Use AudioCompress for volume normalization

git-svn-id: https://svn.musicpd.org/mpd/trunk@4474 09075e82-0dd4-0310-85a5-a0d7c8717e4f

10 files changed, 462 insertions, 99 deletions

diff --git a/AUTHORS b/AUTHORS
index 734779e4..6efbe8ce 100644
--- a/AUTHORS
+++ b/AUTHORS
@@ -7,3 +7,4 @@ command.c and signal handling cleanup -> mackstann
 replayGain -> AliasMrJones
 
 mp4ff copyrighted by M. Bakker, Ahead Software AG, http://www.nero.com
+compress.[ch] copyrighted by M. Hari Nezumi <magenta@trikuare.cx>
diff --git a/src/Makefile.am b/src/Makefile.am
index 08f7ed01..e86c73ea 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -50,6 +50,7 @@ mpd_headers = \
 	mpd_types.h \
 	myfprintf.h \
 	normalize.h \
+	compress.h \
 	outputBuffer.h \
 	path.h \
 	pcm_utils.h \
@@ -95,6 +96,7 @@ mpd_SOURCES = \
 	metadataChunk.c \
 	myfprintf.c \
 	normalize.c \
+	compress.c \
 	outputBuffer.c \
 	path.c \
 	pcm_utils.c \
diff --git a/src/compress.c b/src/compress.c
new file mode 100644
index 00000000..e6c6cd6f
--- /dev/null
+++ b/src/compress.c
@@ -0,0 +1,392 @@
+/* compress.c
+** Compressor logic
+*/
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+
+#include "compress.h"
+
+#ifdef USE_X
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+
+static Display *display;
+static Window window;
+static Visual *visual;
+static int screen;
+static GC blackGC, whiteGC, blueGC, yellowGC, dkyellowGC, redGC;
+#endif
+
+static int *peaks = NULL;
+static int gainCurrent, gainTarget;
+
+static struct {
+	int show_mon;
+	int anticlip;
+	int target;
+	int gainmax;
+	int gainsmooth;
+	int buckets;
+} prefs;
+
+#ifdef USE_X
+static int mon_init = 0;
+#endif
+
+void CompressCfg(int show_mon, int anticlip, int target, int gainmax,
+		 int gainsmooth, int buckets)
+{
+	static int lastsize = 0;
+
+	prefs.show_mon = show_mon;
+	prefs.anticlip = anticlip;
+	prefs.target = target;
+	prefs.gainmax = gainmax;
+	prefs.gainsmooth = gainsmooth;
+	prefs.buckets = buckets;
+
+	/* Allocate the peak structure */
+	peaks = realloc(peaks, sizeof(int)*prefs.buckets);
+
+	if (prefs.buckets > lastsize)
+		memset(peaks + lastsize, 0, sizeof(int)*(prefs.buckets
+							 - lastsize));
+	lastsize = prefs.buckets;
+
+#ifdef USE_X
+	/* Configure the monitor window if needed */
+	if (show_mon && !mon_init)
+	{
+		display = XOpenDisplay(getenv("DISPLAY"));
+
+		/* We really shouldn't try to init X if there's no X */
+		if (!display)
+		{
+			fprintf(stderr,
+				"X not detected; disabling monitor window\n");
+			show_mon = prefs.show_mon = 0;
+		}
+	}
+
+	if (show_mon && !mon_init)
+	{
+		XGCValues gcv;
+		XColor col;
+	
+		gainCurrent = gainTarget = (1 << GAINSHIFT);
