From cf8c93ada4a7643425636ab5c3914c777fc1120f Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Mon, 27 Aug 2012 12:01:16 -0400
Subject: [PATCH 01/10] 8svx: use planar sample format

---
 libavcodec/8svx.c | 49 +++++++++++++++++++++--------------------------
 1 file changed, 22 insertions(+), 27 deletions(-)

diff --git a/libavcodec/8svx.c b/libavcodec/8svx.c
index 328fc65b41..1839617b96 100644
--- a/libavcodec/8svx.c
+++ b/libavcodec/8svx.c
@@ -58,30 +58,25 @@ static const int8_t exponential[16] = { -128, -64, -32, -16, -8, -4, -2, -1,
  * @param[in,out] state starting value. it is saved for use in the next call.
  */
 static void delta_decode(uint8_t *dst, const uint8_t *src, int src_size,
-                         uint8_t *state, const int8_t *table, int channels)
+                         uint8_t *state, const int8_t *table)
 {
     uint8_t val = *state;
 
     while (src_size--) {
         uint8_t d = *src++;
         val = av_clip_uint8(val + table[d & 0xF]);
-        *dst = val;
-        dst += channels;
+        *dst++ = val;
         val = av_clip_uint8(val + table[d >> 4]);
-        *dst = val;
-        dst += channels;
+        *dst++ = val;
     }
 
     *state = val;
 }
 
-static void raw_decode(uint8_t *dst, const int8_t *src, int src_size,
-                       int channels)
+static void raw_decode(uint8_t *dst, const int8_t *src, int src_size)
 {
-    while (src_size--) {
-        *dst = *src++ + 128;
-        dst += channels;
-    }
+    while (src_size--)
+        *dst++ = *src++ + 128;
 }
 
 /** decode a frame */
@@ -90,8 +85,7 @@ static int eightsvx_decode_frame(AVCodecContext *avctx, void *data,
 {
     EightSvxContext *esc = avctx->priv_data;
     int buf_size;
-    uint8_t *out_data;
-    int ret;
+    int ch, ret;
     int is_compr = (avctx->codec_id != AV_CODEC_ID_PCM_S8_PLANAR);
 
     /* for the first packet, copy data to buffer */
@@ -146,22 +140,17 @@ static int eightsvx_decode_frame(AVCodecContext *avctx, void *data,
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return ret;
     }
-    out_data = esc->frame.data[0];
 
-    if (is_compr) {
-    delta_decode(out_data, &esc->data[0][esc->data_idx], buf_size,
-                 &esc->fib_acc[0], esc->table, avctx->channels);
-    if (avctx->channels == 2) {
-        delta_decode(&out_data[1], &esc->data[1][esc->data_idx], buf_size,
-                    &esc->fib_acc[1], esc->table, avctx->channels);
-    }
-    } else {
-        int ch;
-        for (ch = 0; ch < avctx->channels; ch++) {
-            raw_decode((int8_t *)&out_data[ch], &esc->data[ch][esc->data_idx],
-                       buf_size, avctx->channels);
+    for (ch = 0; ch < avctx->channels; ch++) {
+        if (is_compr) {
+            delta_decode(esc->frame.data[ch], &esc->data[ch][esc->data_idx],
+                         buf_size, &esc->fib_acc[ch], esc->table);
+        } else {
+            raw_decode(esc->frame.data[ch], &esc->data[ch][esc->data_idx],
+                       buf_size);
         }
     }
+
     esc->data_idx += buf_size;
 
     *got_frame_ptr   = 1;
@@ -192,7 +181,7 @@ static av_cold int eightsvx_decode_init(AVCodecContext *avctx)
         default:
           return -1;
     }
-    avctx->sample_fmt = AV_SAMPLE_FMT_U8;
+    avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
 
     avcodec_get_frame_defaults(&esc->frame);
     avctx->coded_frame = &esc->frame;
@@ -220,6 +209,8 @@ AVCodec ff_eightsvx_fib_decoder = {
   .decode         = eightsvx_decode_frame,
   .capabilities   = CODEC_CAP_DELAY | CODEC_CAP_DR1,
   .long_name      = NULL_IF_CONFIG_SMALL("8SVX fibonacci"),
+  .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
+                                                    AV_SAMPLE_FMT_NONE },
 };
 
 AVCodec ff_eightsvx_exp_decoder = {
@@ -232,6 +223,8 @@ AVCodec ff_eightsvx_exp_decoder = {
   .decode         = eightsvx_decode_frame,
   .capabilities   = CODEC_CAP_DELAY | CODEC_CAP_DR1,
   .long_name      = NULL_IF_CONFIG_SMALL("8SVX exponential"),
+  .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
+                                                    AV_SAMPLE_FMT_NONE },
 };
 
 AVCodec ff_pcm_s8_planar_decoder = {
@@ -244,4 +237,6 @@ AVCodec ff_pcm_s8_planar_decoder = {
     .decode         = eightsvx_decode_frame,
     .capabilities   = CODEC_CAP_DELAY | CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("PCM signed 8-bit planar"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
+                                                      AV_SAMPLE_FMT_NONE },
 };

From 461ba7e97a89ab3e553b0690ce8f87f9f484bf90 Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Sat, 25 Aug 2012 19:23:55 -0400
Subject: [PATCH 02/10] apedec: output in planar sample format

---
 libavcodec/apedec.c | 39 ++++++++++++++++++++-------------------
 1 file changed, 20 insertions(+), 19 deletions(-)

diff --git a/libavcodec/apedec.c b/libavcodec/apedec.c
index 59179032cb..9d2941ce15 100644
--- a/libavcodec/apedec.c
+++ b/libavcodec/apedec.c
@@ -196,13 +196,13 @@ static av_cold int ape_decode_init(AVCodecContext *avctx)
     s->bps = avctx->bits_per_coded_sample;
     switch (s->bps) {
     case 8:
-        avctx->sample_fmt = AV_SAMPLE_FMT_U8;
+        avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
         break;
     case 16:
-        avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+        avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
         break;
     case 24:
-        avctx->sample_fmt = AV_SAMPLE_FMT_S32;
+        avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
         break;
     default:
         av_log_ask_for_sample(avctx, "Unsupported bits per coded sample %d\n",
@@ -830,7 +830,7 @@ static int ape_decode_frame(AVCodecContext *avctx, void *data,
     uint8_t *sample8;
     int16_t *sample16;
     int32_t *sample24;
-    int i, ret;
+    int i, ch, ret;
     int blockstodecode;
     int bytes_used = 0;
 
