16 *hysteresis, int N, int prev);
+
 #endif /* BANDS_H */

code/opus-1.1/celt/celt.c

@@ -0,0 +1,223 @@
+/* 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 <math.h>
+#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 <stdarg.h>
+#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;i<N;i++)
+   {
+      x0=x[i-T+2];
+      y[i] = x[i]
+               + MULT16_32_Q15(g10,x2)
+               + MULT16_32_Q15(g11,ADD32(x1,x3))
+               + MULT16_32_Q15(g12,ADD32(x0,x4));
+      x4=x3;
+      x3=x2;
+      x2=x1;
+      x1=x0;
+   }
+
+}
+#endif
+
+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)
+{
+   int i;
+   /* printf ("%d %d %f %f\n", T0, T1, g0, g1); */
+   opus_val16 g00, g01, g02, g10, g11, g12;
+   opus_val32 x0, x1, x2, x3, x4;
+   static const opus_val16 gains[3][3] = {
+         {QCONST16(0.3066406250f, 15), QCONST16(0.2170410156f, 15), QCONST16(0.1296386719f, 15)},
+         {QCONST16(0.4638671875f, 15), QCONST16(0.2680664062f, 15), QCONST16(0.f, 15)},
+         {QCONST16(0.7998046875f, 15), QCONST16(0.1000976562f, 15), QCONST16(0.f, 15)}};
+
+   if (g0==0 && g1==0)
+   {
+      /* OPT: Happens to work without the OPUS_MOVE(), but only because the current encoder already copies x to y */
+      if (x!=y)
+         OPUS_MOVE(y, x, N);
+      return;
+   }
+   g00 = MULT16_16_Q15(g0, gains[tapset0][0]);
+   g01 = MULT16_16_Q15(g0, gains[tapset0][1]);
+   g02 = MULT16_16_Q15(g0, gains[tapset0][2]);
+   g10 = MULT16_16_Q15(g1, gains[tapset1][0]);
+   g11 = MULT16_16_Q15(g1, gains[tapset1][1]);
+   g12 = MULT16_16_Q15(g1, gains[tapset1][2]);
+   x1 = x[-T1+1];
+   x2 = x[-T1  ];
+   x3 = x[-T1-1];
+   x4 = x[-T1-2];
+   for (i=0;i<overlap;i++)
+   {
+      opus_val16 f;
+      x0=x[i-T1+2];
+      f = MULT16_16_Q15(window[i],window[i]);
+      y[i] = x[i]
+               + MULT16_32_Q15(MULT16_16_Q15((Q15ONE-f),g00),x[i-T0])
+               + MULT16_32_Q15(MULT16_16_Q15((Q15ONE-f),g01),ADD32(x[i-T0+1],x[i-T0-1]))
+               + MULT16_32_Q15(MULT16_16_Q15((Q15ONE-f),g02),ADD32(x[i-T0+2],x[i-T0-2]))
+               + MULT16_32_Q15(MULT16_16_Q15(f,g10),x2)
+               + MULT16_32_Q15(MULT16_16_Q15(f,g11),ADD32(x1,x3))
+               + MULT16_32_Q15(MULT16_16_Q15(f,g12),ADD32(x0,x4));
+      x4=x3;
+      x3=x2;
+      x2=x1;
+      x1=x0;
+
+   }
+   if (g1==0)
+   {
+      /* OPT: Happens to work without the OPUS_MOVE(), but only because the current encoder already copies x to y */
+      if (x!=y)
+         OPUS_MOVE(y+overlap, x+overlap, N-overlap);
+      return;
+   }
+
+   /* Compute the part with the constant filter. */
+   comb_filter_const(y+i, x+i, T1, N-i, g10, g11, g12);
+}
+
+const signed char tf_select_table[4][8] = {
+      {0, -1, 0, -1,    0,-1, 0,-1},
+      {0, -1, 0, -2,    1, 0, 1,-1},
+      {0, -2, 0, -3,    2, 0, 1,-1},
+      {0, -2, 0, -3,    3, 0, 1,-1},
+};
+
+
+void init_caps(const CELTMode *m,int *cap,int LM,int C)
+{
+   int i;
+   for (i=0;i<m->nbEBands;i++)
+   {
+      int N;
+      N=(m->eBands[i+1]-m->eBands[i])<<LM;
+      cap[i] = (m->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
+          ;
+}