+
+
+
+		screen = DefaultScreen(display);
+		visual = DefaultVisual(display, screen);
+		window = XCreateSimpleWindow(display,
+					     RootWindow(display, screen),
+					     0, 0, prefs.buckets, 128 + 8, 0,
+					     BlackPixel(display, screen),
+					     WhitePixel(display, screen));
+		XStoreName(display, window, "AudioCompress monitor");
+	
+		gcv.foreground = BlackPixel(display, screen);
+		blackGC = XCreateGC(display, window, GCForeground, &gcv);
+		gcv.foreground = WhitePixel(display, screen);
+		whiteGC = XCreateGC(display, window, GCForeground, &gcv);
+		col.red = 0;
+		col.green = 0;
+		col.blue = 65535;
+		XAllocColor(display, DefaultColormap(display, screen), &col);
+		gcv.foreground = col.pixel;
+		blueGC = XCreateGC(display, window, GCForeground, &gcv);
+		col.red = 65535;
+		col.green = 65535;
+		col.blue = 0;
+		XAllocColor(display, DefaultColormap(display, screen), &col);
+		gcv.foreground = col.pixel;
+		yellowGC = XCreateGC(display, window, GCForeground, &gcv);
+		col.red = 32767;
+		col.green = 32767;
+		col.blue = 0;
+		XAllocColor(display, DefaultColormap(display, screen), &col);
+		gcv.foreground = col.pixel;
+		dkyellowGC = XCreateGC(display, window, GCForeground, &gcv);
+		col.red = 65535;
+		col.green = 0;
+		col.blue = 0;
+		XAllocColor(display, DefaultColormap(display, screen), &col);
+		gcv.foreground = col.pixel;
+		redGC = XCreateGC(display, window, GCForeground, &gcv);
+		mon_init = 1;
+	}
+
+	if (mon_init)
+	{
+		if (show_mon)
+			XMapWindow(display, window);
+		else
+			XUnmapWindow(display, window);
+		XResizeWindow(display, window, prefs.buckets, 128 + 8);
+		XFlush(display);
+	}
+#endif
+}
+
+void CompressFree(void)
+{
+#ifdef USE_X
+	if (mon_init)
+	{
+		XFreeGC(display, blackGC);
+		XFreeGC(display, whiteGC);
+		XFreeGC(display, blueGC);
+		XFreeGC(display, yellowGC);
+		XFreeGC(display, dkyellowGC);
+		XFreeGC(display, redGC);
+		XDestroyWindow(display, window);
+		XCloseDisplay(display);
+	}
+#endif
+
+	if (peaks)
+		free(peaks);
+}
+
+void CompressDo(void *data, unsigned int length)
+{
+	int16_t *audio = (int16_t *)data, *ap;
+	int peak, pos;
+	int i;
+	int gr, gf, gn;
+	static int pn = -1;
+#ifdef STATS
+	static int clip = 0;
+#endif
+	static int clipped = 0;
+
+	if (!peaks)
+		return;
+
+	if (pn == -1)
+	{
+		for (i = 0; i < prefs.buckets; i++)
+			peaks[i] = 0;
+	}
+	pn = (pn + 1)%prefs.buckets;
+
+#ifdef DEBUG
+	fprintf(stderr, "modifyNative16(0x%08x, %d)\n",(unsigned int)data,
+		length);
+#endif
+
+	/* Determine peak's value and position */
+	peak = 1;
+	pos = 0;
+
+#ifdef DEBUG
+	fprintf(stderr, "finding peak(b=%d)\n", pn);
+#endif
+
+	ap = audio;
