audio/filter: prepare filter chain for non-interleaved audio

Based on earlier work by Stefano Pigozzi.

There are 2 changes:

1. Instead of mp_audio.audio, mp_audio.planes[0] must be used.

2. mp_audio.len used to contain the size of the audio in bytes. Now
   mp_audio.samples must be used. (Where 1 sample is the smallest unit
   of audio that covers all channels.)

Also, some filters need changes to reject non-interleaved formats
properly.

Nothing uses the non-interleaved features yet, but this is needed so
that things don't just break when doing so.

25 files changed, 163 insertions, 155 deletions

diff --git a/audio/filter/af.c b/audio/filter/af.c
index edee4bef65..95d0e43673 100644
--- a/audio/filter/af.c
+++ b/audio/filter/af.c
@@ -523,8 +523,7 @@ static int af_reinit(struct af_stream *s)
         // Check if this is the first filter
         struct mp_audio in = *af->prev->data;
         // Reset just in case...
-        in.audio = NULL;
-        in.len = 0;
+        mp_audio_set_null_data(&in);
 
         int rv = af->control(af, AF_CONTROL_REINIT, &in);
         if (rv == AF_OK && !mp_audio_config_equals(&in, af->prev->data))
@@ -640,8 +639,8 @@ int af_init(struct af_stream *s)
         return -1;
 
     // Precaution in case caller is misbehaving
-    s->input.audio  = s->output.audio  = NULL;
-    s->input.len    = s->output.len    = 0;
+    mp_audio_set_null_data(&s->input);
+    mp_audio_set_null_data(&s->output);
 
     // Check if this is the first call
     if (s->first->next == s->last) {
@@ -731,36 +730,39 @@ double af_calc_delay(struct af_stream *s)
     return delay;
 }
 
-/* Calculate the minimum output buffer size for given input data d
- * when using the af_resize_local_buffer function. The +t+1 part ensures the
- * value is >= len*mul rounded upwards to whole samples even if the
- * double 'mul' is inexact. */
-static int af_lencalc(double mul, struct mp_audio *d)
-{
-    int t = d->bps * d->nch;
-    return d->len * mul + t + 1;
-}
-
 /* I a local buffer is used (i.e. if the filter doesn't operate on the incoming
  * buffer), this macro must be called to ensure the buffer is big enough. */
 int af_resize_local_buffer(struct af_instance *af, struct mp_audio *data)
 {
-    if (af->data->len >= af_lencalc(af->mul, data))
+    assert(data->format);
+
+    if (!af->data->format && !af->data->planes[0]) {
+        // Dummy initialization
+        mp_audio_set_format(af->data, AF_FORMAT_U8);
+    }
+
+    int oldlen = af->data->samples * af->data->sstride;
+
+    /* Calculate the minimum output buffer size for given input data d
+     * when using the af_resize_local_buffer function. The +x part ensures
+     * the value is >= len*mul rounded upwards to whole samples even if the
+     * double 'mul' is inexact. */
+    int newlen = data->samples * data->sstride * af->mul + data->sstride + 1;
+
+    if (oldlen >= newlen)
         return AF_OK;
 
-    // Calculate new length
-    register int len = af_lencalc(af->mul, data);
     mp_msg(MSGT_AFILTER, MSGL_V, "[libaf] Reallocating memory in module %s, "
-           "old len = %i, new len = %i\n", af->info->name, af->data->len, len);
+           "old len = %i, new len = %i\n", af->info->name, oldlen, newlen);
     // If there is a buffer free it
-    free(af->data->audio);
+    free(af->data->planes[0]);
     // Create new buffer and check that it is OK
-    af->data->audio = malloc(len);
-    if (!af->data->audio) {
+    af->data->planes[0] = malloc(newlen);
+    if (!af->data->planes[0]) {
         mp_msg(MSGT_AFILTER, MSGL_FATAL, "[libaf] Could not allocate memory \n");
         return AF_ERROR;
     }
-    af->data->len = len;
+    af->data->samples = newlen / af->data->sstride;
     return AF_OK;
 }
 
