Windows Media Audio Lossless decoder
authorMashiat Sarker Shakkhar <shahriman_ams@yahoo.com>
Thu, 1 Mar 2012 12:43:00 +0000 (12:43 +0000)
committerDiego Biurrun <diego@biurrun.de>
Fri, 2 Mar 2012 18:10:29 +0000 (19:10 +0100)
Decodes 16-bit WMA Lossless encoded files. 24-bit is not supported yet.

Bitstream parser written by Andreas Öman with contributions from
Baptiste Coudurier and Ulion.

Includes a number of bug-fixes from Benjamin Larsson, Michael Niedermayer and
Konstantin Shishkov, shine and polish by Diego Biurrun.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
Changelog
doc/general.texi
libavcodec/Makefile
libavcodec/allcodecs.c
libavcodec/version.h
libavcodec/wmalosslessdec.c [new file with mode: 0644]

index 9dc2ae9..8703955 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -11,6 +11,7 @@ version <next>:
 - Sun Rasterfile Encoder
 - remove libpostproc
 - ID3v2 attached pictures reading and writing
+- WMA Lossless decoder
 
 
 version 0.8:
index c61951a..9f31081 100644 (file)
@@ -768,6 +768,7 @@ following image formats are supported:
 @item Westwood Audio (SND1)  @tab     @tab  X
 @item Windows Media Audio 1  @tab  X  @tab  X
 @item Windows Media Audio 2  @tab  X  @tab  X
+@item Windows Media Audio Lossless @tab  @tab  X
 @item Windows Media Audio Pro @tab    @tab  X
 @item Windows Media Audio Voice @tab  @tab  X
 @end multitable
index 53e67dc..33a1974 100644 (file)
@@ -418,6 +418,7 @@ OBJS-$(CONFIG_VP6_DECODER)             += vp6.o vp56.o vp56data.o vp56dsp.o \
 OBJS-$(CONFIG_VP8_DECODER)             += vp8.o vp8dsp.o vp56rac.o
 OBJS-$(CONFIG_VQA_DECODER)             += vqavideo.o
 OBJS-$(CONFIG_WAVPACK_DECODER)         += wavpack.o
+OBJS-$(CONFIG_WMALOSSLESS_DECODER)     += wmalosslessdec.o wma.o
 OBJS-$(CONFIG_WMAPRO_DECODER)          += wmaprodec.o wma.o
 OBJS-$(CONFIG_WMAV1_DECODER)           += wmadec.o wma.o aactab.o
 OBJS-$(CONFIG_WMAV1_ENCODER)           += wmaenc.o wma.o aactab.o
index 3d3289c..2844cfc 100644 (file)
@@ -289,6 +289,7 @@ void avcodec_register_all(void)
     REGISTER_DECODER (VMDAUDIO, vmdaudio);
     REGISTER_ENCDEC  (VORBIS, vorbis);
     REGISTER_DECODER (WAVPACK, wavpack);
+    REGISTER_DECODER (WMALOSSLESS, wmalossless);
     REGISTER_DECODER (WMAPRO, wmapro);
     REGISTER_ENCDEC  (WMAV1, wmav1);
     REGISTER_ENCDEC  (WMAV2, wmav2);
index 49f929f..7790f24 100644 (file)
@@ -21,7 +21,7 @@
 #define AVCODEC_VERSION_H
 
 #define LIBAVCODEC_VERSION_MAJOR 54
-#define LIBAVCODEC_VERSION_MINOR  4
+#define LIBAVCODEC_VERSION_MINOR  5
 #define LIBAVCODEC_VERSION_MICRO  0
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
diff --git a/libavcodec/wmalosslessdec.c b/libavcodec/wmalosslessdec.c
new file mode 100644 (file)
index 0000000..d311c39
--- /dev/null
@@ -0,0 +1,1248 @@
+/*
+ * Windows Media Audio Lossless decoder
+ * Copyright (c) 2007 Baptiste Coudurier, Benjamin Larsson, Ulion
+ * Copyright (c) 2008 - 2011 Sascha Sommer, Benjamin Larsson
+ * Copyright (c) 2011 Andreas Öman
+ * Copyright (c) 2011 - 2012 Mashiat Sarker Shakkhar
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "avcodec.h"
+#include "internal.h"
+#include "get_bits.h"
+#include "put_bits.h"
+#include "wma.h"
+
+/** current decoder limitations */
+#define WMALL_MAX_CHANNELS      8                       ///< max number of handled channels
+#define MAX_SUBFRAMES          32                       ///< max number of subframes per channel
+#define MAX_BANDS              29                       ///< max number of scale factor bands
+#define MAX_FRAMESIZE       32768                       ///< maximum compressed frame size
+
+#define WMALL_BLOCK_MIN_BITS    6                       ///< log2 of min block size
+#define WMALL_BLOCK_MAX_BITS   12                       ///< log2 of max block size
+#define WMALL_BLOCK_MAX_SIZE (1 << WMALL_BLOCK_MAX_BITS)    ///< maximum block size
+#define WMALL_BLOCK_SIZES    (WMALL_BLOCK_MAX_BITS - WMALL_BLOCK_MIN_BITS + 1) ///< possible block sizes
+
+
+/**
+ * @brief frame-specific decoder context for a single channel
+ */
+typedef struct {
+    int16_t     prev_block_len;                         ///< length of the previous block
+    uint8_t     transmit_coefs;
+    uint8_t     num_subframes;
+    uint16_t    subframe_len[MAX_SUBFRAMES];            ///< subframe length in samples
+    uint16_t    subframe_offsets[MAX_SUBFRAMES];        ///< subframe positions in the current frame
+    uint8_t     cur_subframe;                           ///< current subframe number
+    uint16_t    decoded_samples;                        ///< number of already processed samples
+    int         quant_step;                             ///< quantization step for the current subframe
+    int         transient_counter;                      ///< number of transient samples from the beginning of the transient zone
+} WmallChannelCtx;
+
+/**
+ * @brief main decoder context
+ */
+typedef struct WmallDecodeCtx {
+    /* generic decoder variables */
+    AVCodecContext  *avctx;
+    AVFrame         frame;
+    uint8_t         frame_data[MAX_FRAMESIZE + FF_INPUT_BUFFER_PADDING_SIZE];  ///< compressed frame data
+    PutBitContext   pb;                             ///< context for filling the frame_data buffer
+
+    /* frame size dependent frame information (set during initialization) */
+    uint32_t        decode_flags;                   ///< used compression features
+    int             len_prefix;                     ///< frame is prefixed with its length
+    int             dynamic_range_compression;      ///< frame contains DRC data
+    uint8_t         bits_per_sample;                ///< integer audio sample size for the unscaled IMDCT output (used to scale to [-1.0, 1.0])
+    uint16_t        samples_per_frame;              ///< number of samples to output
+    uint16_t        log2_frame_size;
+    int8_t          num_channels;                   ///< number of channels in the stream (same as AVCodecContext.num_channels)
+    int8_t          lfe_channel;                    ///< lfe channel index
+    uint8_t         max_num_subframes;
+    uint8_t         subframe_len_bits;              ///< number of bits used for the subframe length
+    uint8_t         max_subframe_len_bit;           ///< flag indicating that the subframe is of maximum size when the first subframe length bit is 1
+    uint16_t        min_samples_per_subframe;
+
+    /* packet decode state */
+    GetBitContext   pgb;                            ///< bitstream reader context for the packet
+    int             next_packet_start;              ///< start offset of the next WMA packet in the demuxer packet
+    uint8_t         packet_offset;                  ///< offset to the frame in the packet
+    uint8_t         packet_sequence_number;         ///< current packet number
+    int             num_saved_bits;                 ///< saved number of bits
+    int             frame_offset;                   ///< frame offset in the bit reservoir
+    int             subframe_offset;                ///< subframe offset in the bit reservoir
+    uint8_t         packet_loss;                    ///< set in case of bitstream error
+    uint8_t         packet_done;                    ///< set when a packet is fully decoded
+
+    /* frame decode state */
+    uint32_t        frame_num;                      ///< current frame number (not used for decoding)
+    GetBitContext   gb;                             ///< bitstream reader context
+    int             buf_bit_size;                   ///< buffer size in bits
+    int16_t         *samples_16;                    ///< current samplebuffer pointer (16-bit)
+    int16_t         *samples_16_end;                ///< maximum samplebuffer pointer
+    int             *samples_32;                    ///< current samplebuffer pointer (24-bit)
+    int             *samples_32_end;                ///< maximum samplebuffer pointer
+    uint8_t         drc_gain;                       ///< gain for the DRC tool
+    int8_t          skip_frame;                     ///< skip output step
+    int8_t          parsed_all_subframes;           ///< all subframes decoded?