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

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<SHR32(ac[0],10))
+            break;
+#else
+         if (error<.001f*ac[0])
+            break;
+#endif
+      }
+   }
+#ifdef FIXED_POINT
+   for (i=0;i<p;i++)
+      _lpc[i] = ROUND16(lpc[i],16);
+#endif
+}
+
+void celt_fir(const opus_val16 *_x,
+         const opus_val16 *num,
+         opus_val16 *_y,
+         int N,
+         int ord,
+         opus_val16 *mem)
+{
+   int i,j;
+   VARDECL(opus_val16, rnum);
+   VARDECL(opus_val16, x);
+   SAVE_STACK;
+
+   ALLOC(rnum, ord, opus_val16);
+   ALLOC(x, N+ord, opus_val16);
+   for(i=0;i<ord;i++)
+      rnum[i] = num[ord-i-1];
+   for(i=0;i<ord;i++)
+      x[i] = mem[ord-i-1];
+   for (i=0;i<N;i++)
+      x[i+ord]=_x[i];
+   for(i=0;i<ord;i++)
+      mem[i] = _x[N-i-1];
+#ifdef SMALL_FOOTPRINT
+   for (i=0;i<N;i++)
+   {
+      opus_val32 sum = SHL32(EXTEND32(_x[i]), SIG_SHIFT);
+      for (j=0;j<ord;j++)
+      {
+         sum = MAC16_16(sum,rnum[j],x[i+j]);
+      }
+      _y[i] = SATURATE16(PSHR32(sum, SIG_SHIFT));
+   }
+#else
+   for (i=0;i<N-3;i+=4)
+   {
+      opus_val32 sum[4]={0,0,0,0};
+      xcorr_kernel(rnum, x+i, sum, ord);
+      _y[i  ] = SATURATE16(ADD32(EXTEND32(_x[i  ]), PSHR32(sum[0], SIG_SHIFT)));
+      _y[i+1] = SATURATE16(ADD32(EXTEND32(_x[i+1]), PSHR32(sum[1], SIG_SHIFT)));
+      _y[i+2] = SATURATE16(ADD32(EXTEND32(_x[i+2]), PSHR32(sum[2], SIG_SHIFT)));
+      _y[i+3] = SATURATE16(ADD32(EXTEND32(_x[i+3]), PSHR32(sum[3], SIG_SHIFT)));
+   }
+   for (;i<N;i++)
+   {
+      opus_val32 sum = 0;
+      for (j=0;j<ord;j++)
+         sum = MAC16_16(sum,rnum[j],x[i+j]);
+      _y[i] = SATURATE16(ADD32(EXTEND32(_x[i]), PSHR32(sum, SIG_SHIFT)));
+   }
+#endif
+   RESTORE_STACK;
+}
+
+void celt_iir(const opus_val32 *_x,
+         const opus_val16 *den,
+         opus_val32 *_y,
+         int N,
+         int ord,
+         opus_val16 *mem)
+{
+#ifdef SMALL_FOOTPRINT
+   int i,j;
+   for (i=0;i<N;i++)
+   {
+      opus_val32 sum = _x[i];
+      for (j=0;j<ord;j++)
+      {
+         sum -= MULT16_16(den[j],mem[j]);
+      }
+      for (j=ord-1;j>=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;i<ord;i++)
+      rden[i] = den[ord-i-1];
+   for(i=0;i<ord;i++)
+      y[i] = -mem[ord-i-1];
+   for(;i<N+ord;i++)
+      y[i]=0;
+   for (i=0;i<N-3;i+=4)
+   {
+      /* Unroll by 4 as if it were an FIR filter */
+      opus_val32 sum[4];
+      sum[0]=_x[i];
+      sum[1]=_x[i+1];
+      sum[2]=_x[i+2];
+      sum[3]=_x[i+3];
+      xcorr_kernel(rden, y+i, sum, ord);
+
+      /* Patch up the result to compensate for the fact that this is an IIR */
+      y[i+ord  ] = -ROUND16(sum[0],SIG_SHIFT);
+      _y[i  ] = sum[0];
+      sum[1] = MAC16_16(sum[1], y[i+ord  ], den[0]);
+      y[i+ord+1] = -ROUND16(sum[1],SIG_SHIFT);
+      _y[i+1] = sum[1];
+      sum[2] = MAC16_16(sum[2], y[i+ord+1], den[0]);
+      sum[2] = MAC16_16(sum[2], y[i+ord  ], den[1]);
+      y[i+ord+2] = -ROUND16(sum[2],SIG_SHIFT);
+      _y[i+2] = sum[2];
+
+      sum[3] = MAC16_16(sum[3], y[i+ord+2], den[0]);
+      sum[3] = MAC16_16(sum[3], y[i+ord+1], den[1]);
+      sum[3] = MAC16_16(sum[3], y[i+ord  ], den[2]);
+      y[i+ord+3] = -ROUND16(sum[3],SIG_SHIFT);
+      _y[i+3] = sum[3];
+   }
+   for (;i<N;i++)
+   {
+      opus_val32 sum = _x[i];
+      for (j=0;j<ord;j++)
+         sum -= MULT16_16(rden[j],y[i+j]);
+      y[i+ord] = ROUND16(sum,SIG_SHIFT);
+      _y[i] = sum;
+   }
+   for(i=0;i<ord;i++)
+      mem[i] = _y[N-i-1];
+   RESTORE_STACK;
+#endif
+}
+
+int _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,
+                   int          arch
+                  )
+{
+   opus_val32 d;
+   int i, k;
+   int fastN=n-lag;
+   int shift;
+   const opus_val16 *xptr;
+   VARDECL(opus_val16, xx);
+   SAVE_STACK;
+   ALLOC(xx, n, opus_val16);
+   celt_assert(n>0);
+   celt_assert(overlap>=0);
+   if (overlap == 0)
+   {
+      xptr = x;
+   } else {
+      for (i=0;i<n;i++)
+         xx[i] = x[i];
+      for (i=0;i<overlap;i++)
+      {
+         xx[i] = MULT16_16_Q15(x[i],window[i]);
+         xx[n-i-1] = MULT16_16_Q15(x[n-i-1],window[i]);
+      }
+      xptr = xx;
+   }
+   shift=0;
+#ifdef FIXED_POINT
+   {
+      opus_val32 ac0;
+      ac0 = 1+(n<<7);
+      if (n&1) ac0 += SHR32(MULT16_16(xptr[0],xptr[0]),9);
+      for(i=(n&1);i<n;i+=2)
+      {
+         ac0 += SHR32(MULT16_16(xptr[i],xptr[i]),9);
+         ac0 += SHR32(MULT16_16(xptr[i+1],xptr[i+1]),9);
+      }
+
+      shift = celt_ilog2(ac0)-30+10;
+      shift = (shift)/2;
+      if (shift>0)
+      {
+         for(i=0;i<n;i++)
+            xx[i] = PSHR32(xptr[i], shift);
+         xptr = xx;
+      } else
+         shift = 0;
+   }
+#endif
+   celt_pitch_xcorr(xptr, xptr, ac, fastN, lag+1, arch);
+   for (k=0;k<=lag;k++)
+   {
+      for (i = k+fastN, d = 0; i < n; i++)
+         d = MAC16_16(d, xptr[i], xptr[i-k]);
+      ac[k] += d;
+   }
+#ifdef FIXED_POINT
+   shift = 2*shift;
+   if (shift<=0)
+      ac[0] += SHL32((opus_int32)1, -shift);
+   if (ac[0] < 268435456)
+   {
+      int shift2 = 29 - EC_ILOG(ac[0]);
+      for (i=0;i<=lag;i++)
+         ac[i] = SHL32(ac[i], shift2);
+      shift -= shift2;
+   } else if (ac[0] >= 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;
+}

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 */

code/opus-1.1/celt/cpu_support.h

@@ -1,11 +1,5 @@
-/* Copyright (c) 2007-2008 CSIRO
+/* Copyright (c) 2010 Xiph.Org Foundation
-   Copyright (c) 2007-2009 Xiph.Org Foundation
+ * Copyright (c) 2013 Parrot */
-   Written by Jean-Marc Valin */
-/**
-   @file pitch.h
-   @brief Pitch analysis
- */
-
 /*
    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
+#ifndef CPU_SUPPORT_H
-#define PITCH_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,
+#if defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_ASM)
-      int len, int C);
+#include "arm/armcpu.h"
 
-void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTRICT y,
+/* We currently support 4 ARM variants:
-                  int len, int max_pitch, int *pitch);
+ * 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,
+#else
-      int N, int *T0, int prev_period, opus_val16 prev_gain);
+#define OPUS_ARCHMASK 0
+
+static OPUS_INLINE int opus_select_arch(void)
+{
+  return 0;
+}
+#endif
 
 #endif

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)<<frac;
+    /*Note that we always need one iteration, since the rounding up above means
+       that we might need to adjust the integer part of the logarithm.*/
+    do{
+      int b;
+      b=(int)(val>>16);
+      l+=b<<frac;
+      val=(val+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)<<frac;
+}
+#endif
+
+/*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
+  }*/
+
+#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-=p;
+        *_y++=0;
+      }
+      else{
+        /*Are the pulses in this dimension negative?*/
+        s=-(_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(++k<len);
+  for(k=2;k<_n;k++)unext(_u+1,_k+1,1);
+  return _u[_k]+_u[_k+1];
+}
+
+/*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 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 */

