tak: demuxer, parser, and decoder
authorPaul B Mahol <onemda@gmail.com>
Tue, 2 Oct 2012 13:43:19 +0000 (13:43 +0000)
committerJustin Ruggles <justin.ruggles@gmail.com>
Fri, 7 Dec 2012 21:15:02 +0000 (16:15 -0500)
Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Justin Ruggles <justin.ruggles@gmail.com>
16 files changed:
Changelog
configure
doc/general.texi
libavcodec/Makefile
libavcodec/allcodecs.c
libavcodec/avcodec.h
libavcodec/codec_desc.c
libavcodec/tak.c [new file with mode: 0644]
libavcodec/tak.h [new file with mode: 0644]
libavcodec/tak_parser.c [new file with mode: 0644]
libavcodec/takdec.c [new file with mode: 0644]
libavcodec/version.h
libavformat/Makefile
libavformat/allformats.c
libavformat/takdec.c [new file with mode: 0644]
libavformat/version.h

index 22374a3..b4a8e60 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -7,6 +7,7 @@ version <next>:
 - audio volume filter
 - deprecated the avconv -vol option. the volume filter is to be used instead.
 - multi-channel ALAC encoding up to 7.1
+- TAK demuxer, parser, and decoder
 
 
 version 9_beta2:
index bf809c6..db0e758 100755 (executable)
--- a/configure
+++ b/configure
@@ -1675,6 +1675,7 @@ sap_muxer_select="rtp_muxer rtp_protocol"
 sdp_demuxer_select="rtpdec"
 smoothstreaming_muxer_select="ismv_muxer"
 spdif_muxer_select="aac_parser"
+tak_demuxer_select="tak_parser"
 tg2_muxer_select="mov_muxer"
 tgp_muxer_select="mov_muxer"
 w64_demuxer_deps="wav_demuxer"
index 7b78308..d973902 100644 (file)
@@ -273,6 +273,7 @@ library:
 @item raw video                 @tab X @tab X
 @item raw id RoQ                @tab X @tab
 @item raw Shorten               @tab   @tab X
+@item raw TAK                   @tab   @tab X
 @item raw TrueHD                @tab X @tab X
 @item raw VC-1                  @tab   @tab X
 @item raw PCM A-law             @tab X @tab X
@@ -800,6 +801,7 @@ following image formats are supported:
 @item SMPTE 302M AES3 audio  @tab     @tab  X
 @item Speex                  @tab  E  @tab  E
     @tab supported through external library libspeex
+@item TAK (Tom's lossless Audio Kompressor)  @tab     @tab  X
 @item True Audio (TTA)       @tab     @tab  X
 @item TrueHD                 @tab     @tab  X
     @tab Used in HD-DVD and Blu-Ray discs.
index 502489d..b535430 100644 (file)
@@ -349,6 +349,7 @@ OBJS-$(CONFIG_SVQ3_DECODER)            += svq3.o svq13.o h263.o h264.o        \
                                           h264_loopfilter.o h264_direct.o     \
                                           h264_sei.o h264_ps.o h264_refs.o    \
                                           h264_cavlc.o h264_cabac.o cabac.o
+OBJS-$(CONFIG_TAK_DECODER)             += takdec.o tak.o
 OBJS-$(CONFIG_TARGA_DECODER)           += targa.o
 OBJS-$(CONFIG_TARGA_ENCODER)           += targaenc.o rle.o
 OBJS-$(CONFIG_THEORA_DECODER)          += xiph.o
@@ -555,6 +556,7 @@ OBJS-$(CONFIG_RTP_MUXER)               += mpeg4audio.o xiph.o
 OBJS-$(CONFIG_RTPDEC)                  += mjpeg.o
 OBJS-$(CONFIG_SPDIF_DEMUXER)           += aacadtsdec.o mpeg4audio.o
 OBJS-$(CONFIG_SPDIF_MUXER)             += dca.o
+OBJS-$(CONFIG_TAK_DEMUXER)             += tak.o
 OBJS-$(CONFIG_WEBM_MUXER)              += mpeg4audio.o mpegaudiodata.o  \
                                           xiph.o flac.o flacdata.o
 OBJS-$(CONFIG_WTV_DEMUXER)             += mpeg4audio.o mpegaudiodata.o
@@ -631,6 +633,7 @@ OBJS-$(CONFIG_MPEGVIDEO_PARSER)        += mpegvideo_parser.o    \
 OBJS-$(CONFIG_PNM_PARSER)              += pnm_parser.o pnm.o
 OBJS-$(CONFIG_RV30_PARSER)             += rv34_parser.o
 OBJS-$(CONFIG_RV40_PARSER)             += rv34_parser.o
+OBJS-$(CONFIG_TAK_PARSER)              += tak_parser.o tak.o
 OBJS-$(CONFIG_VC1_PARSER)              += vc1_parser.o vc1.o vc1data.o \
                                           msmpeg4.o msmpeg4data.o mpeg4video.o \
                                           h263.o
index e5fb351..03949e6 100644 (file)
@@ -295,6 +295,7 @@ void avcodec_register_all(void)
     REGISTER_DECODER (SHORTEN, shorten);
     REGISTER_DECODER (SIPR, sipr);
     REGISTER_DECODER (SMACKAUD, smackaud);
+    REGISTER_DECODER (TAK, tak);
     REGISTER_DECODER (TRUEHD, truehd);
     REGISTER_DECODER (TRUESPEECH, truespeech);
     REGISTER_DECODER (TTA, tta);
@@ -431,6 +432,7 @@ void avcodec_register_all(void)
     REGISTER_PARSER  (PNM, pnm);
     REGISTER_PARSER  (RV30, rv30);
     REGISTER_PARSER  (RV40, rv40);
+    REGISTER_PARSER  (TAK, tak);
     REGISTER_PARSER  (VC1, vc1);
     REGISTER_PARSER  (VORBIS, vorbis);
     REGISTER_PARSER  (VP3, vp3);
index 5e358ca..29e3701 100644 (file)
@@ -407,6 +407,7 @@ enum AVCodecID {
     AV_CODEC_ID_ILBC,
     AV_CODEC_ID_OPUS,
     AV_CODEC_ID_COMFORT_NOISE,
+    AV_CODEC_ID_TAK,
 
     /* subtitle codecs */
     AV_CODEC_ID_FIRST_SUBTITLE = 0x17000,          ///< A dummy ID pointing at the start of subtitle codecs.