diff --git a/audio/filter/af_bs2b.c b/audio/filter/af_bs2b.c
index 0e77b3e4eb..5e0caf28af 100644
--- a/audio/filter/af_bs2b.c
+++ b/audio/filter/af_bs2b.c
@@ -42,7 +42,7 @@ static struct mp_audio *play_##name(struct af_instance *af, struct mp_audio *dat
 { \
     /* filter is called for all pairs of samples available in the buffer */ \
     bs2b_cross_feed_##name(((struct af_bs2b*)(af->priv))->filter, \
-        (type*)(data->audio), data->len/data->bps/2); \
+        (type*)(data->planes[0]), data->samples); \
 \
     return data; \
 }
diff --git a/audio/filter/af_center.c b/audio/filter/af_center.c
index 0cfdbc3b0e..c64c551f1c 100644
--- a/audio/filter/af_center.c
+++ b/audio/filter/af_center.c
@@ -87,9 +87,9 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* data)
 {
   struct mp_audio*    c   = data;	 // Current working data
   af_center_t*  s   = af->setup; // Setup for this instance
-  float*   	a   = c->audio;	 // Audio data
-  int		len = c->len/4;	 // Number of samples in current audio block
+  float*   	a   = c->planes[0];	 // Audio data
   int		nch = c->nch;	 // Number of channels
+  int		len = c->samples*c->nch;	 // Number of samples in current audio block
   int		ch  = s->ch;	 // Channel in which to insert the center audio
   register int  i;
 
diff --git a/audio/filter/af_channels.c b/audio/filter/af_channels.c
index 27445aafe2..fd3b8262f5 100644
--- a/audio/filter/af_channels.c
+++ b/audio/filter/af_channels.c
@@ -169,6 +169,10 @@ static int control(struct af_instance* af, int cmd, void* arg)
     af->data->rate   = ((struct mp_audio*)arg)->rate;
     mp_audio_set_format(af->data, ((struct mp_audio*)arg)->format);
     af->mul          = (double)af->data->nch / ((struct mp_audio*)arg)->nch;
+    mp_audio_force_interleaved_format(af->data);
+    int r = af_test_output(af,(struct mp_audio*)arg);
+    if (r != AF_OK)
+        return r;
     return check_routes(s,((struct mp_audio*)arg)->nch,af->data->nch);
   case AF_CONTROL_COMMAND_LINE:{
     int nch = 0;
@@ -219,7 +223,7 @@ static void uninit(struct af_instance* af)
 {
   free(af->setup);
   if (af->data)
-      free(af->data->audio);
+      free(af->data->planes[0]);
   free(af->data);
 }
 
@@ -235,16 +239,16 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* data)
     return NULL;
 
   // Reset unused channels
-  memset(l->audio,0,c->len / c->nch * l->nch);
+  memset(l->planes[0],0,mp_audio_psize(c) / c->nch * l->nch);
 
   if(AF_OK == check_routes(s,c->nch,l->nch))
     for(i=0;i<s->nr;i++)
-      copy(c->audio,l->audio,c->nch,s->route[i][FR],
-	   l->nch,s->route[i][TO],c->len,c->bps);
+      copy(c->planes[0],l->planes[0],c->nch,s->route[i][FR],
+	   l->nch,s->route[i][TO],mp_audio_psize(c),c->bps);
 
   // Set output data
-  c->audio = l->audio;
-  c->len   = c->len / c->nch * l->nch;
+  c->planes[0] = l->planes[0];
+  c->samples   = c->samples / c->nch * l->nch;
   mp_audio_set_channels(c, &l->channels);
 
   return c;
diff --git a/audio/filter/af_convert24.c b/audio/filter/af_convert24.c
index 18ce156467..96924de344 100644
--- a/audio/filter/af_convert24.c
+++ b/audio/filter/af_convert24.c
@@ -78,12 +78,12 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
         return NULL;
 
     struct mp_audio *out = af->data;
-    size_t len = data->len / data->bps;
+    size_t len = mp_audio_psize(data) / data->bps;
 