+
+    /* subframe/block decode state */
+    int16_t         subframe_len;                   ///< current subframe length
+    int8_t          channels_for_cur_subframe;      ///< number of channels that contain the subframe
+    int8_t          channel_indexes_for_cur_subframe[WMALL_MAX_CHANNELS];
+
+    WmallChannelCtx channel[WMALL_MAX_CHANNELS];    ///< per channel data
+
+    // WMA Lossless-specific
+
+    uint8_t do_arith_coding;
+    uint8_t do_ac_filter;
+    uint8_t do_inter_ch_decorr;
+    uint8_t do_mclms;
+    uint8_t do_lpc;
+
+    int8_t  acfilter_order;
+    int8_t  acfilter_scaling;
+    int64_t acfilter_coeffs[16];
+    int     acfilter_prevvalues[2][16];
+
+    int8_t  mclms_order;
+    int8_t  mclms_scaling;
+    int16_t mclms_coeffs[128];
+    int16_t mclms_coeffs_cur[4];
+    int16_t mclms_prevvalues[64];
+    int16_t mclms_updates[64];
+    int     mclms_recent;
+
+    int     movave_scaling;
+    int     quant_stepsize;
+
+    struct {
+        int order;
+        int scaling;
+        int coefsend;
+        int bitsend;
+        int16_t coefs[256];
+        int16_t lms_prevvalues[512];
+        int16_t lms_updates[512];
+        int recent;
+    } cdlms[2][9];
+
+    int cdlms_ttl[2];
+
+    int bV3RTM;
+
+    int is_channel_coded[2];
+    int update_speed[2];
+
+    int transient[2];
+    int transient_pos[2];
+    int seekable_tile;
+
+    int ave_sum[2];
+
+    int channel_residues[2][2048];
+
+    int lpc_coefs[2][40];
+    int lpc_order;
+    int lpc_scaling;
+    int lpc_intbits;
+
+    int channel_coeffs[2][2048];
+} WmallDecodeCtx;
+
+
+static av_cold int decode_init(AVCodecContext *avctx)
+{
+    WmallDecodeCtx *s  = avctx->priv_data;
+    uint8_t *edata_ptr = avctx->extradata;
+    unsigned int channel_mask;
+    int i, log2_max_num_subframes, num_possible_block_sizes;
+
+    s->avctx = avctx;
+    init_put_bits(&s->pb, s->frame_data, MAX_FRAMESIZE);
+
+    if (avctx->extradata_size >= 18) {
+        s->decode_flags    = AV_RL16(edata_ptr + 14);
+        channel_mask       = AV_RL32(edata_ptr +  2);
+        s->bits_per_sample = AV_RL16(edata_ptr);
+        if (s->bits_per_sample == 16)
+            avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+        else if (s->bits_per_sample == 24) {
+            avctx->sample_fmt = AV_SAMPLE_FMT_S32;
+            av_log_missing_feature(avctx, "bit-depth higher than 16", 0);
+            return AVERROR_PATCHWELCOME;
+        } else {
+            av_log(avctx, AV_LOG_ERROR, "Unknown bit-depth: %d\n",
+                   s->bits_per_sample);
+            return AVERROR_INVALIDDATA;
+        }
+        /* dump the extradata */
+        for (i = 0; i < avctx->extradata_size; i++)
+            av_dlog(avctx, AV_LOG_DEBUG, "[%x] ", avctx->extradata[i]);
+        av_dlog(avctx, AV_LOG_DEBUG, "\n");
+
+    } else {
+        av_log_ask_for_sample(avctx, "Unsupported extradata size\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    /* generic init */
+    s->log2_frame_size = av_log2(avctx->block_align) + 4;
+
+    /* frame info */
+    s->skip_frame  = 1; /* skip first frame */
+    s->packet_loss = 1;
+    s->len_prefix  = s->decode_flags & 0x40;
+
+    /* get frame len */
+    s->samples_per_frame = 1 << ff_wma_get_frame_len_bits(avctx->sample_rate,
+                                                          3, s->decode_flags);
+
+    /* init previous block len */
+    for (i = 0; i < avctx->channels; i++)
+        s->channel[i].prev_block_len = s->samples_per_frame;
+
+    /* subframe info */
+    log2_max_num_subframes  = (s->decode_flags & 0x38) >> 3;
+    s->max_num_subframes    = 1 << log2_max_num_subframes;
+    s->max_subframe_len_bit = 0;
+    s->subframe_len_bits    = av_log2(log2_max_num_subframes) + 1;
+
+    num_possible_block_sizes     = log2_max_num_subframes + 1;
+    s->min_samples_per_subframe  = s->samples_per_frame / s->max_num_subframes;
+    s->dynamic_range_compression = s->decode_flags & 0x80;
+    s->bV3RTM                    = s->decode_flags & 0x100;
+
+    if (s->max_num_subframes > MAX_SUBFRAMES) {
+        av_log(avctx, AV_LOG_ERROR, "invalid number of subframes %i\n",
+               s->max_num_subframes);
+        return AVERROR_INVALIDDATA;
+    }
+
+    s->num_channels = avctx->channels;
+
+    /* extract lfe channel position */
+    s->lfe_channel = -1;
+
+    if (channel_mask & 8) {
+        unsigned int mask;
+        for (mask = 1; mask < 16; mask <<= 1)
+            if (channel_mask & mask)
+                ++s->lfe_channel;
+    }
+
+    if (s->num_channels < 0) {
+        av_log(avctx, AV_LOG_ERROR, "invalid number of channels %d\n",
+               s->num_channels);
+        return AVERROR_INVALIDDATA;
+    } else if (s->num_channels > WMALL_MAX_CHANNELS) {
+        av_log_ask_for_sample(avctx, "unsupported number of channels\n");
+        return AVERROR_PATCHWELCOME;
+    }
+
+    avcodec_get_frame_defaults(&s->frame);
+    avctx->coded_frame    = &s->frame;
+    avctx->channel_layout = channel_mask;
+    return 0;
+}
+
+/**
+ * @brief Decode the subframe length.