@@ -930,27 +930,24 @@ static int ape_decode_frame(AVCodecContext *avctx, void *data,
 
     switch (s->bps) {
     case 8:
-        sample8 = (uint8_t *)s->frame.data[0];
-        for (i = 0; i < blockstodecode; i++) {
-            *sample8++ = (s->decoded[0][i] + 0x80) & 0xff;
-            if (s->channels == 2)
-                *sample8++ = (s->decoded[1][i] + 0x80) & 0xff;
+        for (ch = 0; ch < s->channels; ch++) {
+            sample8 = (uint8_t *)s->frame.data[ch];
+            for (i = 0; i < blockstodecode; i++)
+                *sample8++ = (s->decoded[ch][i] + 0x80) & 0xff;
         }
         break;
     case 16:
-        sample16 = (int16_t *)s->frame.data[0];
-        for (i = 0; i < blockstodecode; i++) {
-            *sample16++ = s->decoded[0][i];
-            if (s->channels == 2)
-                *sample16++ = s->decoded[1][i];
+        for (ch = 0; ch < s->channels; ch++) {
+            sample16 = (int16_t *)s->frame.data[ch];
+            for (i = 0; i < blockstodecode; i++)
+                *sample16++ = s->decoded[ch][i];
         }
         break;
     case 24:
-        sample24 = (int32_t *)s->frame.data[0];
-        for (i = 0; i < blockstodecode; i++) {
-            *sample24++ = s->decoded[0][i] << 8;
-            if (s->channels == 2)
-                *sample24++ = s->decoded[1][i] << 8;
+        for (ch = 0; ch < s->channels; ch++) {
+            sample24 = (int32_t *)s->frame.data[ch];
+            for (i = 0; i < blockstodecode; i++)
+                *sample24++ = s->decoded[ch][i] << 8;
         }
         break;
     }
@@ -995,5 +992,9 @@ AVCodec ff_ape_decoder = {
     .capabilities   = CODEC_CAP_SUBFRAMES | CODEC_CAP_DELAY | CODEC_CAP_DR1,
     .flush          = ape_flush,
     .long_name      = NULL_IF_CONFIG_SMALL("Monkey's Audio"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
+                                                      AV_SAMPLE_FMT_S16P,
+                                                      AV_SAMPLE_FMT_S32P,
+                                                      AV_SAMPLE_FMT_NONE },
     .priv_class     = &ape_decoder_class,
 };

From 9af4eaa8ea03ee8474ec4ad5cf94bfccc954641a Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Sun, 26 Aug 2012 12:48:07 -0400
Subject: [PATCH 03/10] atrac3: use float planar sample format

---
 libavcodec/atrac3.c | 44 +++++++-------------------------------------
 1 file changed, 7 insertions(+), 37 deletions(-)

diff --git a/libavcodec/atrac3.c b/libavcodec/atrac3.c
index a09007c20f..94b355a362 100644
--- a/libavcodec/atrac3.c
+++ b/libavcodec/atrac3.c
@@ -111,7 +111,6 @@ typedef struct {
     //@}
     //@{
     /** data buffers */
-    float              *outSamples[2];
     uint8_t*            decoded_bytes_buffer;
     float               tempBuf[1070];
     //@}
@@ -197,7 +196,7 @@ static int decode_bytes(const uint8_t* inbuffer, uint8_t* out, int bytes){
 }
 
 
-static av_cold int init_atrac3_transforms(ATRAC3Context *q, int is_float) {
+static av_cold int init_atrac3_transforms(ATRAC3Context *q) {
     float enc_window[256];
     int i;
 
@@ -213,7 +212,7 @@ static av_cold int init_atrac3_transforms(ATRAC3Context *q, int is_float) {
         }
 
     /* Initialize the MDCT transform. */
-    return ff_mdct_init(&q->mdct_ctx, 9, 1, is_float ? 1.0 / 32768 : 1.0);
+    return ff_mdct_init(&q->mdct_ctx, 9, 1, 1.0 / 32768);
 }
 
 /**
@@ -226,7 +225,6 @@ static av_cold int atrac3_decode_close(AVCodecContext *avctx)
 
     av_free(q->pUnits);
     av_free(q->decoded_bytes_buffer);
-    av_freep(&q->outSamples[0]);
 
     ff_mdct_end(&q->mdct_ctx);
 
@@ -837,8 +835,6 @@ static int atrac3_decode_frame(AVCodecContext *avctx, void *data,
     ATRAC3Context *q = avctx->priv_data;
     int result;
     const uint8_t* databuf;
-    float   *samples_flt;
-    int16_t *samples_s16;
 
     if (buf_size < avctx->block_align) {
         av_log(avctx, AV_LOG_ERROR,
@@ -852,8 +848,6 @@ static int atrac3_decode_frame(AVCodecContext *avctx, void *data,
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return result;
     }
-    samples_flt = (float   *)q->frame.data[0];
-    samples_s16 = (int16_t *)q->frame.data[0];
 
     /* Check if we need to descramble and what buffer to pass on. */
     if (q->scrambled_stream) {
@@ -863,27 +857,13 @@ static int atrac3_decode_frame(AVCodecContext *avctx, void *data,
         databuf = buf;
     }
 
-    if (q->channels == 1 && avctx->sample_fmt == AV_SAMPLE_FMT_FLT)
-        result = decodeFrame(q, databuf, &samples_flt);
-    else
-        result = decodeFrame(q, databuf, q->outSamples);
+    result = decodeFrame(q, databuf, (float **)q->frame.extended_data);
 
     if (result != 0) {
         av_log(NULL,AV_LOG_ERROR,"Frame decoding error!\n");
         return result;
     }
 
-    /* interleave */
-    if (q->channels == 2 && avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
-        q->fmt_conv.float_interleave(samples_flt,
-                                     (const float **)q->outSamples,
-                                     SAMPLES_PER_FRAME, 2);
-    } else if (avctx->sample_fmt == AV_SAMPLE_FMT_S16) {
-        q->fmt_conv.float_to_int16_interleave(samples_s16,
-                                              (const float **)q->outSamples,
-                                              SAMPLES_PER_FRAME, q->channels);
-    }
-
     *got_frame_ptr   = 1;
     *(AVFrame *)data = q->frame;
 
@@ -1005,12 +985,9 @@ static av_cold int atrac3_decode_init(AVCodecContext *avctx)
         vlcs_initialized = 1;
     }
 
-    if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT)
-        avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
-    else
-        avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+    avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
 
-    if ((ret = init_atrac3_transforms(q, avctx->sample_fmt == AV_SAMPLE_FMT_FLT))) {
+    if ((ret = init_atrac3_transforms(q))) {
         av_log(avctx, AV_LOG_ERROR, "Error initializing MDCT\n");
         av_freep(&q->decoded_bytes_buffer);
         return ret;
@@ -1048,15 +1025,6 @@ static av_cold int atrac3_decode_init(AVCodecContext *avctx)
         return AVERROR(ENOMEM);
     }
 