     if (data->bps == 4) {
         for (int s = 0; s < len; s++) {
-            uint32_t val = *((uint32_t *)data->audio + s);
-            uint8_t *ptr = (uint8_t *)out->audio + s * 3;
+            uint32_t val = *((uint32_t *)data->planes[0] + s);
+            uint8_t *ptr = (uint8_t *)out->planes[0] + s * 3;
             ptr[0] = val >> SHIFT(0);
             ptr[1] = val >> SHIFT(1);
             ptr[2] = val >> SHIFT(2);
@@ -91,24 +91,23 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
         mp_audio_set_format(data, af_fmt_change_bits(data->format, 24));
     } else {
         for (int s = 0; s < len; s++) {
-            uint8_t *ptr = (uint8_t *)data->audio + s * 3;
+            uint8_t *ptr = (uint8_t *)data->planes[0] + s * 3;
             uint32_t val = ptr[0] << SHIFT(0)
                          | ptr[1] << SHIFT(1)
                          | ptr[2] << SHIFT(2);
-            *((uint32_t *)out->audio + s) = val;
+            *((uint32_t *)out->planes[0] + s) = val;
         }
         mp_audio_set_format(data, af_fmt_change_bits(data->format, 32));
     }
 
-    data->audio = out->audio;
-    data->len = len * data->bps;
+    data->planes[0] = out->planes[0];
     return data;
 }
 
 static void uninit(struct af_instance* af)
 {
     if (af->data)
-        free(af->data->audio);
+        free(af->data->planes[0]);
 }
 
 static int af_open(struct af_instance *af)
diff --git a/audio/filter/af_convertsignendian.c b/audio/filter/af_convertsignendian.c
index bfea004bb2..5565438aad 100644
--- a/audio/filter/af_convertsignendian.c
+++ b/audio/filter/af_convertsignendian.c
@@ -24,6 +24,9 @@
 
 static bool test_conversion(int src_format, int dst_format)
 {
+    if ((src_format & AF_FORMAT_PLANAR) ||
+        (dst_format & AF_FORMAT_PLANAR))
+        return false;
     int src_noend = src_format & ~AF_FORMAT_END_MASK;
     int dst_noend = dst_format & ~AF_FORMAT_END_MASK;
     // We can swap endian for all formats, but sign only for integer formats.
@@ -100,13 +103,13 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
 {
     int infmt = data->format;
     int outfmt = af->data->format;
-    size_t len = data->len / data->bps;
+    size_t len = data->samples * data->nch;
 
     if ((infmt & AF_FORMAT_END_MASK) != (outfmt & AF_FORMAT_END_MASK))
-        endian(data->audio, len, data->bps);
+        endian(data->planes[0], len, data->bps);
 
     if ((infmt & AF_FORMAT_SIGN_MASK) != (outfmt & AF_FORMAT_SIGN_MASK))
-        si2us(data->audio, len, data->bps,
+        si2us(data->planes[0], len, data->bps,
               (outfmt & AF_FORMAT_END_MASK) == AF_FORMAT_LE);
 
     mp_audio_set_format(data, outfmt);
diff --git a/audio/filter/af_delay.c b/audio/filter/af_delay.c
index a6515f84cf..c979060fe3 100644
--- a/audio/filter/af_delay.c
+++ b/audio/filter/af_delay.c
@@ -56,6 +56,7 @@ static int control(struct af_instance* af, int cmd, void* arg)
       free(s->q[i]);
 
     mp_audio_copy_config(af->data, (struct mp_audio*)arg);
+    mp_audio_force_interleaved_format(af->data);
 