+ * @param s      context
+ * @param offset sample offset in the frame
+ * @return decoded subframe length on success, < 0 in case of an error
+ */
+static int decode_subframe_length(WmallDecodeCtx *s, int offset)
+{
+    int frame_len_ratio, subframe_len, len;
+
+    /* no need to read from the bitstream when only one length is possible */
+    if (offset == s->samples_per_frame - s->min_samples_per_subframe)
+        return s->min_samples_per_subframe;
+
+    len             = av_log2(s->max_num_subframes - 1) + 1;
+    frame_len_ratio = get_bits(&s->gb, len);
+    subframe_len    = s->min_samples_per_subframe * (frame_len_ratio + 1);
+
+    /* sanity check the length */
+    if (subframe_len < s->min_samples_per_subframe ||
+        subframe_len > s->samples_per_frame) {
+        av_log(s->avctx, AV_LOG_ERROR, "broken frame: subframe_len %i\n",
+               subframe_len);
+        return AVERROR_INVALIDDATA;
+    }
+    return subframe_len;
+}
+
+/**
+ * @brief Decode how the data in the frame is split into subframes.
+ *       Every WMA frame contains the encoded data for a fixed number of
+ *       samples per channel. The data for every channel might be split
+ *       into several subframes. This function will reconstruct the list of
+ *       subframes for every channel.
+ *
+ *       If the subframes are not evenly split, the algorithm estimates the
+ *       channels with the lowest number of total samples.
+ *       Afterwards, for each of these channels a bit is read from the
+ *       bitstream that indicates if the channel contains a subframe with the
+ *       next subframe size that is going to be read from the bitstream or not.
+ *       If a channel contains such a subframe, the subframe size gets added to
+ *       the channel's subframe list.
+ *       The algorithm repeats these steps until the frame is properly divided
+ *       between the individual channels.
+ *
+ * @param s context
+ * @return 0 on success, < 0 in case of an error
+ */
+static int decode_tilehdr(WmallDecodeCtx *s)
+{
+    uint16_t num_samples[WMALL_MAX_CHANNELS] = { 0 }; /* sum of samples for all currently known subframes of a channel */
+    uint8_t  contains_subframe[WMALL_MAX_CHANNELS];   /* flag indicating if a channel contains the current subframe */
+    int channels_for_cur_subframe = s->num_channels;  /* number of channels that contain the current subframe */
+    int fixed_channel_layout = 0;                     /* flag indicating that all channels use the same subfra2me offsets and sizes */
+    int min_channel_len = 0;                          /* smallest sum of samples (channels with this length will be processed first) */
+    int c, tile_aligned;
+
+    /* reset tiling information */
+    for (c = 0; c < s->num_channels; c++)
+        s->channel[c].num_subframes = 0;
+
+    tile_aligned = get_bits1(&s->gb);
+    if (s->max_num_subframes == 1 || tile_aligned)
+        fixed_channel_layout = 1;
+
+    /* loop until the frame data is split between the subframes */
+    do {
+        int subframe_len;
+
+        /* check which channels contain the subframe */
+        for (c = 0; c < s->num_channels; c++) {
+            if (num_samples[c] == min_channel_len) {
+                if (fixed_channel_layout || channels_for_cur_subframe == 1 ||
+                   (min_channel_len == s->samples_per_frame - s->min_samples_per_subframe)) {
+                    contains_subframe[c] = 1;
+                } else {
+                    contains_subframe[c] = get_bits1(&s->gb);
+                }
+            } else
+                contains_subframe[c] = 0;
+        }
+
+        /* get subframe length, subframe_len == 0 is not allowed */
+        if ((subframe_len = decode_subframe_length(s, min_channel_len)) <= 0)
+            return AVERROR_INVALIDDATA;
+        /* add subframes to the individual channels and find new min_channel_len */
+        min_channel_len += subframe_len;
+        for (c = 0; c < s->num_channels; c++) {
+            WmallChannelCtx *chan = &s->channel[c];
+
+            if (contains_subframe[c]) {
+                if (chan->num_subframes >= MAX_SUBFRAMES) {
+                    av_log(s->avctx, AV_LOG_ERROR,
+                           "broken frame: num subframes > 31\n");
+                    return AVERROR_INVALIDDATA;
+                }
+                chan->subframe_len[chan->num_subframes] = subframe_len;
+                num_samples[c] += subframe_len;
+                ++chan->num_subframes;
+                if (num_samples[c] > s->samples_per_frame) {
+                    av_log(s->avctx, AV_LOG_ERROR, "broken frame: "
+                           "channel len(%d) > samples_per_frame(%d)\n",
+                           num_samples[c], s->samples_per_frame);
+                    return AVERROR_INVALIDDATA;
+                }
+            } else if (num_samples[c] <= min_channel_len) {
+                if (num_samples[c] < min_channel_len) {
+                    channels_for_cur_subframe = 0;
+                    min_channel_len = num_samples[c];
+                }
+                ++channels_for_cur_subframe;
+            }
+        }
+    } while (min_channel_len < s->samples_per_frame);
+
+    for (c = 0; c < s->num_channels; c++) {
+        int i, offset = 0;
+        for (i = 0; i < s->channel[c].num_subframes; i++) {
+            s->channel[c].subframe_offsets[i] = offset;
+            offset += s->channel[c].subframe_len[i];
+        }
+    }
+
+    return 0;
+}
+
+static void decode_ac_filter(WmallDecodeCtx *s)
+{
+    int i;
+    s->acfilter_order   = get_bits(&s->gb, 4) + 1;
+    s->acfilter_scaling = get_bits(&s->gb, 4);
+
+    for (i = 0; i < s->acfilter_order; i++)