-    if (avctx->channels > 1 || avctx->sample_fmt == AV_SAMPLE_FMT_S16) {
-        q->outSamples[0] = av_mallocz(SAMPLES_PER_FRAME * avctx->channels * sizeof(*q->outSamples[0]));
-        q->outSamples[1] = q->outSamples[0] + SAMPLES_PER_FRAME;
-        if (!q->outSamples[0]) {
-            atrac3_decode_close(avctx);
-            return AVERROR(ENOMEM);
-        }
-    }
-
     avcodec_get_frame_defaults(&q->frame);
     avctx->coded_frame = &q->frame;
 
@@ -1075,4 +1043,6 @@ AVCodec ff_atrac3_decoder =
     .decode         = atrac3_decode_frame,
     .capabilities   = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("Atrac 3 (Adaptive TRansform Acoustic Coding 3)"),
+    .sample_fmts     = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
+                                                       AV_SAMPLE_FMT_NONE },
 };

From cbf6ee78236659e2ef2c7b0d0b17c270dddfb37c Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Mon, 27 Aug 2012 10:38:36 -0400
Subject: [PATCH 04/10] cook: use planar sample format

---
 libavcodec/cook.c | 40 ++++++++++++++++++++++------------------
 1 file changed, 22 insertions(+), 18 deletions(-)

diff --git a/libavcodec/cook.c b/libavcodec/cook.c
index 1dec398748..c5b17f98da 100644
--- a/libavcodec/cook.c
+++ b/libavcodec/cook.c
@@ -119,9 +119,10 @@ typedef struct cook {
     void (*interpolate)(struct cook *q, float *buffer,
                         int gain_index, int gain_index_next);
 
-    void (*saturate_output)(struct cook *q, int chan, float *out);
+    void (*saturate_output)(struct cook *q, float *out);
 
     AVCodecContext*     avctx;
+    DSPContext          dsp;
     AVFrame             frame;
     GetBitContext       gb;
     /* stream data */
@@ -876,18 +877,15 @@ static inline void decode_bytes_and_gain(COOKContext *q, COOKSubpacket *p,
  * Saturate the output signal and interleave.
  *
  * @param q                 pointer to the COOKContext
- * @param chan              channel to saturate
  * @param out               pointer to the output vector
  */
-static void saturate_output_float(COOKContext *q, int chan, float *out)
+static void saturate_output_float(COOKContext *q, float *out)
 {
-    int j;
-    float *output = q->mono_mdct_output + q->samples_per_channel;
-    for (j = 0; j < q->samples_per_channel; j++) {
-        out[chan + q->nb_channels * j] = av_clipf(output[j], -1.0, 1.0);
-    }
+    q->dsp.vector_clipf(out, q->mono_mdct_output + q->samples_per_channel,
+                        -1.0f, 1.0f, FFALIGN(q->samples_per_channel, 8));
 }
 
+
 /**
  * Final part of subpacket decoding:
  *  Apply modulated lapped transform, gain compensation,
@@ -898,15 +896,14 @@ static void saturate_output_float(COOKContext *q, int chan, float *out)
  * @param gains_ptr         array of current/prev gain pointers
  * @param previous_buffer   pointer to the previous buffer to be used for overlapping
  * @param out               pointer to the output buffer
- * @param chan              0: left or single channel, 1: right channel
  */
 static inline void mlt_compensate_output(COOKContext *q, float *decode_buffer,
                                          cook_gains *gains_ptr, float *previous_buffer,
-                                         float *out, int chan)
+                                         float *out)
 {
     imlt_gain(q, decode_buffer, gains_ptr, previous_buffer);
     if (out)
-        q->saturate_output(q, chan, out);
+        q->saturate_output(q, out);
 }
 
 
@@ -919,7 +916,7 @@ static inline void mlt_compensate_output(COOKContext *q, float *decode_buffer,
  * @param outbuffer         pointer to the outbuffer
  */
 static int decode_subpacket(COOKContext *q, COOKSubpacket *p,
-                            const uint8_t *inbuffer, float *outbuffer)
+                            const uint8_t *inbuffer, float **outbuffer)
 {
     int sub_packet_size = p->size;
     int res;
@@ -942,15 +939,18 @@ static int decode_subpacket(COOKContext *q, COOKSubpacket *p,
     }
 
     mlt_compensate_output(q, q->decode_buffer_1, &p->gains1,
-                          p->mono_previous_buffer1, outbuffer, p->ch_idx);
+                          p->mono_previous_buffer1,
+                          outbuffer ? outbuffer[p->ch_idx] : NULL);
 
     if (p->num_channels == 2)
         if (p->joint_stereo)
             mlt_compensate_output(q, q->decode_buffer_2, &p->gains1,
-                                  p->mono_previous_buffer2, outbuffer, p->ch_idx + 1);
+                                  p->mono_previous_buffer2,
+                                  outbuffer ? outbuffer[p->ch_idx + 1] : NULL);
         else
             mlt_compensate_output(q, q->decode_buffer_2, &p->gains2,
-                                  p->mono_previous_buffer2, outbuffer, p->ch_idx + 1);
+                                  p->mono_previous_buffer2,
+                                  outbuffer ? outbuffer[p->ch_idx + 1] : NULL);
 
     return 0;
 }
@@ -967,7 +967,7 @@ static int cook_decode_frame(AVCodecContext *avctx, void *data,
     const uint8_t *buf = avpkt->data;
     int buf_size = avpkt->size;
     COOKContext *q = avctx->priv_data;
-    float *samples = NULL;
+    float **samples = NULL;
     int i, ret;
     int offset = 0;
     int chidx = 0;
@@ -982,7 +982,7 @@ static int cook_decode_frame(AVCodecContext *avctx, void *data,
             av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
             return ret;
         }
-        samples = (float *) q->frame.data[0];
+        samples = (float **)q->frame.extended_data;
     }
 
     /* estimate subpacket sizes */
@@ -1099,6 +1099,8 @@ static av_cold int cook_decode_init(AVCodecContext *avctx)
     /* Initialize RNG. */
     av_lfg_init(&q->random_state, 0);
 
+    ff_dsputil_init(&q->dsp, avctx);
+
     while (edata_ptr < edata_ptr_end) {
         /* 8 for mono, 16 for stereo, ? for multichannel
            Swap to right endianness so we don't need to care later on. */
@@ -1274,7 +1276,7 @@ static av_cold int cook_decode_init(AVCodecContext *avctx)
         return AVERROR_PATCHWELCOME;
     }
 
-    avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
+    avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
     if (channel_mask)
         avctx->channel_layout = channel_mask;
     else
@@ -1299,4 +1301,6 @@ AVCodec ff_cook_decoder = {
     .decode         = cook_decode_frame,
     .capabilities   = CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("Cook / Cooker / Gecko (RealAudio G2)"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
+                                                      AV_SAMPLE_FMT_NONE },
 };

From 64c312aa297b40bccee607574cfdfa8d14abe9ba Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Mon, 27 Aug 2012 11:43:34 -0400
Subject: [PATCH 05/10] dcadec: use float planar sample format