     // Allocate new delay queues
     for(i=0;i<af->data->nch;i++){
@@ -123,13 +124,13 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* data)
   struct mp_audio*   	c   = data;	 // Current working data
   af_delay_t*  	s   = af->setup; // Setup for this instance
   int 		nch = c->nch;	 // Number of channels
-  int		len = c->len/c->bps; // Number of sample in data chunk
+  int		len = mp_audio_psize(c)/c->bps; // Number of sample in data chunk
   int		ri  = 0;
   int 		ch,i;
   for(ch=0;ch<nch;ch++){
     switch(c->bps){
     case 1:{
-      int8_t* a = c->audio;
+      int8_t* a = c->planes[0];
       int8_t* q = s->q[ch];
       int wi = s->wi[ch];
       ri = s->ri;
@@ -143,7 +144,7 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* data)
       break;
     }
     case 2:{
-      int16_t* a = c->audio;
+      int16_t* a = c->planes[0];
       int16_t* q = s->q[ch];
       int wi = s->wi[ch];
       ri = s->ri;
@@ -157,7 +158,7 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* data)
       break;
     }
     case 4:{
-      int32_t* a = c->audio;
+      int32_t* a = c->planes[0];
       int32_t* q = s->q[ch];
       int wi = s->wi[ch];
       ri = s->ri;
diff --git a/audio/filter/af_drc.c b/audio/filter/af_drc.c
index 9bbbde4831..589844d89b 100644
--- a/audio/filter/af_drc.c
+++ b/audio/filter/af_drc.c
@@ -88,6 +88,7 @@ static int control(struct af_instance* af, int cmd, void* arg)
     // Sanity check
     if(!arg) return AF_ERROR;
 
+    mp_audio_force_interleaved_format((struct mp_audio*)arg);
     mp_audio_copy_config(af->data, (struct mp_audio*)arg);
 
     if(((struct mp_audio*)arg)->format != (AF_FORMAT_S16_NE)){
@@ -119,8 +120,8 @@ static void uninit(struct af_instance* af)
 static void method1_int16(af_drc_t *s, struct mp_audio *c)
 {
   register int i = 0;
-  int16_t *data = (int16_t*)c->audio;	// Audio data
-  int len = c->len/2;		// Number of samples
+  int16_t *data = (int16_t*)c->planes[0];	// Audio data
+  int len = c->samples*c->nch;		// Number of samples
   float curavg = 0.0, newavg, neededmul;
   int tmp;
 
@@ -161,8 +162,8 @@ static void method1_int16(af_drc_t *s, struct mp_audio *c)
 static void method1_float(af_drc_t *s, struct mp_audio *c)
 {
   register int i = 0;
-  float *data = (float*)c->audio;	// Audio data
-  int len = c->len/4;		// Number of samples
+  float *data = (float*)c->planes[0];	// Audio data
+  int len = c->samples*c->nch;		// Number of samples
   float curavg = 0.0, newavg, neededmul, tmp;
 
   for (i = 0; i < len; i++)
@@ -198,8 +199,8 @@ static void method1_float(af_drc_t *s, struct mp_audio *c)
 static void method2_int16(af_drc_t *s, struct mp_audio *c)
 {
   register int i = 0;
-  int16_t *data = (int16_t*)c->audio;	// Audio data
-  int len = c->len/2;		// Number of samples
+  int16_t *data = (int16_t*)c->planes[0];	// Audio data
+  int len = c->samples*c->nch;		// Number of samples
   float curavg = 0.0, newavg, avg = 0.0;
   int tmp, totallen = 0;
 
@@ -248,8 +249,8 @@ static void method2_int16(af_drc_t *s, struct mp_audio *c)
 static void method2_float(af_drc_t *s, struct mp_audio *c)
 {
   register int i = 0;
-  float *data = (float*)c->audio;	// Audio data
-  int len = c->len/4;		// Number of samples
+  float *data = (float*)c->planes[0];	// Audio data
+  int len = c->samples*c->nch;		// Number of samples
   float curavg = 0.0, newavg, avg = 0.0, tmp;
   int totallen = 0;
 
diff --git a/audio/filter/af_equalizer.c b/audio/filter/af_equalizer.c
index cbdcd3f84a..718445c001 100644
--- a/audio/filter/af_equalizer.c
+++ b/audio/filter/af_equalizer.c
@@ -170,9 +170,9 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* data)
 