+        s->acfilter_coeffs[i] = get_bits(&s->gb, s->acfilter_scaling) + 1;
+}
+
+static void decode_mclms(WmallDecodeCtx *s)
+{
+    s->mclms_order   = (get_bits(&s->gb, 4) + 1) * 2;
+    s->mclms_scaling = get_bits(&s->gb, 4);
+    if (get_bits1(&s->gb)) {
+        int i, send_coef_bits;
+        int cbits = av_log2(s->mclms_scaling + 1);
+        assert(cbits == my_log2(s->mclms_scaling + 1));
+        if (1 << cbits < s->mclms_scaling + 1)
+            cbits++;
+
+        send_coef_bits = (cbits ? get_bits(&s->gb, cbits) : 0) + 2;
+
+        for (i = 0; i < s->mclms_order * s->num_channels * s->num_channels; i++)
+            s->mclms_coeffs[i] = get_bits(&s->gb, send_coef_bits);
+
+        for (i = 0; i < s->num_channels; i++) {
+            int c;
+            for (c = 0; c < i; c++)
+                s->mclms_coeffs_cur[i * s->num_channels + c] = get_bits(&s->gb, send_coef_bits);
+        }
+    }
+}
+
+static void decode_cdlms(WmallDecodeCtx *s)
+{
+    int c, i;
+    int cdlms_send_coef = get_bits1(&s->gb);
+
+    for (c = 0; c < s->num_channels; c++) {
+        s->cdlms_ttl[c] = get_bits(&s->gb, 3) + 1;
+        for (i = 0; i < s->cdlms_ttl[c]; i++)
+            s->cdlms[c][i].order = (get_bits(&s->gb, 7) + 1) * 8;
+
+        for (i = 0; i < s->cdlms_ttl[c]; i++)
+            s->cdlms[c][i].scaling = get_bits(&s->gb, 4);
+
+        if (cdlms_send_coef) {
+            for (i = 0; i < s->cdlms_ttl[c]; i++) {
+                int cbits, shift_l, shift_r, j;
+                cbits = av_log2(s->cdlms[c][i].order);
+                if ((1 << cbits) < s->cdlms[c][i].order)
+                    cbits++;
+                s->cdlms[c][i].coefsend = get_bits(&s->gb, cbits) + 1;
+
+                cbits = av_log2(s->cdlms[c][i].scaling + 1);
+                if ((1 << cbits) < s->cdlms[c][i].scaling + 1)
+                    cbits++;
+
+                s->cdlms[c][i].bitsend = get_bits(&s->gb, cbits) + 2;
+                shift_l = 32 - s->cdlms[c][i].bitsend;
+                shift_r = 32 - s->cdlms[c][i].scaling - 2;
+                for (j = 0; j < s->cdlms[c][i].coefsend; j++)
+                    s->cdlms[c][i].coefs[j] =
+                        (get_bits(&s->gb, s->cdlms[c][i].bitsend) << shift_l) >> shift_r;
+            }
+        }
+    }
+}
+
+static int decode_channel_residues(WmallDecodeCtx *s, int ch, int tile_size)
+{
+    int i = 0;
+    unsigned int ave_mean;
+    s->transient[ch] = get_bits1(&s->gb);
+    if (s->transient[ch]) {
+        s->transient_pos[ch] = get_bits(&s->gb, av_log2(tile_size));
+        if (s->transient_pos[ch])
+            s->transient[ch] = 0;
+        s->channel[ch].transient_counter =
+            FFMAX(s->channel[ch].transient_counter, s->samples_per_frame / 2);
+    } else if (s->channel[ch].transient_counter)
+        s->transient[ch] = 1;
+
+    if (s->seekable_tile) {
+        ave_mean = get_bits(&s->gb, s->bits_per_sample);
+        s->ave_sum[ch] = ave_mean << (s->movave_scaling + 1);
+    }
+
+    if (s->seekable_tile) {
+        if (s->do_inter_ch_decorr)
+            s->channel_residues[ch][0] = get_sbits(&s->gb, s->bits_per_sample + 1);
+        else
+            s->channel_residues[ch][0] = get_sbits(&s->gb, s->bits_per_sample);
+        i++;
+    }
+    for (; i < tile_size; i++) {
+        int quo = 0, rem, rem_bits, residue;
+        while(get_bits1(&s->gb)) {
+            quo++;
+            if (get_bits_left(&s->gb) <= 0)
+                return -1;
+        }
+        if (quo >= 32)
+            quo += get_bits_long(&s->gb, get_bits(&s->gb, 5) + 1);
+
+        ave_mean = (s->ave_sum[ch] + (1 << s->movave_scaling)) >> (s->movave_scaling + 1);
+        if (ave_mean <= 1)
+            residue = quo;
+        else {
+            rem_bits = av_ceil_log2(ave_mean);
+            rem      = rem_bits ? get_bits(&s->gb, rem_bits) : 0;
+            residue  = (quo << rem_bits) + rem;
+        }
+
+        s->ave_sum[ch] = residue + s->ave_sum[ch] -
+                         (s->ave_sum[ch] >> s->movave_scaling);
+
+        if (residue & 1)
+            residue = -(residue >> 1) - 1;
+        else
+            residue = residue >> 1;
+        s->channel_residues[ch][i] = residue;
+    }
+
+    return 0;
+
+}
+
+static void decode_lpc(WmallDecodeCtx *s)
+{
+    int ch, i, cbits;
+    s->lpc_order   = get_bits(&s->gb, 5) + 1;
+    s->lpc_scaling = get_bits(&s->gb, 4);
+    s->lpc_intbits = get_bits(&s->gb, 3) + 1;
+    cbits = s->lpc_scaling + s->lpc_intbits;
+    for (ch = 0; ch < s->num_channels; ch++)
+        for (i = 0; i < s->lpc_order; i++)
+            s->lpc_coefs[ch][i] = get_sbits(&s->gb, cbits);
+}
+
+static void clear_codec_buffers(WmallDecodeCtx *s)
+{
+    int ich, ilms;
+
+    memset(s->acfilter_coeffs,     0, sizeof(s->acfilter_coeffs));
+    memset(s->acfilter_prevvalues, 0, sizeof(s->acfilter_prevvalues));
+    memset(s->lpc_coefs,           0, sizeof(s->lpc_coefs));
+
+    memset(s->mclms_coeffs,     0, sizeof(s->mclms_coeffs));
+    memset(s->mclms_coeffs_cur, 0, sizeof(s->mclms_coeffs_cur));
+    memset(s->mclms_prevvalues, 0, sizeof(s->mclms_prevvalues));
+    memset(s->mclms_updates,    0, sizeof(s->mclms_updates));
+
+    for (ich = 0; ich < s->num_channels; ich++) {
+        for (ilms = 0; ilms < s->cdlms_ttl[ich]; ilms++) {
+            memset(s->cdlms[ich][ilms].coefs, 0,
+                   sizeof(s->cdlms[ich][ilms].coefs));
+            memset(s->cdlms[ich][ilms].lms_prevvalues, 0,
+                   sizeof(s->cdlms[ich][ilms].lms_prevvalues));
+            memset(s->cdlms[ich][ilms].lms_updates, 0,
+                   sizeof(s->cdlms[ich][ilms].lms_updates));
+        }
+        s->ave_sum[ich] = 0;
+    }
+}
+
+/**
+ * @brief Reset filter parameters and transient area at new seekable tile.