index a8ff314..0400c31 100644 (file)
@@ -2119,6 +2119,13 @@ static const AVCodecDescriptor codec_descriptors[] = {
         .long_name = NULL_IF_CONFIG_SMALL("RFC 3389 Comfort Noise"),
         .props     = AV_CODEC_PROP_LOSSY,
     },
+    {
+        .id        = AV_CODEC_ID_TAK,
+        .type      = AVMEDIA_TYPE_AUDIO,
+        .name      = "tak",
+        .long_name = NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
+        .props     = AV_CODEC_PROP_LOSSLESS,
+    },
 
     /* subtitle codecs */
     {
diff --git a/libavcodec/tak.c b/libavcodec/tak.c
new file mode 100644 (file)
index 0000000..867a84b
--- /dev/null
@@ -0,0 +1,150 @@
+/*
+ * TAK common code
+ * Copyright (c) 2012 Paul B Mahol
+ *
+ * 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 "libavutil/bswap.h"
+#include "libavutil/crc.h"
+#include "libavutil/intreadwrite.h"
+#include "tak.h"
+
+static const uint16_t frame_duration_type_quants[] = {
+    3, 4, 6, 8, 4096, 8192, 16384, 512, 1024, 2048,
+};
+
+static int tak_get_nb_samples(int sample_rate, enum TAKFrameSizeType type)
+{
+    int nb_samples, max_nb_samples;
+
+    if (type <= TAK_FST_250ms) {
+        nb_samples     = sample_rate * frame_duration_type_quants[type] >>
+                         TAK_FRAME_DURATION_QUANT_SHIFT;
+        max_nb_samples = 16384;
+    } else if (type < FF_ARRAY_ELEMS(frame_duration_type_quants)) {
+        nb_samples     = frame_duration_type_quants[type];
+        max_nb_samples = sample_rate *
+                         frame_duration_type_quants[TAK_FST_250ms] >>
+                         TAK_FRAME_DURATION_QUANT_SHIFT;
+    } else {
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (nb_samples <= 0 || nb_samples > max_nb_samples)
+        return AVERROR_INVALIDDATA;
+
+    return nb_samples;
+}
+
+static int crc_init = 0;
+#if CONFIG_SMALL
+#define CRC_TABLE_SIZE 257
+#else
+#define CRC_TABLE_SIZE 1024
+#endif
+static AVCRC crc_24[CRC_TABLE_SIZE];
+
+av_cold void ff_tak_init_crc(void)
+{
+    if (!crc_init) {
+        av_crc_init(crc_24, 0, 24, 0x864CFBU, sizeof(crc_24));
+        crc_init = 1;
+    }
+}
+
+int ff_tak_check_crc(const uint8_t *buf, unsigned int buf_size)
+{
+    uint32_t crc, CRC;
+
+    if (buf_size < 4)
+        return AVERROR_INVALIDDATA;
+    buf_size -= 3;
+
+    CRC = av_bswap32(AV_RL24(buf + buf_size)) >> 8;
+    crc = av_crc(crc_24, 0xCE04B7U, buf, buf_size);
+    if (CRC != crc)
+        return AVERROR_INVALIDDATA;
+
+    return 0;
+}
+
+void avpriv_tak_parse_streaminfo(GetBitContext *gb, TAKStreamInfo *s)
+{
+    uint64_t channel_mask = 0;
+    int frame_type, i;
+
+    s->codec = get_bits(gb, TAK_ENCODER_CODEC_BITS);
+    skip_bits(gb, TAK_ENCODER_PROFILE_BITS);
+
+    frame_type = get_bits(gb, TAK_SIZE_FRAME_DURATION_BITS);
+    s->samples = get_bits64(gb, TAK_SIZE_SAMPLES_NUM_BITS);
+
+    s->data_type   = get_bits(gb, TAK_FORMAT_DATA_TYPE_BITS);
+    s->sample_rate = get_bits(gb, TAK_FORMAT_SAMPLE_RATE_BITS) +
+                     TAK_SAMPLE_RATE_MIN;
+    s->bps         = get_bits(gb, TAK_FORMAT_BPS_BITS) +
+                     TAK_BPS_MIN;
+    s->channels    = get_bits(gb, TAK_FORMAT_CHANNEL_BITS) +
+                     TAK_CHANNELS_MIN;
+
+    if (get_bits1(gb)) {
+        skip_bits(gb, TAK_FORMAT_VALID_BITS);
+        if (get_bits1(gb)) {
+            for (i = 0; i < s->channels; i++) {