+	for (i = 0; i < length/2; i++)
+	{
+		int val = *ap;
+		if (val > peak)
+		{
+			peak = val;
+			pos = i;
+		} else if (-val > peak)
+		{
+			peak = -val;
+			pos = i;
+		}
+		ap++;
+	}
+	peaks[pn] = peak;
+
+	/* Only draw if needed, of course */
+#ifdef USE_X
+	if (prefs.show_mon)
+	{
+		/* current amplitude */
+		XDrawLine(display, window, whiteGC,
+			  pn, 0,
+			  pn,
+			  127 -
+			  (peaks[pn]*gainCurrent >> (GAINSHIFT + 8)));
+
+		/* amplification */
+		XDrawLine(display, window, yellowGC,
+			  pn,
+			  127 - (peaks[pn]*gainCurrent
+				 >> (GAINSHIFT + 8)),
+			  pn, 127);
+
+		/* peak */
+		XDrawLine(display, window, blackGC,
+			  pn, 127 - (peaks[pn] >> 8), pn, 127);
+
+		/* clip indicator */
+		if (clipped)
+			XDrawLine(display, window, redGC,
+				  (pn + prefs.buckets - 1)%prefs.buckets,
+				  126 - clipped/(length*512),
+				  (pn + prefs.buckets - 1)%prefs.buckets,
+				  127);
+		clipped = 0;
+
+		// target line
+		//XDrawPoint(display, window, redGC,
+		//         pn, 127 - TARGET/256);
+		// amplification edge
+		XDrawLine(display, window, dkyellowGC,
+			  pn,
+			  127 - (peaks[pn]*gainCurrent
+				 >> (GAINSHIFT + 8)),
+			  pn - 1,
+			  127 -
+			  (peaks[(pn + prefs.buckets
+				  - 1)%prefs.buckets]*gainCurrent
+			   >> (GAINSHIFT + 8)));
+	}
+#endif
+
+	for (i = 0; i < prefs.buckets; i++)
+	{
+		if (peaks[i] > peak)
+		{
+			peak = peaks[i];
+			pos = 0;
+		}
+	}
+
+	/* Determine target gain */
+	gn = (1 << GAINSHIFT)*prefs.target/peak;
+
+	if (gn <(1 << GAINSHIFT))
+		gn = 1 << GAINSHIFT;
+
+	gainTarget = (gainTarget *((1 << prefs.gainsmooth) - 1) + gn)
+				      >> prefs.gainsmooth;
+
+	/* Give it an extra insignifigant nudge to counteract possible
+	** rounding error
+	*/
+
+	if (gn < gainTarget)
+		gainTarget--;
+	else if (gn > gainTarget)
+		gainTarget++;
+
+	if (gainTarget > prefs.gainmax << GAINSHIFT)
+		gainTarget = prefs.gainmax << GAINSHIFT;
+
+
+#ifdef USE_X
+	if (prefs.show_mon)
+	{
+		int x;
+
+		/* peak*gain */
+		XDrawPoint(display, window, redGC,
+			   pn,
+			   127 - (peak*gainCurrent
+				  >> (GAINSHIFT + 8)));
+
+		// gain indicator
+		XFillRectangle(display, window, whiteGC, 0, 128,
+			       prefs.buckets, 8);
+		x = (gainTarget - (1 << GAINSHIFT))*prefs.buckets
+			/ ((prefs.gainmax - 1) << GAINSHIFT);
+		XDrawLine(display, window, redGC, x,
+			  128, x, 128 + 8);
+
+		x = (gn - (1 << GAINSHIFT))*prefs.buckets
+			/ ((prefs.gainmax - 1) << GAINSHIFT);
+
+		XDrawLine(display, window, blackGC,
+			  x, 132 - 1,
+			  x, 132 + 1);
+
+		// blue peak line
+		XDrawLine(display, window, blueGC,
+			  0, 127 - (peak >> 8), prefs.buckets,