+ */
+static void reset_codec(WmallDecodeCtx *s)
+{
+    int ich, ilms;
+    s->mclms_recent = s->mclms_order * s->num_channels;
+    for (ich = 0; ich < s->num_channels; ich++) {
+        for (ilms = 0; ilms < s->cdlms_ttl[ich]; ilms++)
+            s->cdlms[ich][ilms].recent = s->cdlms[ich][ilms].order;
+        /* first sample of a seekable subframe is considered as the starting of
+            a transient area which is samples_per_frame samples long */
+        s->channel[ich].transient_counter = s->samples_per_frame;
+        s->transient[ich]     = 1;
+        s->transient_pos[ich] = 0;
+    }
+}
+
+static void mclms_update(WmallDecodeCtx *s, int icoef, int *pred)
+{
+    int i, j, ich, pred_error;
+    int order        = s->mclms_order;
+    int num_channels = s->num_channels;
+    int range        = 1 << (s->bits_per_sample - 1);
+
+    for (ich = 0; ich < num_channels; ich++) {
+        pred_error = s->channel_residues[ich][icoef] - pred[ich];
+        if (pred_error > 0) {
+            for (i = 0; i < order * num_channels; i++)
+                s->mclms_coeffs[i + ich * order * num_channels] +=
+                    s->mclms_updates[s->mclms_recent + i];
+            for (j = 0; j < ich; j++) {
+                if (s->channel_residues[j][icoef] > 0)
+                    s->mclms_coeffs_cur[ich * num_channels + j] += 1;
+                else if (s->channel_residues[j][icoef] < 0)
+                    s->mclms_coeffs_cur[ich * num_channels + j] -= 1;
+            }
+        } else if (pred_error < 0) {
+            for (i = 0; i < order * num_channels; i++)
+                s->mclms_coeffs[i + ich * order * num_channels] -=
+                    s->mclms_updates[s->mclms_recent + i];
+            for (j = 0; j < ich; j++) {
+                if (s->channel_residues[j][icoef] > 0)
+                    s->mclms_coeffs_cur[ich * num_channels + j] -= 1;
+                else if (s->channel_residues[j][icoef] < 0)
+                    s->mclms_coeffs_cur[ich * num_channels + j] += 1;
+            }
+        }
+    }
+
+    for (ich = num_channels - 1; ich >= 0; ich--) {
+        s->mclms_recent--;
+        s->mclms_prevvalues[s->mclms_recent] = s->channel_residues[ich][icoef];
+        if (s->channel_residues[ich][icoef] > range - 1)
+            s->mclms_prevvalues[s->mclms_recent] = range - 1;
+        else if (s->channel_residues[ich][icoef] < -range)
+            s->mclms_prevvalues[s->mclms_recent] = -range;
+
+        s->mclms_updates[s->mclms_recent] = 0;
+        if (s->channel_residues[ich][icoef] > 0)
+            s->mclms_updates[s->mclms_recent] = 1;
+        else if (s->channel_residues[ich][icoef] < 0)
+            s->mclms_updates[s->mclms_recent] = -1;
+    }
+
+    if (s->mclms_recent == 0) {
+        memcpy(&s->mclms_prevvalues[order * num_channels],
+               s->mclms_prevvalues,
+               2 * order * num_channels);
+        memcpy(&s->mclms_updates[order * num_channels],
+               s->mclms_updates,
+               2 * order * num_channels);
+        s->mclms_recent = num_channels * order;
+    }
+}
+
+static void mclms_predict(WmallDecodeCtx *s, int icoef, int *pred)
+{
+    int ich, i;
+    int order        = s->mclms_order;
+    int num_channels = s->num_channels;
+
+    for (ich = 0; ich < num_channels; ich++) {
+        if (!s->is_channel_coded[ich])
+            continue;
+        pred[ich] = 0;
+        for (i = 0; i < order * num_channels; i++)
+            pred[ich] += s->mclms_prevvalues[i + s->mclms_recent] *
+                         s->mclms_coeffs[i + order * num_channels * ich];
+        for (i = 0; i < ich; i++)
+            pred[ich] += s->channel_residues[i][icoef] *
+                         s->mclms_coeffs_cur[i + num_channels * ich];
+        pred[ich] += 1 << s->mclms_scaling - 1;
+        pred[ich] >>= s->mclms_scaling;
+        s->channel_residues[ich][icoef] += pred[ich];
+    }
+}
+
+static void revert_mclms(WmallDecodeCtx *s, int tile_size)
+{
+    int icoef, pred[WMALL_MAX_CHANNELS] = { 0 };
+    for (icoef = 0; icoef < tile_size; icoef++) {
+        mclms_predict(s, icoef, pred);
+        mclms_update(s, icoef, pred);
+    }
+}
+
+static int lms_predict(WmallDecodeCtx *s, int ich, int ilms)
+{
+    int pred = 0, icoef;
+    int recent = s->cdlms[ich][ilms].recent;
+
+    for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+        pred += s->cdlms[ich][ilms].coefs[icoef] *
+                s->cdlms[ich][ilms].lms_prevvalues[icoef + recent];
+
+    return pred;
+}
+
+static void lms_update(WmallDecodeCtx *s, int ich, int ilms,
+                       int input, int residue)
+{
+    int icoef;
+    int recent = s->cdlms[ich][ilms].recent;
+    int range  = 1 << s->bits_per_sample - 1;
+
+    if (residue < 0) {
+        for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+            s->cdlms[ich][ilms].coefs[icoef] -=
+                s->cdlms[ich][ilms].lms_updates[icoef + recent];
+    } else if (residue > 0) {
+        for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+            s->cdlms[ich][ilms].coefs[icoef] +=
+                s->cdlms[ich][ilms].lms_updates[icoef + recent];
+    }
+
+    if (recent)
+        recent--;
+    else {
+        memcpy(&s->cdlms[ich][ilms].lms_prevvalues[s->cdlms[ich][ilms].order],
+               s->cdlms[ich][ilms].lms_prevvalues,
+               2 * s->cdlms[ich][ilms].order);
+        memcpy(&s->cdlms[ich][ilms].lms_updates[s->cdlms[ich][ilms].order],
+               s->cdlms[ich][ilms].lms_updates,
+               2 * s->cdlms[ich][ilms].order);
+        recent = s->cdlms[ich][ilms].order - 1;
+    }
+
+    s->cdlms[ich][ilms].lms_prevvalues[recent] = av_clip(input, -range, range - 1);
+    if (!input)
+        s->cdlms[ich][ilms].lms_updates[recent] = 0;
+    else if (input < 0)
+        s->cdlms[ich][ilms].lms_updates[recent] = -s->update_speed[ich];
+    else
+        s->cdlms[ich][ilms].lms_updates[recent] = s->update_speed[ich];
+
+    s->cdlms[ich][ilms].lms_updates[recent + (s->cdlms[ich][ilms].order >> 4)] >>= 2;
+    s->cdlms[ich][ilms].lms_updates[recent + (s->cdlms[ich][ilms].order >> 3)] >>= 1;