+                int value = get_bits(gb, TAK_FORMAT_CH_LAYOUT_BITS);
+
+                if (value > 0 && value <= 18)
+                    channel_mask |= 1 << (value - 1);
+            }
+        }
+    }
+
+    s->ch_layout     = channel_mask;
+    s->frame_samples = tak_get_nb_samples(s->sample_rate, frame_type);
+}
+
+int ff_tak_decode_frame_header(AVCodecContext *avctx, GetBitContext *gb,
+                               TAKStreamInfo *ti, int log_level_offset)
+{
+    if (get_bits(gb, TAK_FRAME_HEADER_SYNC_ID_BITS) != TAK_FRAME_HEADER_SYNC_ID) {
+        av_log(avctx, AV_LOG_ERROR + log_level_offset, "missing sync id\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    ti->flags     = get_bits(gb, TAK_FRAME_HEADER_FLAGS_BITS);
+    ti->frame_num = get_bits(gb, TAK_FRAME_HEADER_NO_BITS);
+
+    if (ti->flags & TAK_FRAME_FLAG_IS_LAST) {
+        ti->last_frame_samples = get_bits(gb, TAK_FRAME_HEADER_SAMPLE_COUNT_BITS) + 1;
+        skip_bits(gb, 2);
+    } else {
+        ti->last_frame_samples = 0;
+    }
+
+    if (ti->flags & TAK_FRAME_FLAG_HAS_INFO) {
+        avpriv_tak_parse_streaminfo(gb, ti);
+
+        if (get_bits(gb, 6))
+            skip_bits(gb, 25);
+        align_get_bits(gb);
+    }
+
+    skip_bits(gb, 24);
+
+    return 0;
+}
diff --git a/libavcodec/tak.h b/libavcodec/tak.h
new file mode 100644 (file)
index 0000000..fa91149
--- /dev/null
@@ -0,0 +1,166 @@
+/*
+ * TAK decoder/demuxer common code
+ * Copyright (c) 2012 Paul B Mahol
+ *
+ * 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
+ */
+
+/**
+ * @file
+ * TAK (Tom's lossless Audio Kompressor) decoder/demuxer common functions
+ */
+
+#ifndef AVCODEC_TAK_H
+#define AVCODEC_TAK_H
+
+#include <stdint.h>
+
+#define BITSTREAM_READER_LE
+#include "get_bits.h"
+#include "avcodec.h"
+
+#define TAK_FORMAT_DATA_TYPE_BITS               3
+#define TAK_FORMAT_SAMPLE_RATE_BITS            18
+#define TAK_FORMAT_BPS_BITS                     5
+#define TAK_FORMAT_CHANNEL_BITS                 4
+#define TAK_FORMAT_VALID_BITS                   5
+#define TAK_FORMAT_CH_LAYOUT_BITS               6
+#define TAK_SIZE_FRAME_DURATION_BITS            4
+#define TAK_SIZE_SAMPLES_NUM_BITS              35
+#define TAK_LAST_FRAME_POS_BITS                40
+#define TAK_LAST_FRAME_SIZE_BITS               24
+#define TAK_ENCODER_CODEC_BITS                  6
+#define TAK_ENCODER_PROFILE_BITS                4
+#define TAK_ENCODER_VERSION_BITS               24
+#define TAK_SAMPLE_RATE_MIN                  6000
+#define TAK_CHANNELS_MIN                        1
+#define TAK_BPS_MIN                             8
+#define TAK_FRAME_HEADER_FLAGS_BITS             3
+#define TAK_FRAME_HEADER_SYNC_ID           0xA0FF
+#define TAK_FRAME_HEADER_SYNC_ID_BITS          16
+#define TAK_FRAME_HEADER_SAMPLE_COUNT_BITS     14
+#define TAK_FRAME_HEADER_NO_BITS               21
+#define TAK_FRAME_DURATION_QUANT_SHIFT          5
+#define TAK_CRC24_BITS                         24
+
+
+#define TAK_FRAME_FLAG_IS_LAST                0x1
+#define TAK_FRAME_FLAG_HAS_INFO               0x2
+#define TAK_FRAME_FLAG_HAS_METADATA           0x4
+
+#define TAK_MAX_CHANNELS               (1 << TAK_FORMAT_CHANNEL_BITS)
+
+#define TAK_MIN_FRAME_HEADER_BITS      (TAK_FRAME_HEADER_SYNC_ID_BITS + \
+                                        TAK_FRAME_HEADER_FLAGS_BITS   + \