+			  127 - (peak >> 8));
+		XFlush(display);
+		XDrawLine(display, window, whiteGC,
+			  0, 127 - (peak >> 8), prefs.buckets,
+			  127 - (peak >> 8));
+	}
+#endif
+
+	/* See if a peak is going to clip */
+	gn = (1 << GAINSHIFT)*32768/peak;
+
+	if (gn < gainTarget)
+	{
+		gainTarget = gn;
+
+		if (prefs.anticlip)
+			pos = 0;
+
+	} else
+	{
+		/* We're ramping up, so draw it out over the whole frame */
+		pos = length;
+	}
+
+	/* Determine gain rate necessary to make target */
+	if (!pos)
+		pos = 1;
+
+	gr = ((gainTarget - gainCurrent) << 16)/pos;
+
+	/* Do the shiznit */
+	gf = gainCurrent << 16;
+
+#ifdef STATS
+	fprintf(stderr, "\rgain = %2.2f%+.2e ",
+		gainCurrent*1.0/(1 << GAINSHIFT),
+		(gainTarget - gainCurrent)*1.0/(1 << GAINSHIFT));
+#endif
+
+	ap = audio;
+	for (i = 0; i < length/2; i++)
+	{
+		int sample;
+
+		/* Interpolate the gain */
+		gainCurrent = gf >> 16;
+		if (i < pos)
+			gf += gr;
+		else if (i == pos)
+			gf = gainTarget << 16;
+
+		/* Amplify */
+		sample = (*ap)*gainCurrent >> GAINSHIFT;
+		if (sample < -32768)
+		{
+#ifdef STATS
+			clip++;
+#endif
+			clipped += -32768 - sample;
+			sample = -32768;
+		} else if (sample > 32767)
+		{
+#ifdef STATS
+			clip++;
+#endif
+			clipped += sample - 32767;
+			sample = 32767;
+		}
+		*ap++ = sample;
+	}
+#ifdef STATS
+	fprintf(stderr, "clip %d b%-3d ", clip, pn);
+#endif
+
+#ifdef DEBUG
+	fprintf(stderr, "\ndone\n");
+#endif
+}
+
diff --git a/src/compress.h b/src/compress.h
new file mode 100644
index 00000000..1993b47a
--- /dev/null
+++ b/src/compress.h
@@ -0,0 +1,29 @@
+/* compress.h
+** interface to audio compression
+*/
+
+#ifndef COMPRESS_H
+#define COMPRESS_H
+
+/* These are copied from the AudioCompress config.h, mainly because CompressDo
+ * needs GAINSHIFT defined.  The rest are here so they can be used as defaults
+ * to pass to CompressCfg. */
+#define ANTICLIP 0		/* Strict clipping protection */
+#define TARGET 25000		/* Target level */
+#define GAINMAX 32		/* The maximum amount to amplify by */
+#define GAINSHIFT 10		/* How fine-grained the gain is */
+#define GAINSMOOTH 8		/* How much inertia ramping has*/
+#define BUCKETS 400		/* How long of a history to store */
+
+void CompressCfg(int monitor,
+		 int anticlip,
+		 int target,
+		 int maxgain,
+		 int smooth,
+		 int buckets);
+
+void CompressDo(void *data, unsigned int numSamples);
+
+void CompressFree(void);
+
+#endif
diff --git a/src/conf.c b/src/conf.c
index 995cbf89..6fbc4734 100644
--- a/src/conf.c
+++ b/src/conf.c
@@ -373,6 +373,8 @@ int getBoolConfigParam(char *name)
 	if (strcmp("yes", param->value) == 0) return 1;
 	else if (strcmp("no", param->value) == 0) return 0;
 