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.*/

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);
 }
 

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){

code/opus-1.1/celt/fixed_debug.h

@@ -33,9 +33,9 @@
 #define FIXED_DEBUG_H
 
 #include <stdio.h>
+#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)

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))

code/opus-1.1/celt/float_cast.h

@@ -101,7 +101,7 @@
         #include <math.h>
 
         /*      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);

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;

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;
 

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;i<len;i++)
+   {
+      maxval = MAX16(maxval, x[i]);
+      minval = MIN16(minval, x[i]);
+   }
+   return MAX32(EXTEND32(maxval),-EXTEND32(minval));
+}
+#endif
+
+#ifndef OVERRIDE_CELT_MAXABS32
+#ifdef FIXED_POINT
+static OPUS_INLINE opus_val32 celt_maxabs32(const opus_val32 *x, int len)
+{
+   int i;
+   opus_val32 maxval = 0;
+   opus_val32 minval = 0;
+   for (i=0;i<len;i++)
+   {
+      maxval = MAX32(maxval, x[i]);
+      minval = MIN32(minval, x[i]);
+   }
+   return MAX32(maxval, -minval);
+}
+#else
+#define celt_maxabs32(x,len) celt_maxabs16(x,len)
+#endif
+#endif
+
+
 #ifndef FIXED_POINT
 
 #define PI 3.141592653f
@@ -60,7 +95,7 @@ unsigned isqrt32(opus_uint32 _val);
          denorm, +/- inf and NaN are *not* handled */
 
 /** Base-2 log approximation (log2(x)). */
-static inline float celt_log2(float x)
+static OPUS_INLINE float celt_log2(float x)
 {
    int integer;
    float frac;
@@ -78,7 +113,7 @@ static inline float celt_log2(float x)
 }
 
 /** Base-2 exponential approximation (2^x). */
-static inline float celt_exp2(float x)
+static OPUS_INLINE float celt_exp2(float x)
 {
    int integer;
    float frac;
@@ -110,37 +145,16 @@ static inline float celt_exp2(float x)
 
 #ifndef OVERRIDE_CELT_ILOG2
 /** Integer log in base2. Undefined for zero and negative numbers */
-static inline opus_int16 celt_ilog2(opus_int32 x)
+static OPUS_INLINE opus_int16 celt_ilog2(opus_int32 x)
 {
    celt_assert2(x>0, "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;i<len;i++)
-      maxval = MAX16(maxval, ABS16(x[i]));
-   return maxval;
-}
-#endif
-
-#ifndef OVERRIDE_CELT_MAXABS32
-static inline opus_val32 celt_maxabs32(opus_val32 *x, int len)
-{
-   int i;
-   opus_val32 maxval = 0;
-   for (i=0;i<len;i++)
-      maxval = MAX32(maxval, ABS32(x[i]));
-   return maxval;
-}
-#endif
 
 /** Integer log in base2. Defined for zero, but not for negative numbers */
-static inline opus_int16 celt_zlog2(opus_val32 x)
+static OPUS_INLINE opus_int16 celt_zlog2(opus_val32 x)
 {
    return x <= 0 ? 0 : celt_ilog2(x);
 }
@@ -151,7 +165,8 @@ opus_val32 celt_sqrt(opus_val32 x);
 
 opus_val16 celt_cos_norm(opus_val32 x);
 
-static inline opus_val16 celt_log2(opus_val32 x)
+/** Base-2 logarithm approximation (log2(x)). (Q14 input, Q10 output) */
+static OPUS_INLINE opus_val16 celt_log2(opus_val32 x)
 {
    int i;
    opus_val16 n, frac;
@@ -176,8 +191,15 @@ static inline opus_val16 celt_log2(opus_val32 x)
 #define D1 22804
 #define D2 14819
 #define D3 10204
+
+static OPUS_INLINE opus_val32 celt_exp2_frac(opus_val16 x)
+{
+   opus_val16 frac;
+   frac = SHL16(x, 4);
+   return ADD16(D0, MULT16_16_Q15(frac, ADD16(D1, MULT16_16_Q15(frac, ADD16(D2 , MULT16_16_Q15(D3,frac))))));
+}
 /** Base-2 exponential approximation (2^x). (Q10 input, Q16 output) */
-static inline opus_val32 celt_exp2(opus_val16 x)
+static OPUS_INLINE opus_val32 celt_exp2(opus_val16 x)
 {
    int integer;
    opus_val16 frac;
@@ -186,8 +208,7 @@ static inline opus_val32 celt_exp2(opus_val16 x)
       return 0x7f000000;
    else if (integer < -15)
       return 0;
-   frac = SHL16(x-SHL16(integer,10),4);
+   frac = celt_exp2_frac(x-SHL16(integer,10));
-   frac = ADD16(D0, MULT16_16_Q15(frac, ADD16(D1, MULT16_16_Q15(frac, ADD16(D2 , MULT16_16_Q15(D3,frac))))));
    return VSHR32(EXTEND32(frac), -integer-2);
 }
 
@@ -204,7 +225,7 @@ opus_val32 frac_div32(opus_val32 a, opus_val32 b);
 
 /* Atan approximation using a 4th order polynomial. Input is in Q15 format
    and normalized by pi/4. Output is in Q15 format */
-static inline opus_val16 celt_atan01(opus_val16 x)
+static OPUS_INLINE opus_val16 celt_atan01(opus_val16 x)
 {
    return MULT16_16_P15(x, ADD32(M1, MULT16_16_P15(x, ADD32(M2, MULT16_16_P15(x, ADD32(M3, MULT16_16_P15(M4, x)))))));
 }
@@ -215,7 +236,7 @@ static inline opus_val16 celt_atan01(opus_val16 x)
 #undef M4
 