+                                        TAK_FRAME_HEADER_NO_BITS      + \
+                                        TAK_CRC24_BITS)
+
+#define TAK_MIN_FRAME_HEADER_LAST_BITS (TAK_MIN_FRAME_HEADER_BITS + 2 + \
+                                        TAK_FRAME_HEADER_SAMPLE_COUNT_BITS)
+
+#define TAK_ENCODER_BITS               (TAK_ENCODER_CODEC_BITS + \
+                                        TAK_ENCODER_PROFILE_BITS)
+
+#define TAK_SIZE_BITS                  (TAK_SIZE_SAMPLES_NUM_BITS + \
+                                        TAK_SIZE_FRAME_DURATION_BITS)
+
+#define TAK_FORMAT_BITS                (TAK_FORMAT_DATA_TYPE_BITS   + \
+                                        TAK_FORMAT_SAMPLE_RATE_BITS + \
+                                        TAK_FORMAT_BPS_BITS         + \
+                                        TAK_FORMAT_CHANNEL_BITS + 1 + \
+                                        TAK_FORMAT_VALID_BITS   + 1 + \
+                                        TAK_FORMAT_CH_LAYOUT_BITS   * \
+                                        TAK_MAX_CHANNELS)
+
+#define TAK_STREAMINFO_BITS            (TAK_ENCODER_BITS + \
+                                        TAK_SIZE_BITS    + \
+                                        TAK_FORMAT_BITS)
+
+#define TAK_MAX_FRAME_HEADER_BITS      (TAK_MIN_FRAME_HEADER_LAST_BITS + \
+                                        TAK_STREAMINFO_BITS + 31)
+
+#define TAK_STREAMINFO_BYTES           ((TAK_STREAMINFO_BITS       + 7) / 8)
+#define TAK_MAX_FRAME_HEADER_BYTES     ((TAK_MAX_FRAME_HEADER_BITS + 7) / 8)
+#define TAK_MIN_FRAME_HEADER_BYTES     ((TAK_MIN_FRAME_HEADER_BITS + 7) / 8)
+
+enum TAKCodecType {
+    TAK_CODEC_MONO_STEREO  = 2,
+    TAK_CODEC_MULTICHANNEL = 4
+};
+
+enum TAKMetaDataType {
+    TAK_METADATA_END = 0,
+    TAK_METADATA_STREAMINFO,
+    TAK_METADATA_SEEKTABLE,
+    TAK_METADATA_SIMPLE_WAVE_DATA,
+    TAK_METADATA_ENCODER,
+    TAK_METADATA_PADDING,
+    TAK_METADATA_MD5,
+    TAK_METADATA_LAST_FRAME,
+};
+
+enum TAKFrameSizeType {
+    TAK_FST_94ms = 0,
+    TAK_FST_125ms,
+    TAK_FST_188ms,
+    TAK_FST_250ms,
+    TAK_FST_4096,
+    TAK_FST_8192,
+    TAK_FST_16384,
+    TAK_FST_512,
+    TAK_FST_1024,
+    TAK_FST_2048,
+};
+
+typedef struct TAKStreamInfo {
+    int               flags;
+    enum TAKCodecType codec;
+    int               data_type;
+    int               sample_rate;
+    int               channels;
+    int               bps;
+    int               frame_num;
+    int               frame_samples;
+    int               last_frame_samples;
+    uint64_t          ch_layout;
+    int64_t           samples;
+} TAKStreamInfo;
+
+void ff_tak_init_crc(void);
+
+int ff_tak_check_crc(const uint8_t *buf, unsigned int buf_size);
+
+/**
+ * Parse the Streaminfo metadata block.
+ * @param[in]  gb pointer to GetBitContext
+ * @param[out] s  storage for parsed information
+ */
+void avpriv_tak_parse_streaminfo(GetBitContext *gb, TAKStreamInfo *s);
+
+/**
+ * Validate and decode a frame header.
+ * @param      avctx             AVCodecContext to use as av_log() context
+ * @param[in]  gb                GetBitContext from which to read frame header
+ * @param[out] s                 frame information
+ * @param      log_level_offset  log level offset, can be used to silence
+ *                               error messages.