+    s->cdlms[ich][ilms].recent = recent;
+}
+
+static void use_high_update_speed(WmallDecodeCtx *s, int ich)
+{
+    int ilms, recent, icoef;
+    for (ilms = s->cdlms_ttl[ich] - 1; ilms >= 0; ilms--) {
+        recent = s->cdlms[ich][ilms].recent;
+        if (s->update_speed[ich] == 16)
+            continue;
+        if (s->bV3RTM) {
+            for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+                s->cdlms[ich][ilms].lms_updates[icoef + recent] *= 2;
+        } else {
+            for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+                s->cdlms[ich][ilms].lms_updates[icoef] *= 2;
+        }
+    }
+    s->update_speed[ich] = 16;
+}
+
+static void use_normal_update_speed(WmallDecodeCtx *s, int ich)
+{
+    int ilms, recent, icoef;
+    for (ilms = s->cdlms_ttl[ich] - 1; ilms >= 0; ilms--) {
+        recent = s->cdlms[ich][ilms].recent;
+        if (s->update_speed[ich] == 8)
+            continue;
+        if (s->bV3RTM)
+            for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+                s->cdlms[ich][ilms].lms_updates[icoef + recent] /= 2;
+        else
+            for (icoef = 0; icoef < s->cdlms[ich][ilms].order; icoef++)
+                s->cdlms[ich][ilms].lms_updates[icoef] /= 2;
+    }
+    s->update_speed[ich] = 8;
+}
+
+static void revert_cdlms(WmallDecodeCtx *s, int ch,
+                         int coef_begin, int coef_end)
+{
+    int icoef, pred, ilms, num_lms, residue, input;
+
+    num_lms = s->cdlms_ttl[ch];
+    for (ilms = num_lms - 1; ilms >= 0; ilms--) {
+        for (icoef = coef_begin; icoef < coef_end; icoef++) {
+            pred = 1 << (s->cdlms[ch][ilms].scaling - 1);
+            residue = s->channel_residues[ch][icoef];
+            pred += lms_predict(s, ch, ilms);
+            input = residue + (pred >> s->cdlms[ch][ilms].scaling);
+            lms_update(s, ch, ilms, input, residue);
+            s->channel_residues[ch][icoef] = input;
+        }
+    }
+}
+
+static void revert_inter_ch_decorr(WmallDecodeCtx *s, int tile_size)
+{
+    if (s->num_channels != 2)
+        return;
+    else if (s->is_channel_coded[0] && s->is_channel_coded[1]) {
+        int icoef;
+        for (icoef = 0; icoef < tile_size; icoef++) {
+            s->channel_residues[0][icoef] -= s->channel_residues[1][icoef] >> 1;
+            s->channel_residues[1][icoef] += s->channel_residues[0][icoef];
+        }
+    }
+}
+
+static void revert_acfilter(WmallDecodeCtx *s, int tile_size)
+{
+    int ich, pred, i, j;
+    int64_t *filter_coeffs = s->acfilter_coeffs;
+    int scaling            = s->acfilter_scaling;
+    int order              = s->acfilter_order;
+
+    for (ich = 0; ich < s->num_channels; ich++) {
+        int *prevvalues = s->acfilter_prevvalues[ich];
+        for (i = 0; i < order; i++) {
+            pred = 0;
+            for (j = 0; j < order; j++) {
+                if (i <= j)
+                    pred += filter_coeffs[j] * prevvalues[j - i];
+                else
+                    pred += s->channel_residues[ich][i - j - 1] * filter_coeffs[j];
+            }
+            pred >>= scaling;
+            s->channel_residues[ich][i] += pred;
+        }
+        for (i = order; i < tile_size; i++) {
+            pred = 0;
+            for (j = 0; j < order; j++)
+                pred += s->channel_residues[ich][i - j - 1] * filter_coeffs[j];
+            pred >>= scaling;
+            s->channel_residues[ich][i] += pred;
+        }
+        for (j = 0; j < order; j++)
+            prevvalues[j] = s->channel_residues[ich][tile_size - j - 1];
+    }
+}
+
+static int decode_subframe(WmallDecodeCtx *s)
+{
+    int offset        = s->samples_per_frame;
+    int subframe_len  = s->samples_per_frame;
+    int total_samples = s->samples_per_frame * s->num_channels;
+    int i, j, rawpcm_tile, padding_zeroes;
+
+    s->subframe_offset = get_bits_count(&s->gb);
+
+    /* reset channel context and find the next block offset and size
+        == the next block of the channel with the smallest number of
+        decoded samples */
+    for (i = 0; i < s->num_channels; i++) {
+        if (offset > s->channel[i].decoded_samples) {
+            offset = s->channel[i].decoded_samples;
+            subframe_len =
+                s->channel[i].subframe_len[s->channel[i].cur_subframe];
+        }
+    }
+
+    /* get a list of all channels that contain the estimated block */
+    s->channels_for_cur_subframe = 0;
+    for (i = 0; i < s->num_channels; i++) {
+        const int cur_subframe = s->channel[i].cur_subframe;
+        /* subtract already processed samples */
+        total_samples -= s->channel[i].decoded_samples;
+
+        /* and count if there are multiple subframes that match our profile */
+        if (offset == s->channel[i].decoded_samples &&
+            subframe_len == s->channel[i].subframe_len[cur_subframe]) {
+            total_samples -= s->channel[i].subframe_len[cur_subframe];
+            s->channel[i].decoded_samples +=
+                s->channel[i].subframe_len[cur_subframe];
+            s->channel_indexes_for_cur_subframe[s->channels_for_cur_subframe] = i;
+            ++s->channels_for_cur_subframe;
+        }
+    }
+
+    /* check if the frame will be complete after processing the
+        estimated block */
+    if (!total_samples)
+        s->parsed_all_subframes = 1;
+
+
+    s->seekable_tile = get_bits1(&s->gb);
+    if (s->seekable_tile) {
+        clear_codec_buffers(s);
+
+        s->do_arith_coding    = get_bits1(&s->gb);
+        if (s->do_arith_coding) {
+            av_dlog(s->avctx, AV_LOG_DEBUG, "do_arith_coding == 1");
+            abort();
+        }
+        s->do_ac_filter       = get_bits1(&s->gb);
+        s->do_inter_ch_decorr = get_bits1(&s->gb);
+        s->do_mclms           = get_bits1(&s->gb);
+
+        if (s->do_ac_filter)
+            decode_ac_filter(s);
+
+        if (s->do_mclms)
+            decode_mclms(s);
+
+        decode_cdlms(s);
+        s->movave_scaling = get_bits(&s->gb, 3);
+        s->quant_stepsize = get_bits(&s->gb, 8) + 1;