 /* atan2() approximation valid for positive input values */
-static inline opus_val16 celt_atan2p(opus_val16 y, opus_val16 x)
+static OPUS_INLINE opus_val16 celt_atan2p(opus_val16 y, opus_val16 x)
 {
    if (y < x)
    {

code/opus-1.1/celt/mdct.c

@@ -109,12 +109,14 @@ void clt_mdct_forward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar
    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
    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<N4-(overlap>>2);i++)
+      for(;i<N4-((overlap+3)>>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];
-
+      /* Loop to (N4+1)>>1 to handle odd N4. When N4 is odd, the
-      for(i=0;i<N4;i++)
+         middle pair will be computed twice. */
+      for(i=0;i<(N4+1)>>1;i++)
       {
          kiss_fft_scalar re, im, yr, yi;
-         re = fp[0];
+         kiss_twiddle_scalar t0, t1;
-         im = fp[1];
+         re = yp0[0];
+         im = yp0[1];
+         t0 = t[i<<shift];
+         t1 = t[(N4-i)<<shift];
          /* We'd scale up by 2 here, but instead it's done when mixing the windows */
-         yr = S_MUL(re,t[i<<shift]) - S_MUL(im,t[(N4-i)<<shift]);
+         yr = S_MUL(re,t0) - S_MUL(im,t1);
-         yi = S_MUL(im,t[i<<shift]) + S_MUL(re,t[(N4-i)<<shift]);
+         yi = S_MUL(im,t0) + S_MUL(re,t1);
+         re = yp1[0];
+         im = yp1[1];
          /* works because the cos is nearly one */
-         *fp++ = yr - S_MUL(yi,sine);
+         yp0[0] = -(yr - S_MUL(yi,sine));
-         *fp++ = yi + S_MUL(yr,sine);
+         yp1[1] = yi + S_MUL(yr,sine);
+
+         t0 = t[(N4-i-1)<<shift];
+         t1 = t[(i+1)<<shift];
+         /* We'd scale up by 2 here, but instead it's done when mixing the windows */
+         yr = S_MUL(re,t0) - S_MUL(im,t1);
+         yi = S_MUL(im,t0) + S_MUL(re,t1);
+         /* works because the cos is nearly one */
+         yp1[0] = -(yr - S_MUL(yi,sine));
+         yp0[1] = yi + S_MUL(yr,sine);
+         yp0 += 2;
+         yp1 -= 2;
       }
    }
-   /* De-shuffle the components for the middle of the window only */
+
-   {
-      const kiss_fft_scalar * OPUS_RESTRICT fp1 = f;
-      const kiss_fft_scalar * OPUS_RESTRICT fp2 = f+N2-1;
-      kiss_fft_scalar * OPUS_RESTRICT yp = f2;
-      for(i = 0; i < N4; i++)
-      {
-         *yp++ =-*fp1;
-         *yp++ = *fp2;
-         fp1 += 2;
-         fp2 -= 2;
-      }
-   }
-   out -= (N2-overlap)>>1;
    /* Mirror on both sides for TDAC */
    {
-      kiss_fft_scalar * OPUS_RESTRICT fp1 = f2+N4-1;
+      kiss_fft_scalar * OPUS_RESTRICT xp1 = out+overlap-1;
-      kiss_fft_scalar * OPUS_RESTRICT xp1 = out+N2-1;
+      kiss_fft_scalar * OPUS_RESTRICT yp1 = out;
-      kiss_fft_scalar * OPUS_RESTRICT yp1 = out+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++)
+
+      for(i = 0; i < overlap/2; i++)
       {
-         *xp1 = *fp1;
+         kiss_fft_scalar x1, x2;
-         xp1--;
+         x1 = *xp1;
-         fp1--;
+         x2 = *yp1;
-      }
+         *yp1++ = MULT16_32_Q15(*wp2, x2) - MULT16_32_Q15(*wp1, x1);
-      for(; i < N4; i++)
+         *xp1-- = MULT16_32_Q15(*wp1, x2) + MULT16_32_Q15(*wp2, x1);
-      {
-         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);
          wp1++;
          wp2--;
       }

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])<<LM >
+    208) {
+       goto failure;
+   }
+#endif
 
    mode->effEBands = mode->nbEBands;
    while (mode->eBands[mode->effEBands] > mode->shortMdctSize)

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 <string.h>
 #include <stdio.h>
 #include <stdlib.h>
 
 /** 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);
 }

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<N;i++)
+   {
+      opus_val32 sum = SHL32(EXTEND32(x[i]), SIG_SHIFT);
+      sum = MAC16_16(sum,num0,mem0);
+      sum = MAC16_16(sum,num1,mem1);
+      sum = MAC16_16(sum,num2,mem2);
+      sum = MAC16_16(sum,num3,mem3);
+      sum = MAC16_16(sum,num4,mem4);
+      mem4 = mem3;
+      mem3 = mem2;
+      mem2 = mem1;
+      mem1 = mem0;
+      mem0 = x[i];
+      y[i] = ROUND16(sum, SIG_SHIFT);
+   }
+   mem[0]=mem0;
+   mem[1]=mem1;
+   mem[2]=mem2;
+   mem[3]=mem3;
+   mem[4]=mem4;
+}
+
+
 void pitch_downsample(celt_sig * OPUS_RESTRICT x[], opus_val16 * OPUS_RESTRICT x_lp,
-      int len, int C)
+      int len, int C, int arch)
 {
    int i;
    opus_val32 ac[5];
    opus_val16 tmp=Q15ONE;
-   opus_val16 lpc[4], mem[4]={0,0,0,0};
+   opus_val16 lpc[4], mem[5]={0,0,0,0,0};
+   opus_val16 lpc2[5];
+   opus_val16 c1 = QCONST16(.8f,15);
 #ifdef FIXED_POINT
    int shift;
    opus_val32 maxabs = celt_maxabs32(x[0], len);
@@ -136,7 +180,7 @@ void pitch_downsample(celt_sig * OPUS_RESTRICT x[], opus_val16 * OPUS_RESTRICT x
    }
 