---
 libavcodec/dcadec.c | 119 ++++++++++++++++++--------------------------
 1 file changed, 49 insertions(+), 70 deletions(-)

diff --git a/libavcodec/dcadec.c b/libavcodec/dcadec.c
index a0a5ea948d..d38dff81b3 100644
--- a/libavcodec/dcadec.c
+++ b/libavcodec/dcadec.c
@@ -354,11 +354,9 @@ typedef struct {
     DECLARE_ALIGNED(32, float, raXin)[32];
 
     int output;                 ///< type of output
-    float scale_bias;           ///< output scale
 
     DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
-    DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX + 1) * 256];
-    const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
+    float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
 
     uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
     int dca_buffer_size;        ///< how much data is in the dca_buffer
@@ -1007,20 +1005,20 @@ static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
 }
 
 /* downmixing routines */
-#define MIX_REAR1(samples, si1, rs, coef)           \
-    samples[i]     += samples[si1] * coef[rs][0];   \
-    samples[i+256] += samples[si1] * coef[rs][1];
+#define MIX_REAR1(samples, s1, rs, coef)            \
+    samples[0][i] += samples[s1][i] * coef[rs][0];  \
+    samples[1][i] += samples[s1][i] * coef[rs][1];
 
-#define MIX_REAR2(samples, si1, si2, rs, coef)                                     \
-    samples[i]     += samples[si1] * coef[rs][0] + samples[si2] * coef[rs + 1][0]; \
-    samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs + 1][1];
+#define MIX_REAR2(samples, s1, s2, rs, coef)                                          \
+    samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
+    samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
 
 #define MIX_FRONT3(samples, coef)                                      \
-    t = samples[i + c];                                                \
-    u = samples[i + l];                                                \
-    v = samples[i + r];                                                \
-    samples[i]     = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
-    samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
+    t = samples[c][i];                                                 \
+    u = samples[l][i];                                                 \
+    v = samples[r][i];                                                 \
+    samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0];  \
+    samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
 
 #define DOWNMIX_TO_STEREO(op1, op2)             \
     for (i = 0; i < 256; i++) {                 \
@@ -1028,7 +1026,7 @@ static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
         op2                                     \
     }
 
-static void dca_downmix(float *samples, int srcfmt,
+static void dca_downmix(float **samples, int srcfmt,
                         int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
                         const int8_t *channel_mapping)
 {
@@ -1053,36 +1051,36 @@ static void dca_downmix(float *samples, int srcfmt,
     case DCA_STEREO:
         break;
     case DCA_3F:
-        c = channel_mapping[0] * 256;
-        l = channel_mapping[1] * 256;
-        r = channel_mapping[2] * 256;
+        c = channel_mapping[0];
+        l = channel_mapping[1];
+        r = channel_mapping[2];
         DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
         break;
     case DCA_2F1R:
-        s = channel_mapping[2] * 256;
-        DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef), );
+        s = channel_mapping[2];
+        DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
         break;
     case DCA_3F1R:
-        c = channel_mapping[0] * 256;
-        l = channel_mapping[1] * 256;
-        r = channel_mapping[2] * 256;
-        s = channel_mapping[3] * 256;
+        c = channel_mapping[0];
+        l = channel_mapping[1];
+        r = channel_mapping[2];
+        s = channel_mapping[3];
         DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
-                          MIX_REAR1(samples, i + s, 3, coef));
+                          MIX_REAR1(samples, s, 3, coef));
         break;
     case DCA_2F2R:
-        sl = channel_mapping[2] * 256;
-        sr = channel_mapping[3] * 256;
-        DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef), );
+        sl = channel_mapping[2];
+        sr = channel_mapping[3];
+        DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
         break;
     case DCA_3F2R:
-        c  = channel_mapping[0] * 256;
-        l  = channel_mapping[1] * 256;
-        r  = channel_mapping[2] * 256;
-        sl = channel_mapping[3] * 256;
-        sr = channel_mapping[4] * 256;
+        c  = channel_mapping[0];
+        l  = channel_mapping[1];
+        r  = channel_mapping[2];
+        sl = channel_mapping[3];
+        sr = channel_mapping[4];
         DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
-                          MIX_REAR2(samples, i + sl, i + sr, 3, coef));
+                          MIX_REAR2(samples, sl, sr, 3, coef));
         break;
     }
 }
@@ -1279,21 +1277,21 @@ static int dca_filter_channels(DCAContext *s, int block_index)
 /*        static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
                                             0, 8388608.0, 8388608.0 };*/
         qmf_32_subbands(s, k, subband_samples[k],
-                        &s->samples[256 * s->channel_order_tab[k]],
-                        M_SQRT1_2 * s->scale_bias /* pcm_to_double[s->source_pcm_res] */);
+                        s->samples_chanptr[s->channel_order_tab[k]],
+                        M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
     }
 
     /* Down mixing */
     if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
-        dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
+        dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
     }
 
     /* Generate LFE samples for this subsubframe FIXME!!! */
     if (s->output & DCA_LFE) {
         lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
                               s->lfe_data + 2 * s->lfe * (block_index + 4),
-                              &s->samples[256 * dca_lfe_index[s->amode]],
-                              (1.0 / 256.0) * s->scale_bias);
+                              s->samples_chanptr[dca_lfe_index[s->amode]],
+                              1.0 / (256.0 * 32768.0));
         /* Outputs 20bits pcm samples */
     }
 
@@ -1656,8 +1654,7 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data,
     int lfe_samples;
     int num_core_channels = 0;
     int i, ret;
-    float   *samples_flt;
-    int16_t *samples_s16;
+    float  **samples_flt;
     DCAContext *s = avctx->priv_data;
     int channels;
     int core_ss_end;
@@ -1853,33 +1850,26 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data,
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return ret;
     }
-    samples_flt = (float *)   s->frame.data[0];
-    samples_s16 = (int16_t *) s->frame.data[0];
+    samples_flt = (float  **) s->frame.extended_data;
 
     /* filter to get final output */
     for (i = 0; i < (s->sample_blocks / 8); i++) {
+        int ch;
+
+        for (ch = 0; ch < channels; ch++)
+            s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
+
         dca_filter_channels(s, i);
 
         /* If this was marked as a DTS-ES stream we need to subtract back- */
         /* channel from SL & SR to remove matrixed back-channel signal */
         if ((s->source_pcm_res & 1) && s->xch_present) {
-            float *back_chan = s->samples + s->channel_order_tab[s->xch_base_channel]     * 256;
-            float *lt_chan   = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
-            float *rt_chan   = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
+            float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
+            float *lt_chan   = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
+            float *rt_chan   = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
             s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
             s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
         }
-
-        if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
-            s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
-                                         channels);
-            samples_flt += 256 * channels;
-        } else {
-            s->fmt_conv.float_to_int16_interleave(samples_s16,
-                                                  s->samples_chanptr, 256,
-                                                  channels);
-            samples_s16 += 256 * channels;
-        }
     }
 