+
+        reset_codec(s);
+    }
+
+    rawpcm_tile = get_bits1(&s->gb);
+
+    for (i = 0; i < s->num_channels; i++)
+        s->is_channel_coded[i] = 1;
+
+    if (!rawpcm_tile) {
+        for (i = 0; i < s->num_channels; i++)
+            s->is_channel_coded[i] = get_bits1(&s->gb);
+
+        if (s->bV3RTM) {
+            // LPC
+            s->do_lpc = get_bits1(&s->gb);
+            if (s->do_lpc) {
+                decode_lpc(s);
+                av_log_ask_for_sample(s->avctx, "Inverse LPC filter not "
+                                      "implemented. Expect wrong output.\n");
+            }
+        } else
+            s->do_lpc = 0;
+    }
+
+
+    if (get_bits1(&s->gb))
+        padding_zeroes = get_bits(&s->gb, 5);
+    else
+        padding_zeroes = 0;
+
+    if (rawpcm_tile) {
+        int bits = s->bits_per_sample - padding_zeroes;
+        av_dlog(s->avctx, AV_LOG_DEBUG, "RAWPCM %d bits per sample. "
+                "total %d bits, remain=%d\n", bits,
+                bits * s->num_channels * subframe_len, get_bits_count(&s->gb));
+        for (i = 0; i < s->num_channels; i++)
+            for (j = 0; j < subframe_len; j++)
+                s->channel_coeffs[i][j] = get_sbits(&s->gb, bits);
+    } else {
+        for (i = 0; i < s->num_channels; i++)
+            if (s->is_channel_coded[i]) {
+                decode_channel_residues(s, i, subframe_len);
+                if (s->seekable_tile)
+                    use_high_update_speed(s, i);
+                else
+                    use_normal_update_speed(s, i);
+                revert_cdlms(s, i, 0, subframe_len);
+            }
+    }
+    if (s->do_mclms)
+        revert_mclms(s, subframe_len);
+    if (s->do_inter_ch_decorr)
+        revert_inter_ch_decorr(s, subframe_len);
+    if (s->do_ac_filter)
+        revert_acfilter(s, subframe_len);
+
+    /* Dequantize */
+    if (s->quant_stepsize != 1)
+        for (i = 0; i < s->num_channels; i++)
+            for (j = 0; j < subframe_len; j++)
+                s->channel_residues[i][j] *= s->quant_stepsize;
+
+    /* Write to proper output buffer depending on bit-depth */
+    for (i = 0; i < subframe_len; i++)
+        for (j = 0; j < s->num_channels; j++) {
+            if (s->bits_per_sample == 16)
+                *s->samples_16++ = (int16_t) s->channel_residues[j][i];
+            else
+                *s->samples_32++ = s->channel_residues[j][i];
+        }
+
+    /* handled one subframe */
+    for (i = 0; i < s->channels_for_cur_subframe; i++) {
+        int c = s->channel_indexes_for_cur_subframe[i];
+        if (s->channel[c].cur_subframe >= s->channel[c].num_subframes) {
+            av_log(s->avctx, AV_LOG_ERROR, "broken subframe\n");
+            return AVERROR_INVALIDDATA;
+        }
+        ++s->channel[c].cur_subframe;
+    }
+    return 0;
+}
+
+/**
+ * @brief Decode one WMA frame.
+ * @param s codec context
+ * @return 0 if the trailer bit indicates that this is the last frame,
+ *         1 if there are additional frames
+ */
+static int decode_frame(WmallDecodeCtx *s)
+{
+    GetBitContext* gb = &s->gb;
+    int more_frames = 0, len = 0, i, ret;
+
+    s->frame.nb_samples = s->samples_per_frame;
+    if ((ret = s->avctx->get_buffer(s->avctx, &s->frame)) < 0) {
+        /* return an error if no frame could be decoded at all */
+        av_log(s->avctx, AV_LOG_ERROR,
+               "not enough space for the output samples\n");
+        s->packet_loss = 1;
+        return ret;
+    }
+    s->samples_16 = (int16_t *)s->frame.data[0];
+    s->samples_32 = (int32_t *)s->frame.data[0];
+
+    /* get frame length */
+    if (s->len_prefix)
+        len = get_bits(gb, s->log2_frame_size);
+
+    /* decode tile information */
+    if (decode_tilehdr(s)) {
+        s->packet_loss = 1;
+        return 0;
+    }
+
+    /* read drc info */
+    if (s->dynamic_range_compression)
+        s->drc_gain = get_bits(gb, 8);
+
+    /* no idea what these are for, might be the number of samples
+       that need to be skipped at the beginning or end of a stream */
+    if (get_bits1(gb)) {
+        int skip;
+
+        /* usually true for the first frame */
+        if (get_bits1(gb)) {
+            skip = get_bits(gb, av_log2(s->samples_per_frame * 2));
+            av_dlog(s->avctx, AV_LOG_DEBUG, "start skip: %i\n", skip);
+        }
+
+        /* sometimes true for the last frame */
+        if (get_bits1(gb)) {
+            skip = get_bits(gb, av_log2(s->samples_per_frame * 2));
+            av_dlog(s->avctx, AV_LOG_DEBUG, "end skip: %i\n", skip);
+        }
+
+    }
+
+    /* reset subframe states */
+    s->parsed_all_subframes = 0;
+    for (i = 0; i < s->num_channels; i++) {
+        s->channel[i].decoded_samples = 0;
+        s->channel[i].cur_subframe    = 0;
+    }
+
+    /* decode all subframes */
+    while (!s->parsed_all_subframes) {
+        if (decode_subframe(s) < 0) {
+            s->packet_loss = 1;
+            return 0;
+        }
+    }
+
+    av_dlog(s->avctx, AV_LOG_DEBUG, "Frame done\n");
+
+    if (s->skip_frame)
+        s->skip_frame = 0;
+
+    if (s->len_prefix) {
+        if (len != (get_bits_count(gb) - s->frame_offset) + 2) {
+            /* FIXME: not sure if this is always an error */
+            av_log(s->avctx, AV_LOG_ERROR,
+                   "frame[%i] would have to skip %i bits\n", s->frame_num,
+                   len - (get_bits_count(gb) - s->frame_offset) - 1);
+            s->packet_loss = 1;
+            return 0;
+        }
+
+        /* skip the rest of the frame data */
+        skip_bits_long(gb, len - (get_bits_count(gb) - s->frame_offset) - 1);
+    }
+
+    /* decode trailer bit */
+    more_frames = get_bits1(gb);
+    ++s->frame_num;
+    return more_frames;
+}
+
+/**
+ * @brief Calculate remaining input buffer length.
+ * @param s  codec context
+ * @param gb bitstream reader context
+ * @return remaining size in bits
+ */
+static int remaining_bits(WmallDecodeCtx *s, GetBitContext *gb)
+{
+    return s->buf_bit_size - get_bits_count(gb);
+}
+
+/**
+ * @brief Fill the bit reservoir with a (partial) frame.