    _celt_autocorr(x_lp, ac, NULL, 0,
-                  4, len>>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);
+   /* Add a zero */
-
+   lpc2[0] = lpc[0] + QCONST16(.8f,SIG_SHIFT);
-   mem[0]=0;
+   lpc2[1] = lpc[1] + MULT16_16_Q15(c1,lpc[0]);
-   lpc[0]=QCONST16(.8f,12);
+   lpc2[2] = lpc[2] + MULT16_16_Q15(c1,lpc[1]);
-   celt_fir(x_lp, lpc, x_lp, len>>1, 1, mem);
+   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;i<max_pitch;i++)
+   {
+      opus_val32 sum = 0;
+      for (j=0;j<len;j++)
+         sum = MAC16_16(sum, x[j],y[i+j]);
+      xcorr[i] = sum;
+#ifdef FIXED_POINT
+      maxcorr = MAX32(maxcorr, sum);
+#endif
+   }
+#ifdef FIXED_POINT
+   return maxcorr;
+#endif
+}
+
+#else /* Unrolled version of the pitch correlation -- runs faster on x86 and ARM */
+
+#ifdef FIXED_POINT
+opus_val32
+#else
+void
+#endif
+celt_pitch_xcorr_c(const opus_val16 *_x, const opus_val16 *_y, opus_val32 *xcorr, int len, int max_pitch)
+{
+   int i,j;
+   /*The EDSP version requires that max_pitch is at least 1, and that _x is
+      32-bit aligned.
+     Since it's hard to put asserts in assembly, put them here.*/
+   celt_assert(max_pitch>0);
+   celt_assert((((unsigned char *)_x-(unsigned char *)NULL)&3)==0);
+#ifdef FIXED_POINT
+   opus_val32 maxcorr=1;
+#endif
+   for (i=0;i<max_pitch-3;i+=4)
+   {
+      opus_val32 sum[4]={0,0,0,0};
+      xcorr_kernel(_x, _y+i, sum, len);
+      xcorr[i]=sum[0];
+      xcorr[i+1]=sum[1];
+      xcorr[i+2]=sum[2];
+      xcorr[i+3]=sum[3];
+#ifdef FIXED_POINT
+      sum[0] = MAX32(sum[0], sum[1]);
+      sum[2] = MAX32(sum[2], sum[3]);
+      sum[0] = MAX32(sum[0], sum[2]);
+      maxcorr = MAX32(maxcorr, sum[0]);
+#endif
+   }
+   /* In case max_pitch isn't a multiple of 4, do non-unrolled version. */
+   for (;i<max_pitch;i++)
+   {
+      opus_val32 sum = 0;
+      for (j=0;j<len;j++)
+         sum = MAC16_16(sum, _x[j],_y[i+j]);
+      xcorr[i] = sum;
+#ifdef FIXED_POINT
+      maxcorr = MAX32(maxcorr, sum);
+#endif
+   }
+#ifdef FIXED_POINT
+   return maxcorr;
+#endif
+}
+
+#endif
 void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTRICT y,
-                  int len, int max_pitch, int *pitch)
+                  int len, int max_pitch, int *pitch, int arch)
 {
    int i, j;
    int lag;
@@ -179,8 +303,8 @@ void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTR
    VARDECL(opus_val16, y_lp4);
    VARDECL(opus_val32, xcorr);
 #ifdef FIXED_POINT
-   opus_val32 maxcorr=1;
+   opus_val32 maxcorr;
-   opus_val16 xmax, ymax;
+   opus_val32 xmax, ymax;
    int shift=0;
 #endif
    int offset;
@@ -204,7 +328,7 @@ void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTR
 #ifdef FIXED_POINT
    xmax = celt_maxabs16(x_lp4, len>>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<len>>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<max_pitch>>2;i++)
-   {
-      opus_val32 sum = 0;
-      for (j=0;j<len>>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;
+   ALLOC(yy_lookup, maxperiod+1, opus_val32);
-   for (i=0;i<N;i++)
+   dual_inner_prod(x, x, x-T0, N, &xx, &xy);
+   yy_lookup[0] = xx;
+   yy=xx;
+   for (i=1;i<=maxperiod;i++)
    {
-      xy = MAC16_16(xy, x[i], x[i-T0]);
+      yy = yy+MULT16_16(x[-i],x[-i])-MULT16_16(x[N-i],x[N-i]);
-      xx = MAC16_16(xx, x[i], x[i]);
+      yy_lookup[i] = MAX32(0, yy);
-      yy = MAC16_16(yy, x[i-T0],x[i-T0]);
    }
+   yy = yy_lookup[T0];
    best_xy = xy;
    best_yy = yy;
 #ifdef FIXED_POINT
@@ -331,6 +455,7 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod,
       int T1, T1b;
       opus_val16 g1;
       opus_val16 cont=0;
+      opus_val16 thresh;
       T1 = (2*T0+k)/(2*k);
       if (T1 < minperiod)
          break;
@@ -345,15 +470,9 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod,
       {
          T1b = (2*second_check[k]*T0+k)/(2*k);
       }
-      xy=yy=0;
+      dual_inner_prod(x, &x[-T1], &x[-T1b], N, &xy, &xy2);
-      for (i=0;i<N;i++)
+      xy += xy2;
-      {
+      yy = yy_lookup[T1] + yy_lookup[T1b];
-         xy = MAC16_16(xy, x[i], x[i-T1]);
-         yy = MAC16_16(yy, x[i-T1], x[i-T1]);
-
-         xy = MAC16_16(xy, x[i], x[i-T1b]);
-         yy = MAC16_16(yy, x[i-T1b], x[i-T1b]);
-      }
 #ifdef FIXED_POINT
       {
          opus_val32 x2y2;
@@ -372,7 +491,14 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod,
          cont = HALF32(prev_gain);
       else
          cont = 0;
-      if (g1 > 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_<minperiod0)
       *T0_=minperiod0;
+   RESTORE_STACK;
    return pg;
 }