     /* update lfe history */
@@ -1904,7 +1894,6 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data,
 static av_cold int dca_decode_init(AVCodecContext *avctx)
 {
     DCAContext *s = avctx->priv_data;
-    int i;
 
     s->avctx = avctx;
     dca_init_vlcs();
@@ -1915,16 +1904,7 @@ static av_cold int dca_decode_init(AVCodecContext *avctx)
     ff_dcadsp_init(&s->dcadsp);
     ff_fmt_convert_init(&s->fmt_conv, avctx);
 
-    for (i = 0; i < DCA_PRIM_CHANNELS_MAX + 1; i++)
-        s->samples_chanptr[i] = s->samples + i * 256;
-
-    if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
-        avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
-        s->scale_bias     = 1.0 / 32768.0;
-    } else {
-        avctx->sample_fmt = AV_SAMPLE_FMT_S16;
-        s->scale_bias     = 1.0;
-    }
+    avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
 
     /* allow downmixing to stereo */
     if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
@@ -1964,8 +1944,7 @@ AVCodec ff_dca_decoder = {
     .close           = dca_decode_end,
     .long_name       = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
     .capabilities    = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
-    .sample_fmts     = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
-                                                       AV_SAMPLE_FMT_S16,
+    .sample_fmts     = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
                                                        AV_SAMPLE_FMT_NONE },
     .profiles        = NULL_IF_CONFIG_SMALL(profiles),
 };

From 3fca0d72105a607926adea702065b1a00adb0b5d Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Mon, 27 Aug 2012 12:58:25 -0400
Subject: [PATCH 06/10] iac/imc: use planar sample format

---
 libavcodec/imc.c | 26 +++++++++++---------------
 1 file changed, 11 insertions(+), 15 deletions(-)

diff --git a/libavcodec/imc.c b/libavcodec/imc.c
index 8e6b5bfb31..1156e8aac2 100644
--- a/libavcodec/imc.c
+++ b/libavcodec/imc.c
@@ -241,7 +241,7 @@ static av_cold int imc_decode_init(AVCodecContext *avctx)
         return ret;
     }
     ff_dsputil_init(&q->dsp, avctx);
-    avctx->sample_fmt     = AV_SAMPLE_FMT_FLT;
+    avctx->sample_fmt     = AV_SAMPLE_FMT_FLTP;
     avctx->channel_layout = avctx->channels == 1 ? AV_CH_LAYOUT_MONO
                                                  : AV_CH_LAYOUT_STEREO;
 
@@ -661,7 +661,7 @@ static void imc_imdct256(IMCContext *q, IMCChannel *chctx, int channels)
     int i;
     float re, im;
     float *dst1 = q->out_samples;
-    float *dst2 = q->out_samples + (COEFFS - 1) * channels;
+    float *dst2 = q->out_samples + (COEFFS - 1);
 
     /* prerotation */
     for (i = 0; i < COEFFS / 2; i++) {
@@ -683,8 +683,8 @@ static void imc_imdct256(IMCContext *q, IMCChannel *chctx, int channels)
                + (q->mdct_sine_window[i * 2] * re);
         *dst2 =  (q->mdct_sine_window[i * 2] * chctx->last_fft_im[i])
                - (q->mdct_sine_window[COEFFS - 1 - i * 2] * re);
-        dst1 += channels * 2;
-        dst2 -= channels * 2;
+        dst1 += 2;
+        dst2 -= 2;
         chctx->last_fft_im[i] = im;
     }
 }
@@ -785,7 +785,6 @@ static int imc_decode_block(AVCodecContext *avctx, IMCContext *q, int ch)
         chctx->decoder_reset = 1;
 
     if (chctx->decoder_reset) {
-        memset(q->out_samples, 0, COEFFS * sizeof(*q->out_samples));
         for (i = 0; i < BANDS; i++)
             chctx->old_floor[i] = 1.0;
         for (i = 0; i < COEFFS; i++)
@@ -944,7 +943,7 @@ static int imc_decode_frame(AVCodecContext *avctx, void *data,
     }
 
     for (i = 0; i < avctx->channels; i++) {
-        q->out_samples = (float*)q->frame.data[0] + i;
+        q->out_samples = (float *)q->frame.extended_data[i];
 
         q->dsp.bswap16_buf(buf16, (const uint16_t*)buf, IMC_BLOCK_SIZE / 2);
 
@@ -957,15 +956,8 @@ static int imc_decode_frame(AVCodecContext *avctx, void *data,
     }
 
     if (avctx->channels == 2) {
-        float *src = (float*)q->frame.data[0], t1, t2;
-
-        for (i = 0; i < COEFFS; i++) {
-            t1     = src[0];
-            t2     = src[1];
-            src[0] = t1 + t2;
-            src[1] = t1 - t2;
-            src   += 2;
-        }
+        q->dsp.butterflies_float((float *)q->frame.extended_data[0],
+                                 (float *)q->frame.extended_data[1], COEFFS);
     }
 
     *got_frame_ptr   = 1;
@@ -995,6 +987,8 @@ AVCodec ff_imc_decoder = {
     .decode         = imc_decode_frame,
     .capabilities   = CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("IMC (Intel Music Coder)"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
+                                                      AV_SAMPLE_FMT_NONE },
 };
 
 AVCodec ff_iac_decoder = {
@@ -1007,4 +1001,6 @@ AVCodec ff_iac_decoder = {
     .decode         = imc_decode_frame,
     .capabilities   = CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("IAC (Indeo Audio Coder)"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
+                                                      AV_SAMPLE_FMT_NONE },
 };

From 1a3459033dc94d3f6e1b7e7c4de227fda369f2bf Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Tue, 28 Aug 2012 09:11:45 -0400
Subject: [PATCH 07/10] mpc7/8: use planar sample format

---
 libavcodec/mpc.c  | 14 +++++---------
 libavcodec/mpc.h  |  2 +-
 libavcodec/mpc7.c |  6 ++++--
 libavcodec/mpc8.c |  7 +++++--
 4 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/libavcodec/mpc.c b/libavcodec/mpc.c
index 6b15a33e5a..5a54a9bad9 100644
--- a/libavcodec/mpc.c
+++ b/libavcodec/mpc.c
@@ -43,28 +43,24 @@ void ff_mpc_init(void)
 /**
  * Process decoded Musepack data and produce PCM
  */
-static void mpc_synth(MPCContext *c, int16_t *out, int channels)
+static void mpc_synth(MPCContext *c, int16_t **out, int channels)
 {
     int dither_state = 0;
     int i, ch;
-    OUT_INT samples[MPA_MAX_CHANNELS * MPA_FRAME_SIZE], *samples_ptr;
 
     for(ch = 0;  ch < channels; ch++){
-        samples_ptr = samples + ch;
         for(i = 0; i < SAMPLES_PER_BAND; i++) {
             ff_mpa_synth_filter_fixed(&c->mpadsp,
                                 c->synth_buf[ch], &(c->synth_buf_offset[ch]),
                                 ff_mpa_synth_window_fixed, &dither_state,
-                                samples_ptr, channels,
+                                out[ch] + 32 * i, 1,
                                 c->sb_samples[ch][i]);
-            samples_ptr += 32 * channels;
         }
     }
-    for(i = 0; i < MPC_FRAME_SIZE*channels; i++)
-        *out++=samples[i];
 }
 