+	ERROR("%s is not \"yes\" or \"no\" on line %i\n", name, param->line);
+
 	return -2;
 }
 
diff --git a/src/main.c b/src/main.c
index 7dcc5240..40e0274e 100644
--- a/src/main.c
+++ b/src/main.c
@@ -40,6 +40,7 @@
 #include "dbUtils.h"
 #include "../config.h"
 #include "utils.h"
+#include "normalize.h"
 
 #include <stdio.h>
 #include <sys/select.h>
@@ -552,6 +553,7 @@ int main(int argc, char *argv[])
 	initPaths();
 	initAudioConfig();
 	initAudioDriver();
+	initNormalization();
 	initPlaylist();
 	openDB(&options, argv[0]);
 
@@ -601,6 +603,7 @@ int main(int argc, char *argv[])
 
 	finishPlaylist();
 	freePlayerData();
+	finishNormalization();
 	finishAudioDriver();
 	finishAudioConfig();
 	finishVolume();
diff --git a/src/normalize.c b/src/normalize.c
index 8c48f13b..31394cbf 100644
--- a/src/normalize.c
+++ b/src/normalize.c
@@ -16,85 +16,32 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
-#include <math.h>
-#include <limits.h>
-
+#include "compress.h"
 #include "conf.h"
 #include "normalize.h"
-#include "playlist.h"
-
-/* silence level, apparently this is Wrong (tm) */
-#define SILENCE_LEVEL (SHRT_MAX * 0.01)
-/* not sure what this is :) */
-#define MID           (SHRT_MAX * 0.25)
 
-#define MUL_MIN 0.1
-#define MUL_MAX 5.0
-#define NSAMPLES 128
-#define MIN_SAMPLE_SIZE 32000
+#include <stdlib.h>
 
-#define clamp(a,min,max) (((a)>(max))?(max):(((a)<(min))?(min):(a)))
+int normalizationEnabled;
 
-void normalizeData(char *buffer, int bufferSize, AudioFormat *format)
+void initNormalization()
 {
-	static float multiplier = 1.0;
-	static int current_id = 0;
-	float average = 0.0;
-	static int old_song = 0;
-	int new_song = 0;
-	int total_length = 0;
-	int temp = 0;
-	int i = 0;
-	float root_mean_square = 0.0; /* the rms of the data */
-	mpd_sint16 *data = (mpd_sint16 *) buffer; /* the audio data */
-	int length = bufferSize / 2; /* the number of samples */
-	static struct {
-		float avg; /* average sample 'level' */
-		int len;   /* sample size (used to weigh sample) */
-	} mem[NSAMPLES];
-
-	/* operate only on 16 bit, 2 channel audio */
-	if (format->bits != 16 && format->channels != 2) return;
-
-	/* calculate the root mean square of the data */
-	for (i = 0; i < length; i++)
-		root_mean_square += (float)(data[i] * data[i]);
-
-	root_mean_square = sqrt(root_mean_square / (float)length);
+	normalizationEnabled = getBoolConfigParam(CONF_VOLUME_NORMALIZATION);
+	if (normalizationEnabled == -1) normalizationEnabled = 0;
+	else if (normalizationEnabled < 0) exit(EXIT_FAILURE);
 
-	/* reset the multiplier if the song has changed */
-	if (old_song != (new_song = getPlaylistCurrentSong())) {
-		old_song = new_song;
-		/* re-zero 'mem' */
-		for (i = 0; i < NSAMPLES; i++) {
-			mem[i].avg = 0.0;
-			mem[i].len = 0;
-		}
-		current_id = 0;
-	}
-
-	/* and now do magic tricks */
-	for (i = 0; i < NSAMPLES; i++) {
-		average += mem[i].avg * (float)mem[i].len;
-		total_length += mem[i].len;
-	}
+	if (normalizationEnabled)
+		CompressCfg(0, ANTICLIP, TARGET, GAINMAX, GAINSMOOTH, BUCKETS);
+}
 
-	if (total_length > MIN_SAMPLE_SIZE) {
-		average /= (float) total_length;
-		if (average >= SILENCE_LEVEL) {
-			multiplier = MID / average;
-			/* clamp multiplier */
-			multiplier = clamp(multiplier, MUL_MIN, MUL_MAX);
-		}
-	}
+void finishNormalization()
+{
+	if (normalizationEnabled) CompressFree();
+}
 
-	/* scale and clamp the samples */
-	for (i = 0; i < length; i++) {
-		temp = data[i] * multiplier;
-		data[i] = clamp(temp, SHRT_MIN, SHRT_MAX);
-	}
+void normalizeData(char *buffer, int bufferSize, AudioFormat *format)
+{
+	if ((format->bits != 16) || (format->channels != 2)) return;
 