code/opus-1.1/celt/pitch.h

@@ -0,0 +1,173 @@
+/* Copyright (c) 2007-2008 CSIRO
+   Copyright (c) 2007-2009 Xiph.Org Foundation
+   Written by Jean-Marc Valin */
+/**
+   @file pitch.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_H
+#define PITCH_H
+
+#include "modes.h"
+#include "cpu_support.h"
+
+#if defined(__SSE__) && !defined(FIXED_POINT)
+#include "x86/pitch_sse.h"
+#endif
+
+#if defined(OPUS_ARM_ASM) && defined(FIXED_POINT)
+# include "arm/pitch_arm.h"
+#endif
+
+void pitch_downsample(celt_sig * OPUS_RESTRICT x[], opus_val16 * OPUS_RESTRICT x_lp,
+      int len, int C, int arch);
+
+void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTRICT y,
+                  int len, int max_pitch, int *pitch, int arch);
+
+opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod,
+      int N, int *T0, int prev_period, opus_val16 prev_gain);
+
+/* OPT: This is the kernel you really want to optimize. It gets used a lot
+   by the prefilter and by the PLC. */
+#ifndef 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;
+   opus_val16 y_0, y_1, y_2, y_3;
+   celt_assert(len>=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<len-3;j+=4)
+   {
+      opus_val16 tmp;
+      tmp = *x++;
+      y_3=*y++;
+      sum[0] = MAC16_16(sum[0],tmp,y_0);
+      sum[1] = MAC16_16(sum[1],tmp,y_1);
+      sum[2] = MAC16_16(sum[2],tmp,y_2);
+      sum[3] = MAC16_16(sum[3],tmp,y_3);
+      tmp=*x++;
+      y_0=*y++;
+      sum[0] = MAC16_16(sum[0],tmp,y_1);
+      sum[1] = MAC16_16(sum[1],tmp,y_2);
+      sum[2] = MAC16_16(sum[2],tmp,y_3);
+      sum[3] = MAC16_16(sum[3],tmp,y_0);
+      tmp=*x++;
+      y_1=*y++;
+      sum[0] = MAC16_16(sum[0],tmp,y_2);
+      sum[1] = MAC16_16(sum[1],tmp,y_3);
+      sum[2] = MAC16_16(sum[2],tmp,y_0);
+      sum[3] = MAC16_16(sum[3],tmp,y_1);
+      tmp=*x++;
+      y_2=*y++;
+      sum[0] = MAC16_16(sum[0],tmp,y_3);
+      sum[1] = MAC16_16(sum[1],tmp,y_0);
+      sum[2] = MAC16_16(sum[2],tmp,y_1);
+      sum[3] = MAC16_16(sum[3],tmp,y_2);
+   }
+   if (j++<len)
+   {
+      opus_val16 tmp = *x++;
+      y_3=*y++;
+      sum[0] = MAC16_16(sum[0],tmp,y_0);
+      sum[1] = MAC16_16(sum[1],tmp,y_1);
+      sum[2] = MAC16_16(sum[2],tmp,y_2);
+      sum[3] = MAC16_16(sum[3],tmp,y_3);
+   }
+   if (j++<len)
+   {
+      opus_val16 tmp=*x++;
+      y_0=*y++;
+      sum[0] = MAC16_16(sum[0],tmp,y_1);
+      sum[1] = MAC16_16(sum[1],tmp,y_2);
+      sum[2] = MAC16_16(sum[2],tmp,y_3);
+      sum[3] = MAC16_16(sum[3],tmp,y_0);
+   }
+   if (j<len)
+   {
+      opus_val16 tmp=*x++;
+      y_1=*y++;
+      sum[0] = MAC16_16(sum[0],tmp,y_2);
+      sum[1] = MAC16_16(sum[1],tmp,y_3);
+      sum[2] = MAC16_16(sum[2],tmp,y_0);
+      sum[3] = MAC16_16(sum[3],tmp,y_1);
+   }
+}
+#endif /* OVERRIDE_XCORR_KERNEL */
+
+#ifndef 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;
+   opus_val32 xy01=0;
+   opus_val32 xy02=0;
+   for (i=0;i<N;i++)
+   {
+      xy01 = MAC16_16(xy01, x[i], y01[i]);
+      xy02 = MAC16_16(xy02, x[i], y02[i]);
+   }
+   *xy1 = xy01;
+   *xy2 = xy02;
+}
+#endif
+
+#ifdef FIXED_POINT
+opus_val32
+#else
+void
+#endif
+celt_pitch_xcorr_c(const opus_val16 *_x, const opus_val16 *_y,
+      opus_val32 *xcorr, int len, int max_pitch);
+
+#if !defined(OVERRIDE_PITCH_XCORR)
+/*Is run-time CPU detection enabled on this platform?*/
+# if defined(OPUS_HAVE_RTCD)
+extern
+#  if defined(FIXED_POINT)
+opus_val32
+#  else
+void
+#  endif
+(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
+      const opus_val16 *, opus_val32 *, int, int);
+
+#  define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
+  ((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \
+        xcorr, len, max_pitch))
+# else
+#  define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
+  ((void)(arch),celt_pitch_xcorr_c(_x, _y, xcorr, len, max_pitch))
+# endif
+#endif
+
+#endif

code/opus-1.1/celt/quant_bands.c

@@ -40,8 +40,8 @@
 #include "rate.h"
 
 #ifdef FIXED_POINT
-/* Mean energy in each band quantized in Q6 */
+/* Mean energy in each band quantized in Q4 */
-static const signed char eMeans[25] = {
+const signed char eMeans[25] = {
       103,100, 92, 85, 81,
        77, 72, 70, 78, 75,
        73, 71, 78, 74, 69,
@@ -49,8 +49,8 @@ static const signed char eMeans[25] = {
        60, 60, 60, 60, 60
 };
 #else
-/* Mean energy in each band quantized in Q6 and converted back to float */
+/* Mean energy in each band quantized in Q4 and converted back to float */
-static const opus_val16 eMeans[25] = {
+const opus_val16 eMeans[25] = {
       6.437500f, 6.250000f, 5.750000f, 5.312500f, 5.062500f,
       4.812500f, 4.500000f, 4.375000f, 4.875000f, 4.687500f,
       4.562500f, 4.437500f, 4.875000f, 4.625000f, 4.312500f,
@@ -157,7 +157,7 @@ static int quant_coarse_energy_impl(const CELTMode *m, int start, int end,
       const opus_val16 *eBands, opus_val16 *oldEBands,
       opus_int32 budget, opus_int32 tell,
       const unsigned char *prob_model, opus_val16 *error, ec_enc *enc,
-      int C, int LM, int intra, opus_val16 max_decay)
+      int C, int LM, int intra, opus_val16 max_decay, int lfe)
 {
    int i, c;
    int badness = 0;
@@ -222,6 +222,8 @@ static int quant_coarse_energy_impl(const CELTMode *m, int start, int end,
             if (bits_left < 16)
                qi = IMAX(-1, qi);
          }
+         if (lfe && i>=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;i<start;i++)
-         eBands[i+c*m->nbEBands] = 0;
-      for (;i<end;i++)
-      {
-         opus_val16 lg = ADD16(oldEBands[i+c*m->nbEBands],
-                         SHL16((opus_val16)eMeans[i],6));
-         eBands[i+c*m->nbEBands] = PSHR32(celt_exp2(lg),4);
-      }
-      for (;i<m->nbEBands;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)
 {

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);
 

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<prev?7:9)*band_width<<LM<<BITRES)>>4)
+            if (codedBands<=start+2 || (band_bits > ((j<prev?7:9)*band_width<<LM<<BITRES)>>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;
 }

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

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 <valgrind/memcheck.h>
+#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 */

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 X Residual signal to quantise/encode (returns quantised version)
- * @param W Perceptual weight to use when optimising (currently unused)
  * @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
  */