-void ff_mpc_dequantize_and_synth(MPCContext * c, int maxband, void *data, int channels)
+void ff_mpc_dequantize_and_synth(MPCContext * c, int maxband, int16_t **out,
+                                 int channels)
 {
     int i, j, ch;
     Band *bands = c->bands;
@@ -100,5 +96,5 @@ void ff_mpc_dequantize_and_synth(MPCContext * c, int maxband, void *data, int ch
         }
     }
 
-    mpc_synth(c, data, channels);
+    mpc_synth(c, out, channels);
 }
diff --git a/libavcodec/mpc.h b/libavcodec/mpc.h
index 1a6e7943af..2ee3c6d416 100644
--- a/libavcodec/mpc.h
+++ b/libavcodec/mpc.h
@@ -73,6 +73,6 @@ typedef struct {
 } MPCContext;
 
 void ff_mpc_init(void);
-void ff_mpc_dequantize_and_synth(MPCContext *c, int maxband, void *dst, int channels);
+void ff_mpc_dequantize_and_synth(MPCContext *c, int maxband, int16_t **out, int channels);
 
 #endif /* AVCODEC_MPC_H */
diff --git a/libavcodec/mpc7.c b/libavcodec/mpc7.c
index 0d7e9dfc28..b013eeb3d5 100644
--- a/libavcodec/mpc7.c
+++ b/libavcodec/mpc7.c
@@ -94,7 +94,7 @@ static av_cold int mpc7_decode_init(AVCodecContext * avctx)
             c->IS, c->MSS, c->gapless, c->lastframelen, c->maxbands);
     c->frames_to_skip = 0;
 
-    avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+    avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
     avctx->channel_layout = AV_CH_LAYOUT_STEREO;
 
     if(vlc_initialized) return 0;
@@ -293,7 +293,7 @@ static int mpc7_decode_frame(AVCodecContext * avctx, void *data,
         for(ch = 0; ch < 2; ch++)
             idx_to_quant(c, &gb, bands[i].res[ch], c->Q[ch] + off);
 
-    ff_mpc_dequantize_and_synth(c, mb, c->frame.data[0], 2);
+    ff_mpc_dequantize_and_synth(c, mb, (int16_t **)c->frame.extended_data, 2);
 
     bits_used = get_bits_count(&gb);
     bits_avail = buf_size * 8;
@@ -340,4 +340,6 @@ AVCodec ff_mpc7_decoder = {
     .flush          = mpc7_decode_flush,
     .capabilities   = CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("Musepack SV7"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16P,
+                                                      AV_SAMPLE_FMT_NONE },
 };
diff --git a/libavcodec/mpc8.c b/libavcodec/mpc8.c
index 79225c4275..91e228b07c 100644
--- a/libavcodec/mpc8.c
+++ b/libavcodec/mpc8.c
@@ -135,7 +135,7 @@ static av_cold int mpc8_decode_init(AVCodecContext * avctx)
     c->MSS = get_bits1(&gb);
     c->frames = 1 << (get_bits(&gb, 3) * 2);
 
-    avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+    avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
     avctx->channel_layout = (avctx->channels==2) ? AV_CH_LAYOUT_STEREO : AV_CH_LAYOUT_MONO;
 
     if(vlc_initialized) return 0;
@@ -405,7 +405,8 @@ static int mpc8_decode_frame(AVCodecContext * avctx, void *data,
         }
     }
 
-    ff_mpc_dequantize_and_synth(c, maxband - 1, c->frame.data[0],
+    ff_mpc_dequantize_and_synth(c, maxband - 1,
+                                (int16_t **)c->frame.extended_data,
                                 avctx->channels);
 
     c->cur_frame++;
@@ -438,4 +439,6 @@ AVCodec ff_mpc8_decoder = {
     .flush          = mpc8_decode_flush,
     .capabilities   = CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("Musepack SV8"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16P,
+                                                      AV_SAMPLE_FMT_NONE },
 };

From a34be78546d23e3dda05ef2e04d8f2345bc6eb94 Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Tue, 28 Aug 2012 13:04:49 -0400
Subject: [PATCH 08/10] ralf: use planar sample format

---
 libavcodec/ralf.c | 35 ++++++++++++++++++++---------------
 1 file changed, 20 insertions(+), 15 deletions(-)

diff --git a/libavcodec/ralf.c b/libavcodec/ralf.c
index 6866f79609..3244424f74 100644
--- a/libavcodec/ralf.c
+++ b/libavcodec/ralf.c
@@ -149,7 +149,7 @@ static av_cold int decode_init(AVCodecContext *avctx)
                avctx->sample_rate, avctx->channels);
         return AVERROR_INVALIDDATA;
     }
-    avctx->sample_fmt     = AV_SAMPLE_FMT_S16;
+    avctx->sample_fmt     = AV_SAMPLE_FMT_S16P;
     avctx->channel_layout = (avctx->channels == 2) ? AV_CH_LAYOUT_STEREO
                                                    : AV_CH_LAYOUT_MONO;
 
@@ -338,7 +338,8 @@ static void apply_lpc(RALFContext *ctx, int ch, int length, int bits)
     }
 }
 