+ * @return non-zero on error, 0 if OK
+ */
+int ff_tak_decode_frame_header(AVCodecContext *avctx, GetBitContext *gb,
+                               TAKStreamInfo *s, int log_level_offset);
+
+#endif /* AVCODEC_TAK_H */
diff --git a/libavcodec/tak_parser.c b/libavcodec/tak_parser.c
new file mode 100644 (file)
index 0000000..295df24
--- /dev/null
@@ -0,0 +1,128 @@
+/*
+ * TAK parser
+ * Copyright (c) 2012 Michael Niedermayer
+ *
+ * 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
+ */
+
+/**
+ * @file
+ * TAK parser
+ **/
+
+#include "tak.h"
+#include "parser.h"
+
+typedef struct TAKParseContext {
+    ParseContext  pc;
+    TAKStreamInfo ti;
+    int           index;
+} TAKParseContext;
+
+static av_cold int tak_init(AVCodecParserContext *s)
+{
+    ff_tak_init_crc();
+    return 0;
+}
+
+static int tak_parse(AVCodecParserContext *s, AVCodecContext *avctx,
+                     const uint8_t **poutbuf, int *poutbuf_size,
+                     const uint8_t *buf, int buf_size)
+{
+    TAKParseContext *t = s->priv_data;
+    ParseContext *pc   = &t->pc;
+    int next           = END_NOT_FOUND;
+    GetBitContext gb;
+    int consumed = 0;
+    int needed   = buf_size ? TAK_MAX_FRAME_HEADER_BYTES : 8;
+
+    if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
+        TAKStreamInfo ti;
+        init_get_bits(&gb, buf, buf_size);
+        if (!ff_tak_decode_frame_header(avctx, &gb, &ti, 127))
+            s->duration = t->ti.last_frame_samples ? t->ti.last_frame_samples
+                                                   : t->ti.frame_samples;
+        *poutbuf      = buf;
+        *poutbuf_size = buf_size;
+        return buf_size;
+    }
+
+    while (buf_size || t->index + needed <= pc->index) {
+        if (buf_size && t->index + TAK_MAX_FRAME_HEADER_BYTES > pc->index) {
+            int tmp_buf_size       = FFMIN(2 * TAK_MAX_FRAME_HEADER_BYTES,
+                                           buf_size);
+            const uint8_t *tmp_buf = buf;
+
+            ff_combine_frame(pc, END_NOT_FOUND, &tmp_buf, &tmp_buf_size);
+            consumed += tmp_buf_size;
+            buf      += tmp_buf_size;
+            buf_size -= tmp_buf_size;
+        }
+
+        for (; t->index + needed <= pc->index; t->index++)
+            if (pc->buffer[t->index]     == 0xFF &&
+                pc->buffer[t->index + 1] == 0xA0) {
+                TAKStreamInfo ti;
+
+                init_get_bits(&gb, pc->buffer + t->index,
+                              8 * (pc->index - t->index));
+                if (!ff_tak_decode_frame_header(avctx, &gb,
+                                                pc->frame_start_found ? &ti
+                                                                      : &t->ti,
+                                                127) &&
+                    !ff_tak_check_crc(pc->buffer + t->index,
+                                      get_bits_count(&gb) / 8)) {
+                    if (!pc->frame_start_found) {
+                        pc->frame_start_found = 1;
+                        s->duration           = t->ti.last_frame_samples ?
+                                                t->ti.last_frame_samples :
+                                                t->ti.frame_samples;
+                    } else {
+                        pc->frame_start_found = 0;
+                        next                  = t->index - pc->index;
+                        t->index              = 0;
+                        goto found;
+                    }
+                }
+            }
+    }
+
+found:
+    if (consumed && !buf_size && next == END_NOT_FOUND ||
+        ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
+        *poutbuf      = NULL;
+        *poutbuf_size = 0;
+        return buf_size + consumed;
+    }
+
+    if (next != END_NOT_FOUND) {
+        next        += consumed;
+        pc->overread = FFMAX(0, -next);
+    }
+
+    *poutbuf      = buf;
+    *poutbuf_size = buf_size;
+    return next;
+}
+
+AVCodecParser ff_tak_parser = {
+    .codec_ids      = { AV_CODEC_ID_TAK },
+    .priv_data_size = sizeof(TAKParseContext),
+    .parser_init    = tak_init,
+    .parser_parse   = tak_parse,
+    .parser_close   = ff_parse_close,
+};
diff --git a/libavcodec/takdec.c b/libavcodec/takdec.c
new file mode 100644 (file)
index 0000000..87fcf83
--- /dev/null
@@ -0,0 +1,929 @@
+/*
+ * TAK decoder
+ * Copyright (c) 2012 Paul B Mahol
+ *
+ * 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
+ */
+
+/**
+ * @file
+ * TAK (Tom's lossless Audio Kompressor) decoder
+ * @author Paul B Mahol
+ */
+
+#include "libavutil/samplefmt.h"
+#include "tak.h"
+#include "avcodec.h"
+#include "dsputil.h"
+#include "internal.h"
+#include "unary.h"
+
+#define MAX_SUBFRAMES     8                         // max number of subframes per channel
+#define MAX_PREDICTORS  256