   while(ci--){
     float*	g   = s->g[ci];      // Gain factor
-    float*	in  = ((float*)c->audio)+ci;
-    float*	out = ((float*)c->audio)+ci;
-    float* 	end = in + c->len/4; // Block loop end
+    float*	in  = ((float*)c->planes[0])+ci;
+    float*	out = ((float*)c->planes[0])+ci;
+    float* 	end = in + c->samples*c->nch; // Block loop end
 
     while(in < end){
       register int	k  = 0;		// Frequency band index
diff --git a/audio/filter/af_export.c b/audio/filter/af_export.c
index 45143075a1..a6ebdc322e 100644
--- a/audio/filter/af_export.c
+++ b/audio/filter/af_export.c
@@ -213,9 +213,9 @@ static struct mp_audio* play( struct af_instance* af, struct mp_audio* data )
 {
   struct mp_audio*   	c   = data;	     // Current working data
   af_export_t* 	s   = af->setup;     // Setup for this instance
-  int16_t* 	a   = c->audio;	     // Incomming sound
+  int16_t* 	a   = c->planes[0];	     // Incomming sound
   int 		nch = c->nch;	     // Number of channels
-  int		len = c->len/c->bps; // Number of sample in data chunk
+  int		len = c->samples*c->nch; // Number of sample in data chunk
   int 		sz  = s->sz;         // buffer size (in samples)
   int 		flag = 0;	     // Set to 1 if buffer is filled
 
diff --git a/audio/filter/af_extrastereo.c b/audio/filter/af_extrastereo.c
index 4cf27f2724..5b3792763b 100644
--- a/audio/filter/af_extrastereo.c
+++ b/audio/filter/af_extrastereo.c
@@ -49,6 +49,7 @@ static int control(struct af_instance* af, int cmd, void* arg)
     if(!arg) return AF_ERROR;
 
     mp_audio_copy_config(af->data, (struct mp_audio*)arg);
+    mp_audio_force_interleaved_format(af->data);
     mp_audio_set_num_channels(af->data, 2);
     if (af->data->format == AF_FORMAT_FLOAT_NE)
     {
@@ -83,8 +84,8 @@ static struct mp_audio* play_s16(struct af_instance* af, struct mp_audio* data)
 {
   af_extrastereo_t *s = af->setup;
   register int i = 0;
-  int16_t *a = (int16_t*)data->audio;	// Audio data
-  int len = data->len/2;		// Number of samples
+  int16_t *a = (int16_t*)data->planes[0];	// Audio data
+  int len = data->samples*data->nch;		// Number of samples
   int avg, l, r;
 
   for (i = 0; i < len; i+=2)
@@ -105,8 +106,8 @@ static struct mp_audio* play_float(struct af_instance* af, struct mp_audio* data
 {
   af_extrastereo_t *s = af->setup;
   register int i = 0;
-  float *a = (float*)data->audio;	// Audio data
-  int len = data->len/4;		// Number of samples
+  float *a = (float*)data->planes[0];	// Audio data
+  int len = data->samples * data->nch;	// Number of samples
   float avg, l, r;
 
   for (i = 0; i < len; i+=2)
diff --git a/audio/filter/af_format.c b/audio/filter/af_format.c
index 642ef927bb..b551ddba42 100644
--- a/audio/filter/af_format.c
+++ b/audio/filter/af_format.c
@@ -105,8 +105,9 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
     struct mp_audio *r = &priv->temp;
 
     *r = *af->data;
-    r->audio = data->audio;
-    r->len = data->len;
+    for (int n = 0; n < r->nch; n++)
+        r->planes[n] = data->planes[n];
+    r->samples = data->samples;
 
     return r;
 }
diff --git a/audio/filter/af_hrtf.c b/audio/filter/af_hrtf.c
index 5b80bf0eec..01148dc6a6 100644
--- a/audio/filter/af_hrtf.c
+++ b/audio/filter/af_hrtf.c
@@ -367,7 +367,7 @@ static void uninit(struct af_instance *af)
 	free(af->setup);
     }
     if(af->data)
-	free(af->data->audio);
+	free(af->data->planes[0]);
     free(af->data);
 }
 