-static int decode_block(AVCodecContext *avctx, GetBitContext *gb, int16_t *dst)
+static int decode_block(AVCodecContext *avctx, GetBitContext *gb,
+                        int16_t *dst0, int16_t *dst1)
 {
     RALFContext *ctx = avctx->priv_data;
     int len, ch, ret;
@@ -382,35 +383,35 @@ static int decode_block(AVCodecContext *avctx, GetBitContext *gb, int16_t *dst)
     switch (dmode) {
     case 0:
         for (i = 0; i < len; i++)
-            *dst++ = ch0[i] + ctx->bias[0];
+            dst0[i] = ch0[i] + ctx->bias[0];
         break;
     case 1:
         for (i = 0; i < len; i++) {
-            *dst++ = ch0[i] + ctx->bias[0];
-            *dst++ = ch1[i] + ctx->bias[1];
+            dst0[i] = ch0[i] + ctx->bias[0];
+            dst1[i] = ch1[i] + ctx->bias[1];
         }
         break;
     case 2:
         for (i = 0; i < len; i++) {
             ch0[i] += ctx->bias[0];
-            *dst++ = ch0[i];
-            *dst++ = ch0[i] - (ch1[i] + ctx->bias[1]);
+            dst0[i] = ch0[i];
+            dst1[i] = ch0[i] - (ch1[i] + ctx->bias[1]);
         }
         break;
     case 3:
         for (i = 0; i < len; i++) {
             t  = ch0[i] + ctx->bias[0];
             t2 = ch1[i] + ctx->bias[1];
-            *dst++ = t + t2;
-            *dst++ = t;
+            dst0[i] = t + t2;
+            dst1[i] = t;
         }
         break;
     case 4:
         for (i = 0; i < len; i++) {
             t  =   ch1[i] + ctx->bias[1];
             t2 = ((ch0[i] + ctx->bias[0]) << 1) | (t & 1);
-            *dst++ = (t2 + t) / 2;
-            *dst++ = (t2 - t) / 2;
+            dst0[i] = (t2 + t) / 2;
+            dst1[i] = (t2 - t) / 2;
         }
         break;
     }
@@ -424,7 +425,8 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr,
                         AVPacket *avpkt)
 {
     RALFContext *ctx = avctx->priv_data;
-    int16_t *samples;
+    int16_t *samples0;
+    int16_t *samples1;
     int ret;
     GetBitContext gb;
     int table_size, table_bytes, i;
@@ -465,7 +467,8 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr,
         av_log(avctx, AV_LOG_ERROR, "Me fail get_buffer()? That's unpossible!\n");
         return ret;
     }
-    samples = (int16_t*)ctx->frame.data[0];
+    samples0 = (int16_t *)ctx->frame.data[0];
+    samples1 = (int16_t *)ctx->frame.data[1];
 
     if (src_size < 5) {
         av_log(avctx, AV_LOG_ERROR, "too short packets are too short!\n");
@@ -498,8 +501,8 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr,
             break;
         }
         init_get_bits(&gb, block_pointer, ctx->block_size[i] * 8);
-        if (decode_block(avctx, &gb, samples + ctx->sample_offset
-                                               * avctx->channels) < 0) {
+        if (decode_block(avctx, &gb, samples0 + ctx->sample_offset,
+                                     samples1 + ctx->sample_offset) < 0) {
             av_log(avctx, AV_LOG_ERROR, "Sir, I got carsick in your office. Not decoding the rest of packet.\n");
             break;
         }
@@ -533,4 +536,6 @@ AVCodec ff_ralf_decoder = {
     .flush          = decode_flush,
     .capabilities   = CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("RealAudio Lossless"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16P,
+                                                      AV_SAMPLE_FMT_NONE },
 };

From 1478a3601e450f2044103e6e3449105957b25eac Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Tue, 28 Aug 2012 15:29:40 -0400
Subject: [PATCH 09/10] twinvq: use planar sample format

---
 libavcodec/twinvq.c | 33 +++++++++++++--------------------
 1 file changed, 13 insertions(+), 20 deletions(-)

diff --git a/libavcodec/twinvq.c b/libavcodec/twinvq.c
index ab22f64734..d009196445 100644
--- a/libavcodec/twinvq.c
+++ b/libavcodec/twinvq.c
@@ -666,7 +666,7 @@ static void imdct_and_window(TwinContext *tctx, enum FrameType ftype, int wtype,
 }
 
 static void imdct_output(TwinContext *tctx, enum FrameType ftype, int wtype,
-                         float *out)
+                         float **out)
 {
     const ModeTab *mtab = tctx->mtab;
     int size1, size2;
@@ -685,24 +685,15 @@ static void imdct_output(TwinContext *tctx, enum FrameType ftype, int wtype,
 
     size2 = tctx->last_block_pos[0];
     size1 = mtab->size - size2;
+
+    memcpy(&out[0][0    ], prev_buf,         size1 * sizeof(out[0][0]));
+    memcpy(&out[0][size1], tctx->curr_frame, size2 * sizeof(out[0][0]));
+
     if (tctx->avctx->channels == 2) {
-        tctx->dsp.butterflies_float_interleave(out, prev_buf,
-                                               &prev_buf[2*mtab->size],
-                                               size1);
-
-        out += 2 * size1;
-
-        tctx->dsp.butterflies_float_interleave(out, tctx->curr_frame,
-                                               &tctx->curr_frame[2*mtab->size],
-                                               size2);
-    } else {
-        memcpy(out, prev_buf, size1 * sizeof(*out));
-
-        out += size1;
-
-        memcpy(out, tctx->curr_frame, size2 * sizeof(*out));
+        memcpy(&out[1][0],     &prev_buf[2*mtab->size],         size1 * sizeof(out[1][0]));
+        memcpy(&out[1][size1], &tctx->curr_frame[2*mtab->size], size2 * sizeof(out[1][0]));
+        tctx->dsp.butterflies_float(out[0], out[1], mtab->size);
     }
-
 }
 
 static void dec_bark_env(TwinContext *tctx, const uint8_t *in, int use_hist,
@@ -825,7 +816,7 @@ static int twin_decode_frame(AVCodecContext * avctx, void *data,
     TwinContext *tctx = avctx->priv_data;
     GetBitContext gb;
     const ModeTab *mtab = tctx->mtab;
-    float *out = NULL;
+    float **out = NULL;
     enum FrameType ftype;
     int window_type, ret;
     static const enum FrameType wtype_to_ftype_table[] = {
@@ -846,7 +837,7 @@ static int twin_decode_frame(AVCodecContext * avctx, void *data,
             av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
             return ret;
         }
-        out = (float *)tctx->frame.data[0];
+        out = (float **)tctx->frame.extended_data;
     }
 
     init_get_bits(&gb, buf, buf_size * 8);
@@ -1119,7 +1110,7 @@ static av_cold int twin_decode_init(AVCodecContext *avctx)
     int isampf, ibps;
 
     tctx->avctx       = avctx;
-    avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
+    avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
 
     if (!avctx->extradata || avctx->extradata_size < 12) {
         av_log(avctx, AV_LOG_ERROR, "Missing or incomplete extradata\n");
@@ -1184,4 +1175,6 @@ AVCodec ff_twinvq_decoder = {
     .decode         = twin_decode_frame,
     .capabilities   = CODEC_CAP_DR1,
     .long_name      = NULL_IF_CONFIG_SMALL("VQF TwinVQ"),
+    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
+                                                      AV_SAMPLE_FMT_NONE },
 };

From bfcd4b6a1691d20aebc6d2308424c2a88334a9f0 Mon Sep 17 00:00:00 2001
From: Justin Ruggles <justin.ruggles@gmail.com>
Date: Fri, 21 Sep 2012 10:33:56 -0400
Subject: [PATCH 10/10] adpcmdec: set AVCodec.sample_fmts