-	mem[current_id].len = bufferSize / 2;
-	mem[current_id].avg = multiplier * root_mean_square;
-	current_id = (current_id + 1) % NSAMPLES; /* increment current_id */
+	CompressDo(buffer, bufferSize);
 }
diff --git a/src/normalize.h b/src/normalize.h
index f131669a..73e44288 100644
--- a/src/normalize.h
+++ b/src/normalize.h
@@ -16,6 +16,17 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#ifndef NORMALIZE_H
+#define NORMALIZE_H
+
 #include "audio.h"
 
+extern int normalizationEnabled;
+
+void initNormalization();
+
+void finishNormalization();
+
 void normalizeData(char *buffer, int bufferSize, AudioFormat *format);
+
+#endif /* !NORMALIZE_H */
diff --git a/src/outputBuffer.c b/src/outputBuffer.c
index 90f2ef2d..fe868d54 100644
--- a/src/outputBuffer.c
+++ b/src/outputBuffer.c
@@ -78,22 +78,6 @@ int sendDataToOutputBuffer(OutputBuffer * cb, InputStream * inStream,
 	size_t datalen;
 	static char *convBuffer = NULL;
 	static long convBufferLen = 0;
-	static int normalEnable = -1;
-	ConfigParam *param;
-
-	if (normalEnable == -1) {
-		normalEnable = getBoolConfigParam(CONF_VOLUME_NORMALIZATION);
-		if (normalEnable == -1) {
-			/* not set */
-			normalEnable = 0;
-		} else if (normalEnable < 0) {
-			param = getConfigParam(CONF_VOLUME_NORMALIZATION);
-			WARNING("%s is not \"yes\" or \"no\" on line %i, "
-			        "disabling\n", CONF_VOLUME_NORMALIZATION,
-			        param->line);
-			normalEnable = 0;
-		}
-	}
 
 	if (cmpAudioFormat(&(cb->audioFormat), &(dc->audioFormat)) == 0) {
 		data = dataIn;
@@ -115,11 +99,10 @@ int sendDataToOutputBuffer(OutputBuffer * cb, InputStream * inStream,
 				       &(cb->audioFormat), data);
 	}
 
-	if (replayGainInfo && (replayGainState != REPLAYGAIN_OFF)) {
+	if (replayGainInfo && (replayGainState != REPLAYGAIN_OFF))
 		doReplayGain(replayGainInfo, data, datalen, &cb->audioFormat);
-	} else if (normalEnable) {
+	else if (normalizationEnabled)
 		normalizeData(data, datalen, &cb->audioFormat);
-	}
 
 	while (datalen) {
 		if (currentChunk != cb->end) {
diff --git a/src/playlist.c b/src/playlist.c
index 37c2aeb5..2f4feee7 100644
--- a/src/playlist.c
+++ b/src/playlist.c
@@ -167,15 +167,8 @@ void initPlaylist(void)
 	}
 
 	playlist_saveAbsolutePaths = getBoolConfigParam(CONF_SAVE_ABSOLUTE_PATHS);
-	if (playlist_saveAbsolutePaths == -1) {
-		/* not set */
-		playlist_saveAbsolutePaths = 0;
-	} else if (playlist_saveAbsolutePaths < 0) {
-		param = getConfigParam(CONF_SAVE_ABSOLUTE_PATHS);
-		ERROR("%s is not \"yes\" or \"no\" on line %i\n",
-		      CONF_SAVE_ABSOLUTE_PATHS, param->line);
-		exit(EXIT_FAILURE);
-	}
+	if (playlist_saveAbsolutePaths == -1) playlist_saveAbsolutePaths = 0;
+	else if (playlist_saveAbsolutePaths < 0) exit(EXIT_FAILURE);
 
 	playlist.songs = malloc(sizeof(Song *) * playlist_max_length);
 	playlist.songMod = malloc(sizeof(mpd_uint32) * playlist_max_length);