@@ -385,9 +385,9 @@ damped (without any real 3D acoustical image, however).
 static struct mp_audio* play(struct af_instance *af, struct mp_audio *data)
 {
     af_hrtf_t *s = af->setup;
-    short *in = data->audio; // Input audio data
+    short *in = data->planes[0]; // Input audio data
     short *out = NULL; // Output audio data
-    short *end = in + data->len / sizeof(short); // Loop end
+    short *end = in + data->samples * data->nch; // Loop end
     float common, left, right, diff, left_b, right_b;
     const int dblen = s->dlbuflen, hlen = s->hrflen, blen = s->basslen;
 
@@ -425,7 +425,7 @@ static struct mp_audio* play(struct af_instance *af, struct mp_audio *data)
 		 "channel\n");
     }
 
-    out = af->data->audio;
+    out = af->data->planes[0];
 
     /* MPlayer's 5 channel layout (notation for the variable):
      *
@@ -565,8 +565,7 @@ static struct mp_audio* play(struct af_instance *af, struct mp_audio *data)
     }
 
     /* Set output data */
-    data->audio = af->data->audio;
-    data->len   = data->len / data->nch * 2;
+    data->planes[0] = af->data->planes[0];
     mp_audio_set_num_channels(data, 2);
 
     return data;
diff --git a/audio/filter/af_karaoke.c b/audio/filter/af_karaoke.c
index b24ba0d877..d9c62aa85c 100644
--- a/audio/filter/af_karaoke.c
+++ b/audio/filter/af_karaoke.c
@@ -51,9 +51,9 @@ static void uninit(struct af_instance* af)
 static struct mp_audio* play(struct af_instance* af, struct mp_audio* data)
 {
 	struct mp_audio*	c	= data;		 // Current working data
-	float*		a	= c->audio;	 // Audio data
-	int			len	= c->len/4;	 // Number of samples in current audio block
+	float*		a	= c->planes[0];	 // Audio data
 	int			nch	= c->nch;	 // Number of channels
+	int			len	= c->samples*nch;	 // Number of samples in current audio block
 	register int  i;
 
 	/*
diff --git a/audio/filter/af_ladspa.c b/audio/filter/af_ladspa.c
index 73b7430201..6acdc150d7 100644
--- a/audio/filter/af_ladspa.c
+++ b/audio/filter/af_ladspa.c
@@ -497,10 +497,6 @@ static int control(struct af_instance *af, int cmd, void *arg) {
         mp_audio_copy_config(af->data, (struct mp_audio*)arg);
         mp_audio_set_format(af->data, AF_FORMAT_FLOAT_NE);
 
-        /* arg->len is not set here yet, so init of buffers and connecting the
-         * filter, has to be done in play() :-/
-         */
-
         return af_test_output(af, (struct mp_audio*)arg);
     case AF_CONTROL_COMMAND_LINE: {
         char *buf;
@@ -710,8 +706,8 @@ static void uninit(struct af_instance *af) {
 static struct mp_audio* play(struct af_instance *af, struct mp_audio *data) {
     af_ladspa_t *setup = af->setup;
     const LADSPA_Descriptor *pdes = setup->plugin_descriptor;
-    float *audio = (float*)data->audio;
-    int nsamples = data->len/4; /* /4 because it's 32-bit float */
+    float *audio = (float*)data->planes[0];
+    int nsamples = data->samples*data->nch;
     int nch = data->nch;
     int rate = data->rate;
     int i, p;
@@ -723,10 +719,6 @@ static struct mp_audio* play(struct af_instance *af, struct mp_audio *data) {
      * plugin, connect ports and activate plugin
      */
 
-    /* 2004-12-07: Also check if the buffersize has to be changed!
-     *             data->len is not constant per se! re-init buffers.
-     */
-
     if ( (setup->bufsize != nsamples/nch) || (setup->nch != nch) ) {
 