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 <xmmintrin.h>
+#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;i<N-3;i+=4)
+   {
+      __m128 xi = _mm_loadu_ps(x+i);
+      __m128 y1i = _mm_loadu_ps(y01+i);
+      __m128 y2i = _mm_loadu_ps(y02+i);
+      xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(xi, y1i));
+      xsum2 = _mm_add_ps(xsum2,_mm_mul_ps(xi, y2i));
+   }
+   /* Horizontal sum */
+   xsum1 = _mm_add_ps(xsum1, _mm_movehl_ps(xsum1, xsum1));
+   xsum1 = _mm_add_ss(xsum1, _mm_shuffle_ps(xsum1, xsum1, 0x55));
+   _mm_store_ss(xy1, xsum1);
+   xsum2 = _mm_add_ps(xsum2, _mm_movehl_ps(xsum2, xsum2));
+   xsum2 = _mm_add_ss(xsum2, _mm_shuffle_ps(xsum2, xsum2, 0x55));
+   _mm_store_ss(xy2, xsum2);
+   for (;i<N;i++)
+   {
+      *xy1 = MAC16_16(*xy1, x[i], y01[i]);
+      *xy2 = MAC16_16(*xy2, x[i], y02[i]);
+   }
+}
+
+#define OVERRIDE_COMB_FILTER_CONST
+static OPUS_INLINE void comb_filter_const(opus_val32 *y, opus_val32 *x, int T, int N,
+      opus_val16 g10, opus_val16 g11, opus_val16 g12)
+{
+   int i;
+   __m128 x0v;
+   __m128 g10v, g11v, g12v;
+   g10v = _mm_load1_ps(&g10);
+   g11v = _mm_load1_ps(&g11);
+   g12v = _mm_load1_ps(&g12);
+   x0v = _mm_loadu_ps(&x[-T-2]);
+   for (i=0;i<N-3;i+=4)
+   {
+      __m128 yi, yi2, x1v, x2v, x3v, x4v;
+      const opus_val32 *xp = &x[i-T-2];
+      yi = _mm_loadu_ps(x+i);
+      x4v = _mm_loadu_ps(xp+4);
+#if 0
+      /* Slower version with all loads */
+      x1v = _mm_loadu_ps(xp+1);
+      x2v = _mm_loadu_ps(xp+2);
+      x3v = _mm_loadu_ps(xp+3);
+#else
+      x2v = _mm_shuffle_ps(x0v, x4v, 0x4e);
+      x1v = _mm_shuffle_ps(x0v, x2v, 0x99);
+      x3v = _mm_shuffle_ps(x2v, x4v, 0x99);
+#endif
+
+      yi = _mm_add_ps(yi, _mm_mul_ps(g10v,x2v));
+#if 0 /* Set to 1 to make it bit-exact with the non-SSE version */
+      yi = _mm_add_ps(yi, _mm_mul_ps(g11v,_mm_add_ps(x3v,x1v)));
+      yi = _mm_add_ps(yi, _mm_mul_ps(g12v,_mm_add_ps(x4v,x0v)));
+#else
+      /* Use partial sums */
+      yi2 = _mm_add_ps(_mm_mul_ps(g11v,_mm_add_ps(x3v,x1v)),
+                       _mm_mul_ps(g12v,_mm_add_ps(x4v,x0v)));
+      yi = _mm_add_ps(yi, yi2);
+#endif
+      x0v=x4v;
+      _mm_storeu_ps(y+i, yi);
+   }
+#ifdef CUSTOM_MODES
+   for (;i<N;i++)
+   {
+      y[i] = x[i]
+               + MULT16_32_Q15(g10,x[i-T])
+               + MULT16_32_Q15(g11,ADD32(x[i-T+1],x[i-T-1]))
+               + MULT16_32_Q15(g12,ADD32(x[i-T+2],x[i-T-2]));
+   }
+#endif
+}
+
+#endif

code/opus-1.1/include/opus.h

@@ -520,7 +520,7 @@ OPUS_EXPORT void opus_decoder_destroy(OpusDecoder *st);
   * @param [in] len <tt>opus_int32</tt>: size of data
   * @param [out] out_toc <tt>char*</tt>: TOC pointer
   * @param [out] frames <tt>char*[48]</tt> encapsulated frames
-  * @param [out] size <tt>short[48]</tt> sizes of the encapsulated frames
+  * @param [out] size <tt>opus_int16[48]</tt> sizes of the encapsulated frames
   * @param [out] payload_offset <tt>int*</tt>: 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 <tt>char*</tt>: Opus packet
   * @param [in] len <tt>opus_int32</tt>: 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 <tt>char*</tt>: Opus packet
   * @param [in] len <tt>opus_int32</tt>: 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 <tt>float*</tt>: Input PCM and modified PCM
+  * @param [in] frame_size <tt>int</tt> Number of samples per channel to process
+  * @param [in] channels <tt>int</tt>: Number of channels
+  * @param [in,out] softclip_mem <tt>float*</tt>: 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 <tt>const unsigned char*</tt>: The buffer containing the
+  *                                                   packet to pad.
+  * @param len <tt>opus_int32</tt>: The size of the packet.
+  *                                 This must be at least 1.
+  * @param new_len <tt>opus_int32</tt>: 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 <tt>const unsigned char*</tt>: The buffer containing the
+  *                                                   packet to strip.
+  * @param len <tt>opus_int32</tt>: 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 <tt>const unsigned char*</tt>: The buffer containing the
+  *                                                   packet to pad.
+  * @param len <tt>opus_int32</tt>: The size of the packet.
+  *                                 This must be at least 1.
+  * @param new_len <tt>opus_int32</tt>: The desired size of the packet after padding.
+  *                                 This must be at least 1.
+  * @param nb_streams <tt>opus_int32</tt>: 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 <tt>const unsigned char*</tt>: The buffer containing the
+  *                                                   packet to strip.
+  * @param len <tt>opus_int32</tt>: The size of the packet.
+  *                                 This must be at least 1.
+  * @param nb_streams <tt>opus_int32</tt>: 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

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 OPUS_EXPORT
-#define OPUS_CUSTOM_EXPORT_STATIC OPUS_EXPORT
+# define OPUS_CUSTOM_EXPORT_STATIC OPUS_EXPORT
 #else
-#define OPUS_CUSTOM_EXPORT
+# define OPUS_CUSTOM_EXPORT
-#ifdef CELT_C
+# ifdef OPUS_BUILD
-#define OPUS_CUSTOM_EXPORT_STATIC static inline
+#  define OPUS_CUSTOM_EXPORT_STATIC static OPUS_INLINE
-#else
+# else
-#define OPUS_CUSTOM_EXPORT_STATIC
+#  define OPUS_CUSTOM_EXPORT_STATIC
-#endif
+# 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 <tt>OpusCustomMode *</tt>: 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 <tt>OpusCustomEncoder*</tt>: Encoder state
+  * @param [in] mode <tt>OpusCustomMode *</tt>: Contains all the information about the characteristics of
+  *  the stream (must be the same characteristics as used for the
+  *  decoder)
+  * @param [in] channels <tt>int</tt>: 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 <tt>OpusCustomMode*</tt>: 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 <tt>OpusCustomEncoder*</tt>: Encoder state
-  * @param [in] mode <tt>OpusCustomMode *</tt>: Contains all the information about the characteristics of
-  *  the stream (must be the same characteristics as used for the
-  *  decoder)
-  * @param [in] channels <tt>int</tt>: 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 <tt>OpusCustomEncoder*</tt>: 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 <tt>OpusCustomMode</tt>: Contains all the information about the characteristics of the
-  *          stream (must be the same characteristics as used for the encoder)
-  * @param [in] channels <tt>int</tt>: Number of channels
-  * @param [out] error <tt>int*</tt>: 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 <tt>OpusCustomMode</tt>: Contains all the information about the characteristics of the
+  *          stream (must be the same characteristics as used for the encoder)
+  * @param [in] channels <tt>int</tt>: Number of channels
+  * @param [out] error <tt>int*</tt>: 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 <tt>OpusCustomDecoder*</tt>: State to be freed.
   */