+
+typedef struct MCDParam {
+    int8_t present;                                 // decorrelation parameter availability for this channel
+    int8_t index;                                   // index into array of decorrelation types
+    int8_t chan1;
+    int8_t chan2;
+} MCDParam;
+
+typedef struct TAKDecContext {
+    AVCodecContext *avctx;                          // parent AVCodecContext
+    AVFrame         frame;                          // AVFrame for decoded output
+    DSPContext      dsp;
+    TAKStreamInfo   ti;
+    GetBitContext   gb;                             // bitstream reader initialized to start at the current frame
+
+    int             uval;
+    int             nb_samples;                     // number of samples in the current frame
+    uint8_t        *decode_buffer;
+    unsigned int    decode_buffer_size;
+    int32_t        *decoded[TAK_MAX_CHANNELS];      // decoded samples for each channel
+
+    int8_t          lpc_mode[TAK_MAX_CHANNELS];
+    int8_t          sample_shift[TAK_MAX_CHANNELS]; // shift applied to every sample in the channel
+    int             subframe_scale;
+
+    int8_t          dmode;                          // channel decorrelation type in the current frame
+
+    MCDParam        mcdparams[TAK_MAX_CHANNELS];    // multichannel decorrelation parameters
+
+    int16_t        *residues;
+    unsigned int    residues_buf_size;
+} TAKDecContext;
+
+static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
+
+static const uint16_t predictor_sizes[] = {
+    4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
+};
+
+static const struct CParam {
+    int init;
+    int escape;
+    int scale;
+    int aescape;
+    int bias;
+} xcodes[50] = {
+    { 0x01, 0x0000001, 0x0000001, 0x0000003, 0x0000008 },
+    { 0x02, 0x0000003, 0x0000001, 0x0000007, 0x0000006 },
+    { 0x03, 0x0000005, 0x0000002, 0x000000E, 0x000000D },
+    { 0x03, 0x0000003, 0x0000003, 0x000000D, 0x0000018 },
+    { 0x04, 0x000000B, 0x0000004, 0x000001C, 0x0000019 },
+    { 0x04, 0x0000006, 0x0000006, 0x000001A, 0x0000030 },
+    { 0x05, 0x0000016, 0x0000008, 0x0000038, 0x0000032 },
+    { 0x05, 0x000000C, 0x000000C, 0x0000034, 0x0000060 },
+    { 0x06, 0x000002C, 0x0000010, 0x0000070, 0x0000064 },
+    { 0x06, 0x0000018, 0x0000018, 0x0000068, 0x00000C0 },
+    { 0x07, 0x0000058, 0x0000020, 0x00000E0, 0x00000C8 },
+    { 0x07, 0x0000030, 0x0000030, 0x00000D0, 0x0000180 },
+    { 0x08, 0x00000B0, 0x0000040, 0x00001C0, 0x0000190 },
+    { 0x08, 0x0000060, 0x0000060, 0x00001A0, 0x0000300 },
+    { 0x09, 0x0000160, 0x0000080, 0x0000380, 0x0000320 },
+    { 0x09, 0x00000C0, 0x00000C0, 0x0000340, 0x0000600 },
+    { 0x0A, 0x00002C0, 0x0000100, 0x0000700, 0x0000640 },
+    { 0x0A, 0x0000180, 0x0000180, 0x0000680, 0x0000C00 },
+    { 0x0B, 0x0000580, 0x0000200, 0x0000E00, 0x0000C80 },
+    { 0x0B, 0x0000300, 0x0000300, 0x0000D00, 0x0001800 },
+    { 0x0C, 0x0000B00, 0x0000400, 0x0001C00, 0x0001900 },
+    { 0x0C, 0x0000600, 0x0000600, 0x0001A00, 0x0003000 },
+    { 0x0D, 0x0001600, 0x0000800, 0x0003800, 0x0003200 },
+    { 0x0D, 0x0000C00, 0x0000C00, 0x0003400, 0x0006000 },
+    { 0x0E, 0x0002C00, 0x0001000, 0x0007000, 0x0006400 },
+    { 0x0E, 0x0001800, 0x0001800, 0x0006800, 0x000C000 },
+    { 0x0F, 0x0005800, 0x0002000, 0x000E000, 0x000C800 },
+    { 0x0F, 0x0003000, 0x0003000, 0x000D000, 0x0018000 },
+    { 0x10, 0x000B000, 0x0004000, 0x001C000, 0x0019000 },
+    { 0x10, 0x0006000, 0x0006000, 0x001A000, 0x0030000 },
+    { 0x11, 0x0016000, 0x0008000, 0x0038000, 0x0032000 },
+    { 0x11, 0x000C000, 0x000C000, 0x0034000, 0x0060000 },
+    { 0x12, 0x002C000, 0x0010000, 0x0070000, 0x0064000 },
+    { 0x12, 0x0018000, 0x0018000, 0x0068000, 0x00C0000 },
+    { 0x13, 0x0058000, 0x0020000, 0x00E0000, 0x00C8000 },
+    { 0x13, 0x0030000, 0x0030000, 0x00D0000, 0x0180000 },
+    { 0x14, 0x00B0000, 0x0040000, 0x01C0000, 0x0190000 },
+    { 0x14, 0x0060000, 0x0060000, 0x01A0000, 0x0300000 },
+    { 0x15, 0x0160000, 0x0080000, 0x0380000, 0x0320000 },
+    { 0x15, 0x00C0000, 0x00C0000, 0x0340000, 0x0600000 },
+    { 0x16, 0x02C0000, 0x0100000, 0x0700000, 0x0640000 },
+    { 0x16, 0x0180000, 0x0180000, 0x0680000, 0x0C00000 },
+    { 0x17, 0x0580000, 0x0200000, 0x0E00000, 0x0C80000 },
+    { 0x17, 0x0300000, 0x0300000, 0x0D00000, 0x1800000 },
+    { 0x18, 0x0B00000, 0x0400000, 0x1C00000, 0x1900000 },
+    { 0x18, 0x0600000, 0x0600000, 0x1A00000, 0x3000000 },
+    { 0x19, 0x1600000, 0x0800000, 0x3800000, 0x3200000 },
+    { 0x19, 0x0C00000, 0x0C00000, 0x3400000, 0x6000000 },
+    { 0x1A, 0x2C00000, 0x1000000, 0x7000000, 0x6400000 },
+    { 0x1A, 0x1800000, 0x1800000, 0x6800000, 0xC000000 },
+};
+
+static av_cold void tak_init_static_data(AVCodec *codec)
+{
+    ff_tak_init_crc();
+}
+
+static int set_bps_params(AVCodecContext *avctx)
+{