         /* if setup->nch==0, it's the first call, if not, something has
diff --git a/audio/filter/af_lavcac3enc.c b/audio/filter/af_lavcac3enc.c
index 9ca193017a..d4abf2e942 100644
--- a/audio/filter/af_lavcac3enc.c
+++ b/audio/filter/af_lavcac3enc.c
@@ -161,7 +161,7 @@ static void uninit(struct af_instance* af)
     af_ac3enc_t *s = af->setup;
 
     if (af->data)
-        free(af->data->audio);
+        free(af->data->planes[0]);
     free(af->data);
     if (s) {
         av_free_packet(&s->pkt);
@@ -183,7 +183,7 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* audio)
     int left, outsize = 0;
     char *buf, *src;
     int max_output_len;
-    int frame_num = (audio->len + s->pending_len) / s->expect_len;
+    int frame_num = (mp_audio_psize(audio) + s->pending_len) / s->expect_len;
     int samplesize = af_fmt2bits(s->in_sampleformat) / 8;
 
     if (s->add_iec61937_header)
@@ -191,23 +191,23 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* audio)
     else
         max_output_len = AC3_MAX_CODED_FRAME_SIZE * frame_num;
 
-    if (af->data->len < max_output_len) {
+    if (mp_audio_psize(af->data) < max_output_len) {
         mp_msg(MSGT_AFILTER, MSGL_V, "[libaf] Reallocating memory in module %s, "
-               "old len = %i, new len = %i\n", af->info->name, af->data->len,
-                max_output_len);
-        free(af->data->audio);
-        af->data->audio = malloc(max_output_len);
-        if (!af->data->audio) {
+               "old len = %i, new len = %i\n", af->info->name,
+               mp_audio_psize(af->data), max_output_len);
+        free(af->data->planes[0]);
+        af->data->planes[0] = malloc(max_output_len);
+        if (!af->data->planes[0]) {
             mp_msg(MSGT_AFILTER, MSGL_FATAL, "[libaf] Could not allocate memory \n");
             return NULL;
         }
-        af->data->len = max_output_len;
+        af->data->samples = max_output_len / af->data->sstride;
     }
 
     l = af->data;           // Local data
-    buf = l->audio;
-    src = c->audio;
-    left = c->len;
+    buf = l->planes[0];
+    src = c->planes[0];
+    left = mp_audio_psize(c);
 
 
     while (left > 0) {
@@ -296,7 +296,7 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* audio)
             len = AC3_FRAME_SIZE * 2 * 2;
         }
 
-        assert(buf + len <= (char *)af->data->audio + af->data->len);
+        assert(buf + len <= (char *)af->data->planes[0] + mp_audio_psize(af->data));
         assert(s->pkt.size <= len - header_len);
 
         memcpy(buf + header_len, s->pkt.data, s->pkt.size);
@@ -304,10 +304,10 @@ static struct mp_audio* play(struct af_instance* af, struct mp_audio* audio)
         outsize += len;
         buf += len;
     }
-    c->audio = l->audio;
+    c->planes[0] = l->planes[0];
     mp_audio_set_num_channels(c, 2);
     mp_audio_set_format(c, af->data->format);
-    c->len   = outsize;
+    c->samples = outsize / c->sstride;
     mp_msg(MSGT_AFILTER, MSGL_DBG2, "play return size %d, pending %d\n",
            outsize, s->pending_len);
     return c;
diff --git a/audio/filter/af_lavfi.c b/audio/filter/af_lavfi.c
index 3c6d981f1d..f08e7ff85c 100644
--- a/audio/filter/af_lavfi.c
+++ b/audio/filter/af_lavfi.c
@@ -181,6 +181,7 @@ static int control(struct af_instance *af, int cmd, void *arg)
 
         if (af_to_avformat(in->format) == AV_SAMPLE_FMT_NONE)
             mp_audio_set_format(in, AF_FORMAT_FLOAT_NE);
+        mp_audio_force_interleaved_format(in);
 
         if (!mp_chmap_is_lavc(&in->channels))
             mp_chmap_reorder_to_lavc(&in->channels); // will always work
@@ -193,6 +194,7 @@ static int control(struct af_instance *af, int cmd, void *arg)
         out->rate = l_out->sample_rate;
 
         mp_audio_set_format(out, af_from_avformat(l_out->format));
+        mp_audio_force_interleaved_format(out);
 
         struct mp_chmap out_cm;
         mp_chmap_from_lavc(&out_cm, l_out->channel_layout);
@@ -228,7 +230,7 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
     int out_frame_size = r->bps * r->channels.num;
 