---
 libavcodec/adpcm.c | 60 ++++++++++++++++++++++++----------------------
 1 file changed, 32 insertions(+), 28 deletions(-)

diff --git a/libavcodec/adpcm.c b/libavcodec/adpcm.c
index a48cee2f9b..f11b899dd3 100644
--- a/libavcodec/adpcm.c
+++ b/libavcodec/adpcm.c
@@ -1265,7 +1265,10 @@ static int adpcm_decode_frame(AVCodecContext *avctx, void *data,
 }
 
 
-#define ADPCM_DECODER(id_, name_, long_name_)               \
+static const enum AVSampleFormat sample_fmts_s16[]  = { AV_SAMPLE_FMT_S16,
+                                                        AV_SAMPLE_FMT_NONE };
+
+#define ADPCM_DECODER(id_, sample_fmts_, name_, long_name_) \
 AVCodec ff_ ## name_ ## _decoder = {                        \
     .name           = #name_,                               \
     .type           = AVMEDIA_TYPE_AUDIO,                   \
@@ -1275,33 +1278,34 @@ AVCodec ff_ ## name_ ## _decoder = {                        \
     .decode         = adpcm_decode_frame,                   \
     .capabilities   = CODEC_CAP_DR1,                        \
     .long_name      = NULL_IF_CONFIG_SMALL(long_name_),     \
+    .sample_fmts    = sample_fmts_,                         \
 }
 
 /* Note: Do not forget to add new entries to the Makefile as well. */
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_4XM, adpcm_4xm, "ADPCM 4X Movie");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_CT, adpcm_ct, "ADPCM Creative Technology");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA, adpcm_ea, "ADPCM Electronic Arts");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_MAXIS_XA, adpcm_ea_maxis_xa, "ADPCM Electronic Arts Maxis CDROM XA");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_R1, adpcm_ea_r1, "ADPCM Electronic Arts R1");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_R2, adpcm_ea_r2, "ADPCM Electronic Arts R2");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_R3, adpcm_ea_r3, "ADPCM Electronic Arts R3");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_XAS, adpcm_ea_xas, "ADPCM Electronic Arts XAS");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_AMV, adpcm_ima_amv, "ADPCM IMA AMV");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_APC, adpcm_ima_apc, "ADPCM IMA CRYO APC");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_DK3, adpcm_ima_dk3, "ADPCM IMA Duck DK3");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_DK4, adpcm_ima_dk4, "ADPCM IMA Duck DK4");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_EA_EACS, adpcm_ima_ea_eacs, "ADPCM IMA Electronic Arts EACS");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_EA_SEAD, adpcm_ima_ea_sead, "ADPCM IMA Electronic Arts SEAD");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_ISS, adpcm_ima_iss, "ADPCM IMA Funcom ISS");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_QT, adpcm_ima_qt, "ADPCM IMA QuickTime");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_SMJPEG, adpcm_ima_smjpeg, "ADPCM IMA Loki SDL MJPEG");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_WAV, adpcm_ima_wav, "ADPCM IMA WAV");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_WS, adpcm_ima_ws, "ADPCM IMA Westwood");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_MS, adpcm_ms, "ADPCM Microsoft");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_SBPRO_2, adpcm_sbpro_2, "ADPCM Sound Blaster Pro 2-bit");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_SBPRO_3, adpcm_sbpro_3, "ADPCM Sound Blaster Pro 2.6-bit");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_SBPRO_4, adpcm_sbpro_4, "ADPCM Sound Blaster Pro 4-bit");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_SWF, adpcm_swf, "ADPCM Shockwave Flash");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_THP, adpcm_thp, "ADPCM Nintendo Gamecube THP");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_XA, adpcm_xa, "ADPCM CDROM XA");
-ADPCM_DECODER(AV_CODEC_ID_ADPCM_YAMAHA, adpcm_yamaha, "ADPCM Yamaha");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_4XM,         sample_fmts_s16,  adpcm_4xm,         "ADPCM 4X Movie");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_CT,          sample_fmts_s16,  adpcm_ct,          "ADPCM Creative Technology");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA,          sample_fmts_s16,  adpcm_ea,          "ADPCM Electronic Arts");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_MAXIS_XA, sample_fmts_s16,  adpcm_ea_maxis_xa, "ADPCM Electronic Arts Maxis CDROM XA");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_R1,       sample_fmts_s16,  adpcm_ea_r1,       "ADPCM Electronic Arts R1");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_R2,       sample_fmts_s16,  adpcm_ea_r2,       "ADPCM Electronic Arts R2");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_R3,       sample_fmts_s16,  adpcm_ea_r3,       "ADPCM Electronic Arts R3");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_EA_XAS,      sample_fmts_s16,  adpcm_ea_xas,      "ADPCM Electronic Arts XAS");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_AMV,     sample_fmts_s16,  adpcm_ima_amv,     "ADPCM IMA AMV");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_APC,     sample_fmts_s16,  adpcm_ima_apc,     "ADPCM IMA CRYO APC");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_DK3,     sample_fmts_s16,  adpcm_ima_dk3,     "ADPCM IMA Duck DK3");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_DK4,     sample_fmts_s16,  adpcm_ima_dk4,     "ADPCM IMA Duck DK4");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_EA_EACS, sample_fmts_s16,  adpcm_ima_ea_eacs, "ADPCM IMA Electronic Arts EACS");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_EA_SEAD, sample_fmts_s16,  adpcm_ima_ea_sead, "ADPCM IMA Electronic Arts SEAD");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_ISS,     sample_fmts_s16,  adpcm_ima_iss,     "ADPCM IMA Funcom ISS");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_QT,      sample_fmts_s16,  adpcm_ima_qt,      "ADPCM IMA QuickTime");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_SMJPEG,  sample_fmts_s16,  adpcm_ima_smjpeg,  "ADPCM IMA Loki SDL MJPEG");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_WAV,     sample_fmts_s16,  adpcm_ima_wav,     "ADPCM IMA WAV");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_IMA_WS,      sample_fmts_s16,  adpcm_ima_ws,      "ADPCM IMA Westwood");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_MS,          sample_fmts_s16,  adpcm_ms,          "ADPCM Microsoft");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_SBPRO_2,     sample_fmts_s16,  adpcm_sbpro_2,     "ADPCM Sound Blaster Pro 2-bit");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_SBPRO_3,     sample_fmts_s16,  adpcm_sbpro_3,     "ADPCM Sound Blaster Pro 2.6-bit");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_SBPRO_4,     sample_fmts_s16,  adpcm_sbpro_4,     "ADPCM Sound Blaster Pro 4-bit");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_SWF,         sample_fmts_s16,  adpcm_swf,         "ADPCM Shockwave Flash");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_THP,         sample_fmts_s16,  adpcm_thp,         "ADPCM Nintendo Gamecube THP");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_XA,          sample_fmts_s16,  adpcm_xa,          "ADPCM CDROM XA");
+ADPCM_DECODER(AV_CODEC_ID_ADPCM_YAMAHA,      sample_fmts_s16,  adpcm_yamaha,      "ADPCM Yamaha");