+    switch (avctx->bits_per_coded_sample) {
+    case 8:
+        avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
+        break;
+    case 16:
+        avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
+        break;
+    case 24:
+        avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
+        break;
+    default:
+        av_log(avctx, AV_LOG_ERROR, "unsupported bits per sample: %d\n",
+               avctx->bits_per_coded_sample);
+        return AVERROR_INVALIDDATA;
+    }
+    avctx->bits_per_raw_sample = avctx->bits_per_coded_sample;
+
+    return 0;
+}
+
+static void set_sample_rate_params(AVCodecContext *avctx)
+{
+    TAKDecContext *s  = avctx->priv_data;
+    int shift         = 3 - (avctx->sample_rate / 11025);
+    shift             = FFMAX(0, shift);
+    s->uval           = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << shift;
+    s->subframe_scale = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << 1;
+}
+
+static av_cold int tak_decode_init(AVCodecContext *avctx)
+{
+    TAKDecContext *s = avctx->priv_data;
+
+    ff_dsputil_init(&s->dsp, avctx);
+
+    s->avctx = avctx;
+    avcodec_get_frame_defaults(&s->frame);
+    avctx->coded_frame = &s->frame;
+
+    set_sample_rate_params(avctx);
+
+    return set_bps_params(avctx);
+}
+
+static void decode_lpc(int32_t *coeffs, int mode, int length)
+{
+    int i;
+
+    if (length < 2)
+        return;
+
+    if (mode == 1) {
+        int a1 = *coeffs++;
+        for (i = 0; i < length - 1 >> 1; i++) {
+            *coeffs   += a1;
+            coeffs[1] += *coeffs;
+            a1         = coeffs[1];
+            coeffs    += 2;
+        }
+        if (length - 1 & 1)
+            *coeffs += a1;
+    } else if (mode == 2) {
+        int a1    = coeffs[1];
+        int a2    = a1 + *coeffs;
+        coeffs[1] = a2;
+        if (length > 2) {
+            coeffs += 2;
+            for (i = 0; i < length - 2 >> 1; i++) {
+                int a3    = *coeffs + a1;
+                int a4    = a3 + a2;
+                *coeffs   = a4;
+                a1        = coeffs[1] + a3;
+                a2        = a1 + a4;
+                coeffs[1] = a2;
+                coeffs   += 2;
+            }
+            if (length & 1)
+                *coeffs += a1 + a2;
+        }
+    } else if (mode == 3) {
+        int a1    = coeffs[1];
+        int a2    = a1 + *coeffs;
+        coeffs[1] = a2;
+        if (length > 2) {
+            int a3  = coeffs[2];
+            int a4  = a3 + a1;
+            int a5  = a4 + a2;
+            coeffs += 3;
+            for (i = 0; i < length - 3; i++) {
+                a3     += *coeffs;
+                a4     += a3;
+                a5     += a4;
+                *coeffs = a5;
+                coeffs++;
+            }
+        }
+    }
+}
+
+static int decode_segment(GetBitContext *gb, int mode, int32_t *decoded,
+                          int len)
+{
+    struct CParam code;
+    int i;
+
+    if (!mode) {
+        memset(decoded, 0, len * sizeof(*decoded));
+        return 0;
+    }
+
+    if (mode > FF_ARRAY_ELEMS(xcodes))
+        return AVERROR_INVALIDDATA;
+    code = xcodes[mode - 1];
+
+    for (i = 0; i < len; i++) {
+        int x = get_bits_long(gb, code.init);
+        if (x >= code.escape && get_bits1(gb)) {
+            x |= 1 << code.init;
+            if (x >= code.aescape) {
+                int scale = get_unary(gb, 1, 9);
+                if (scale == 9) {
+                    int scale_bits = get_bits(gb, 3);
+                    if (scale_bits > 0) {
+                        if (scale_bits == 7) {
+                            scale_bits += get_bits(gb, 5);
+                            if (scale_bits > 29)
+                                return AVERROR_INVALIDDATA;
+                        }
+                        scale = get_bits_long(gb, scale_bits) + 1;
+                        x    += code.scale * scale;
+                    }
+                    x += code.bias;
+                } else
+                    x += code.scale * scale - code.escape;
+            } else
+                x -= code.escape;
+        }
+        decoded[i] = (x >> 1) ^ -(x & 1);
+    }
+
+    return 0;
+}
+
+static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
+{
+    GetBitContext *gb = &s->gb;
+    int i, mode, ret;
+
+    if (length > s->nb_samples)
+        return AVERROR_INVALIDDATA;
+
+    if (get_bits1(gb)) {
+        int wlength, rval;
+        int coding_mode[128];
+
+        wlength = length / s->uval;
+
+        rval = length - (wlength * s->uval);
+
+        if (rval < s->uval / 2)
+            rval += s->uval;
+        else
+            wlength++;
+
+        if (wlength <= 1 || wlength > 128)
+            return AVERROR_INVALIDDATA;
+
+        coding_mode[0] = mode = get_bits(gb, 6);
+
+        for (i = 1; i < wlength; i++) {