+ * @param s      codec context
+ * @param gb     bitstream reader context
+ * @param len    length of the partial frame
+ * @param append decides whether to reset the buffer or not
+ */
+static void save_bits(WmallDecodeCtx *s, GetBitContext* gb, int len,
+                      int append)
+{
+    int buflen;
+    PutBitContext tmp;
+
+    /* when the frame data does not need to be concatenated, the input buffer
+        is reset and additional bits from the previous frame are copied
+        and skipped later so that a fast byte copy is possible */
+
+    if (!append) {
+        s->frame_offset   = get_bits_count(gb) & 7;
+        s->num_saved_bits = s->frame_offset;
+        init_put_bits(&s->pb, s->frame_data, MAX_FRAMESIZE);
+    }
+
+    buflen = (s->num_saved_bits + len + 8) >> 3;
+
+    if (len <= 0 || buflen > MAX_FRAMESIZE) {
+        av_log_ask_for_sample(s->avctx, "input buffer too small\n");
+        s->packet_loss = 1;
+        return;
+    }
+
+    s->num_saved_bits += len;
+    if (!append) {
+        avpriv_copy_bits(&s->pb, gb->buffer + (get_bits_count(gb) >> 3),
+                         s->num_saved_bits);
+    } else {
+        int align = 8 - (get_bits_count(gb) & 7);
+        align = FFMIN(align, len);
+        put_bits(&s->pb, align, get_bits(gb, align));
+        len -= align;
+        avpriv_copy_bits(&s->pb, gb->buffer + (get_bits_count(gb) >> 3), len);
+    }
+    skip_bits_long(gb, len);
+
+    tmp = s->pb;
+    flush_put_bits(&tmp);
+
+    init_get_bits(&s->gb, s->frame_data, s->num_saved_bits);
+    skip_bits(&s->gb, s->frame_offset);
+}
+
+/**
+ * @brief Decode a single WMA packet.
+ * @param avctx     codec context
+ * @param data      the output buffer
+ * @param data_size number of bytes that were written to the output buffer
+ * @param avpkt     input packet
+ * @return number of bytes that were read from the input buffer
+ */
+static int decode_packet(AVCodecContext *avctx, void *data, int *got_frame_ptr,
+                         AVPacket* avpkt)
+{
+    WmallDecodeCtx *s = avctx->priv_data;
+    GetBitContext* gb  = &s->pgb;
+    const uint8_t* buf = avpkt->data;
+    int buf_size       = avpkt->size;
+    int num_bits_prev_frame, packet_sequence_number,
+        seekable_frame_in_packet, spliced_packet;
+
+    if (s->packet_done || s->packet_loss) {
+        s->packet_done = 0;
+
+        /* sanity check for the buffer length */
+        if (buf_size < avctx->block_align)
+            return 0;
+
+        s->next_packet_start = buf_size - avctx->block_align;
+        buf_size             = avctx->block_align;
+        s->buf_bit_size      = buf_size << 3;
+
+        /* parse packet header */
+        init_get_bits(gb, buf, s->buf_bit_size);
+        packet_sequence_number   = get_bits(gb, 4);
+        seekable_frame_in_packet = get_bits1(gb);
+        spliced_packet           = get_bits1(gb);
+
+        /* get number of bits that need to be added to the previous frame */
+        num_bits_prev_frame = get_bits(gb, s->log2_frame_size);
+
+        /* check for packet loss */
+        if (!s->packet_loss &&
+            ((s->packet_sequence_number + 1) & 0xF) != packet_sequence_number) {
+            s->packet_loss = 1;
+            av_log(avctx, AV_LOG_ERROR, "Packet loss detected! seq %x vs %x\n",
+                   s->packet_sequence_number, packet_sequence_number);
+        }
+        s->packet_sequence_number = packet_sequence_number;
+
+        if (num_bits_prev_frame > 0) {
+            int remaining_packet_bits = s->buf_bit_size - get_bits_count(gb);
+            if (num_bits_prev_frame >= remaining_packet_bits) {
+                num_bits_prev_frame = remaining_packet_bits;
+                s->packet_done = 1;
+            }
+
+            /* Append the previous frame data to the remaining data from the
+             * previous packet to create a full frame. */
+            save_bits(s, gb, num_bits_prev_frame, 1);
+
+            /* decode the cross packet frame if it is valid */
+            if (!s->packet_loss)
+                decode_frame(s);
+        } else if (s->num_saved_bits - s->frame_offset) {
+            av_dlog(avctx, AV_LOG_DEBUG, "ignoring %x previously saved bits\n",
+                    s->num_saved_bits - s->frame_offset);
+        }
+
+        if (s->packet_loss) {
+            /* Reset number of saved bits so that the decoder does not start
+             * to decode incomplete frames in the s->len_prefix == 0 case. */
+            s->num_saved_bits = 0;
+            s->packet_loss    = 0;
+        }
+
+    } else {
+        int frame_size;
+
+        s->buf_bit_size = (avpkt->size - s->next_packet_start) << 3;
+        init_get_bits(gb, avpkt->data, s->buf_bit_size);
+        skip_bits(gb, s->packet_offset);
+
+        if (s->len_prefix && remaining_bits(s, gb) > s->log2_frame_size &&
+            (frame_size = show_bits(gb, s->log2_frame_size)) &&
+            frame_size <= remaining_bits(s, gb)) {
+            save_bits(s, gb, frame_size, 0);
+            s->packet_done = !decode_frame(s);
+        } else if (!s->len_prefix
+                   && s->num_saved_bits > get_bits_count(&s->gb)) {
+            /* when the frames do not have a length prefix, we don't know the
+             * compressed length of the individual frames however, we know what
+             * part of a new packet belongs to the previous frame therefore we
+             * save the incoming packet first, then we append the "previous
+             * frame" data from the next packet so that we get a buffer that
+             * only contains full frames */
+            s->packet_done = !decode_frame(s);
+        } else {
+            s->packet_done = 1;
+        }
+    }
+
+    if (s->packet_done && !s->packet_loss &&
+        remaining_bits(s, gb) > 0) {
+        /* save the rest of the data so that it can be decoded
+         * with the next packet */
+        save_bits(s, gb, remaining_bits(s, gb), 0);
+    }
+
+    *(AVFrame *)data = s->frame;
+    *got_frame_ptr   = 1;
+    s->packet_offset = get_bits_count(gb) & 7;
+
+    return (s->packet_loss) ? AVERROR_INVALIDDATA : get_bits_count(gb) >> 3;
+}
+
+
+AVCodec ff_wmalossless_decoder = {
+    .name           = "wmalossless",
+    .type           = AVMEDIA_TYPE_AUDIO,
+    .id             = CODEC_ID_WMALOSSLESS,
+    .priv_data_size = sizeof(WmallDecodeCtx),
+    .init           = decode_init,
+    .decode         = decode_packet,
+    .capabilities   = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1,
+    .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Audio Lossless"),
+};