     AVFrame *frame = av_frame_alloc();
-    frame->nb_samples = data->len / in_frame_size;
+    frame->nb_samples = data->samples;
     frame->format = l_in->format;
 
     // Timebase is 1/sample_rate
@@ -238,7 +240,7 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
     av_frame_set_channel_layout(frame, l_in->channel_layout);
     av_frame_set_sample_rate(frame, l_in->sample_rate);
 
-    frame->data[0] = data->audio;
+    frame->data[0] = data->planes[0];
     frame->extended_data = frame->data;
 
     if (av_buffersrc_add_frame(p->in, frame) < 0) {
@@ -268,11 +270,11 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
         av_frame_free(&frame);
     }
 
-    r->audio = p->out_buffer;
-    r->len = out_len;
+    r->planes[0] = p->out_buffer;
+    r->samples = out_len / r->sstride;
 
-    p->bytes_in += data->len;
-    p->bytes_out += r->len;
+    p->bytes_in += data->samples * data->sstride;
+    p->bytes_out += r->samples * r->sstride;
 
     if (out_pts != AV_NOPTS_VALUE) {
         int64_t num_in_frames = p->bytes_in / in_frame_size;
@@ -280,8 +282,7 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
 
         double out_time = out_pts * av_q2d(p->timebase_out);
         // Need pts past the last output sample.
-        int out_frames = r->len / out_frame_size;
-        out_time += out_frames / (double)r->rate;
+        out_time += r->samples / (double)r->rate;
 
         af->delay = (in_time - out_time) * r->rate * out_frame_size;
     }
diff --git a/audio/filter/af_lavrresample.c b/audio/filter/af_lavrresample.c
index 860e5a52d3..a6abd32aed 100644
--- a/audio/filter/af_lavrresample.c
+++ b/audio/filter/af_lavrresample.c
@@ -244,8 +244,10 @@ static int control(struct af_instance *af, int cmd, void *arg)
 
         if (af_to_avformat(in->format) == AV_SAMPLE_FMT_NONE)
             mp_audio_set_format(in, AF_FORMAT_FLOAT_NE);
+        mp_audio_force_interleaved_format(in);
         if (af_to_avformat(out->format) == AV_SAMPLE_FMT_NONE)
             mp_audio_set_format(out, in->format);
+        mp_audio_force_interleaved_format(out);
 
         af->mul     = (double) (out->rate * out->nch) / (in->rate * in->nch);
         af->delay   = out->nch * s->opts.filter_size / FFMIN(af->mul, 1);
@@ -306,28 +308,30 @@ static struct mp_audio *play(struct af_instance *af, struct mp_audio *data)
     struct mp_audio *out  = af->data;
 
 
-    int in_size     = data->len;
-    int in_samples  = in_size / (data->bps * data->nch);
+    int in_samples  = data->samples;
+    int in_size     = data->samples * data->sstride;
     int out_samples = avresample_available(s->avrctx) +
         av_rescale_rnd(get_delay(s) + in_samples,
                        s->ctx.out_rate, s->ctx.in_rate, AV_ROUND_UP);
-    int out_size    = out->bps * out_samples * out->nch;
+    int out_size    = out_samples * out->sstride;
 
-    if (talloc_get_size(out->audio) < out_size)
-        out->audio = talloc_realloc_size(out, out->audio, out_size);
+    if (talloc_get_size(out->planes[0]) < out_size)
+        out->planes[0] = talloc_realloc_size(out, out->planes[0], out_size);
 
     af->delay = out->bps * av_rescale_r