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)
+# if defined(WIN32)
-#  define OPUS_EXPORT __attribute__ ((visibility ("default")))
-# elif defined(WIN32) && !defined(__MINGW32__)
 #  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 <tt>opus_int32 *</tt>: 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 <b>really</b>
+  * know what you are doing.
+  * @see OPUS_GET_EXPERT_VARIABLE_DURATION
+  * @param[in] x <tt>opus_int32</tt>: Allowed values:
+  * <dl>
+  * <dt>OPUS_FRAMESIZE_ARG</dt><dd>Select frame size from the argument (default).</dd>
+  * <dt>OPUS_FRAMESIZE_2_5_MS</dt><dd>Use 2.5 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_5_MS</dt><dd>Use 2.5 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_10_MS</dt><dd>Use 10 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_20_MS</dt><dd>Use 20 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_40_MS</dt><dd>Use 40 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_60_MS</dt><dd>Use 60 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_VARIABLE</dt><dd>Optimize the frame size dynamically.</dd>
+  * </dl>
+  * @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 <tt>opus_int32 *</tt>: Returns one of the following values:
+  * <dl>
+  * <dt>OPUS_FRAMESIZE_ARG</dt><dd>Select frame size from the argument (default).</dd>
+  * <dt>OPUS_FRAMESIZE_2_5_MS</dt><dd>Use 2.5 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_5_MS</dt><dd>Use 2.5 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_10_MS</dt><dd>Use 10 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_20_MS</dt><dd>Use 20 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_40_MS</dt><dd>Use 40 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_60_MS</dt><dd>Use 60 ms frames.</dd>
+  * <dt>OPUS_FRAMESIZE_VARIABLE</dt><dd>Optimize the frame size dynamically.</dd>
+  * </dl>
+  * @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

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 <tt>OpusMSEncoder*</tt>: Multistream encoder state.
   * @param[in] pcm <tt>const opus_int16*</tt>: The input signal as interleaved

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,

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                                  */
 );
 

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;
 }

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

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)        */
 )

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<d;j++) {
+        mem[ j ] = in[ d - j - 1 ];
+    }
+    celt_fir( in + d, num, out + d, len - d, d, mem );
+    for ( j = 0; j < d; j++ ) {
+        out[ j ] = 0;
+    }
+#else
     for( ix = d; ix < len; ix++ ) {
         in_ptr = &in[ ix - 1 ];
 
@@ -82,4 +102,5 @@ void silk_LPC_analysis_filter(
 
     /* Set first d output samples to zero */
     silk_memset( out, 0, d * sizeof( opus_int16 ) );
+#endif
 }

code/opus-1.1/silk/LPC_inv_pred_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

code/opus-1.1/silk/LP_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
@@ -38,7 +38,7 @@ POSSIBILITY OF SUCH DAMAGE.
 #include "main.h"
 
 /* Helper function, interpolates the filter taps */
-static inline void silk_LP_interpolate_filter_taps(
+static OPUS_INLINE void silk_LP_interpolate_filter_taps(
     opus_int32           B_Q28[ TRANSITION_NB ],
     opus_int32           A_Q28[ TRANSITION_NA ],
     const opus_int       ind,

code/opus-1.1/silk/MacroCount.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
@@ -34,11 +34,11 @@ POSSIBILITY OF SUCH DAMAGE.
 
 extern opus_int64 ops_count;
 
-static inline opus_int64 silk_SaveCount(){
+static OPUS_INLINE opus_int64 silk_SaveCount(){
     return(ops_count);
 }
 
-static inline opus_int64 silk_SaveResetCount(){
+static OPUS_INLINE opus_int64 silk_SaveResetCount(){
     opus_int64 ret;
 
     ret = ops_count;
@@ -46,12 +46,12 @@ static inline opus_int64 silk_SaveResetCount(){
     return(ret);
 }
 
-static inline silk_PrintCount(){
+static OPUS_INLINE silk_PrintCount(){
     printf("ops_count = %d \n ", (opus_int32)ops_count);
 }
 
 #undef silk_MUL
-static inline opus_int32 silk_MUL(opus_int32 a32, opus_int32 b32){
+static OPUS_INLINE opus_int32 silk_MUL(opus_int32 a32, opus_int32 b32){
     opus_int32 ret;
     ops_count += 4;
     ret = a32 * b32;
@@ -59,14 +59,14 @@ static inline opus_int32 silk_MUL(opus_int32 a32, opus_int32 b32){
 }
 
 #undef silk_MUL_uint
-static inline opus_uint32 silk_MUL_uint(opus_uint32 a32, opus_uint32 b32){
+static OPUS_INLINE opus_uint32 silk_MUL_uint(opus_uint32 a32, opus_uint32 b32){
     opus_uint32 ret;
     ops_count += 4;
     ret = a32 * b32;
     return ret;
 }
 #undef silk_MLA
-static inline opus_int32 silk_MLA(opus_int32 a32, opus_int32 b32, opus_int32 c32){
+static OPUS_INLINE opus_int32 silk_MLA(opus_int32 a32, opus_int32 b32, opus_int32 c32){
     opus_int32 ret;
     ops_count += 4;
     ret = a32 + b32 * c32;
@@ -74,7 +74,7 @@ static inline opus_int32 silk_MLA(opus_int32 a32, opus_int32 b32, opus_int32 c32
 }
 
 #undef silk_MLA_uint
-static inline opus_int32 silk_MLA_uint(opus_uint32 a32, opus_uint32 b32, opus_uint32 c32){
+static OPUS_INLINE opus_int32 silk_MLA_uint(opus_uint32 a32, opus_uint32 b32, opus_uint32 c32){
     opus_uint32 ret;
     ops_count += 4;
     ret = a32 + b32 * c32;
@@ -82,14 +82,14 @@ static inline opus_int32 silk_MLA_uint(opus_uint32 a32, opus_uint32 b32, opus_ui
 }
 
 #undef silk_SMULWB
-static inline opus_int32 silk_SMULWB(opus_int32 a32, opus_int32 b32){
+static OPUS_INLINE opus_int32 silk_SMULWB(opus_int32 a32, opus_int32 b32){
     opus_int32 ret;
     ops_count += 5;
     ret = (a32 >> 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;

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 )

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)   */

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

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

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 ]          */

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;

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 ];
+    VARDECL( opus_int32, err_Q26 );
-    opus_int32       RD_Q25[       NLSF_VQ_MAX_SURVIVORS ];
+    VARDECL( opus_int32, RD_Q25 );
-    opus_int         tempIndices1[ NLSF_VQ_MAX_SURVIVORS ];
+    VARDECL( opus_int, tempIndices1 );
-    opus_int8        tempIndices2[ NLSF_VQ_MAX_SURVIVORS * MAX_LPC_ORDER ];
+    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 ];
 }