+            int c = get_unary(gb, 1, 6);
+
+            switch (c) {
+            case 6:
+                mode = get_bits(gb, 6);
+                break;
+            case 5:
+            case 4:
+            case 3: {
+                /* mode += sign ? (1 - c) : (c - 1) */
+                int sign = get_bits1(gb);
+                mode    += (-sign ^ (c - 1)) + sign;
+                break;
+            }
+            case 2:
+                mode++;
+                break;
+            case 1:
+                mode--;
+                break;
+            }
+            coding_mode[i] = mode;
+        }
+
+        i = 0;
+        while (i < wlength) {
+            int len = 0;
+
+            mode = coding_mode[i];
+            do {
+                if (i >= wlength - 1)
+                    len += rval;
+                else
+                    len += s->uval;
+                i++;
+
+                if (i == wlength)
+                    break;
+            } while (coding_mode[i] == mode);
+
+            if ((ret = decode_segment(gb, mode, decoded, len)) < 0)
+                return ret;
+            decoded += len;
+        }
+    } else {
+        mode = get_bits(gb, 6);
+        if ((ret = decode_segment(gb, mode, decoded, length)) < 0)
+            return ret;
+    }
+
+    return 0;
+}
+
+static int get_bits_esc4(GetBitContext *gb)
+{
+    if (get_bits1(gb))
+        return get_bits(gb, 4) + 1;
+    else
+        return 0;
+}
+
+static void decode_filter_coeffs(TAKDecContext *s, int filter_order, int size,
+                                 int filter_quant, int16_t *filter)
+{
+    GetBitContext *gb = &s->gb;
+    int i, j, a, b;
+    int filter_tmp[MAX_PREDICTORS];
+    int16_t predictors[MAX_PREDICTORS];
+
+    predictors[0] = get_sbits(gb, 10);
+    predictors[1] = get_sbits(gb, 10);
+    predictors[2] = get_sbits(gb, size) << (10 - size);
+    predictors[3] = get_sbits(gb, size) << (10 - size);
+    if (filter_order > 4) {
+        int av_uninit(code_size);
+        int code_size_base = size - get_bits1(gb);
+
+        for (i = 4; i < filter_order; i++) {
+            if (!(i & 3))
+                code_size = code_size_base - get_bits(gb, 2);
+            predictors[i] = get_sbits(gb, code_size) << (10 - size);
+        }
+    }
+
+    filter_tmp[0] = predictors[0] << 6;
+    for (i = 1; i < filter_order; i++) {
+        int *p1 = &filter_tmp[0];
+        int *p2 = &filter_tmp[i - 1];
+
+        for (j = 0; j < (i + 1) / 2; j++) {
+            int tmp = *p1 + (predictors[i] * *p2 + 256 >> 9);
+            *p2     = *p2 + (predictors[i] * *p1 + 256 >> 9);
+            *p1     = tmp;
+            p1++;
+            p2--;
+        }
+
+        filter_tmp[i] = predictors[i] << 6;
+    }
+
+    a = 1 << (32 - (15 - filter_quant));
+    b = 1 << ((15 - filter_quant) - 1);
+    for (i = 0, j = filter_order - 1; i < filter_order / 2; i++, j--) {
+        filter[j] = a - ((filter_tmp[i] + b) >> (15 - filter_quant));
+        filter[i] = a - ((filter_tmp[j] + b) >> (15 - filter_quant));
+    }
+}
+
+static int decode_subframe(TAKDecContext *s, int32_t *decoded,
+                           int subframe_size, int prev_subframe_size)
+{
+    LOCAL_ALIGNED_16(int16_t, filter, [MAX_PREDICTORS]) = { 0, };
+    GetBitContext *gb = &s->gb;
+    int i, ret;
+    int dshift, size, filter_quant, filter_order;
+
+    if (!get_bits1(gb))
+        return decode_residues(s, decoded, subframe_size);
+
+    filter_order = predictor_sizes[get_bits(gb, 4)];
+
+    if (prev_subframe_size > 0 && get_bits1(gb)) {
+        if (filter_order > prev_subframe_size)
+            return AVERROR_INVALIDDATA;
+
+        decoded       -= filter_order;
+        subframe_size += filter_order;
+
+        if (filter_order > subframe_size)
+            return AVERROR_INVALIDDATA;
+    } else {
+        int lpc_mode;
+
+        if (filter_order > subframe_size)
+            return AVERROR_INVALIDDATA;
+
+        lpc_mode = get_bits(gb, 2);
+        if (lpc_mode > 2)
+            return AVERROR_INVALIDDATA;
+
+        if ((ret = decode_residues(s, decoded, filter_order)) < 0)
+            return ret;
+
+        if (lpc_mode)
+            decode_lpc(decoded, lpc_mode, filter_order);
+    }
+
+    dshift = get_bits_esc4(gb);
+    size   = get_bits1(gb) + 6;
+
+    filter_quant = 10;
+    if (get_bits1(gb)) {
+        filter_quant -= get_bits(gb, 3) + 1;
+        if (filter_quant < 3)
+            return AVERROR_INVALIDDATA;
+    }
+
+    decode_filter_coeffs(s, filter_order, size, filter_quant, filter);
+
+    if ((ret = decode_residues(s, &decoded[filter_order],
+                               subframe_size - filter_order)) < 0)
+        return ret;
+
+    av_fast_malloc(&s->residues, &s->residues_buf_size,
+                   FFALIGN(subframe_size + 16, 16) * sizeof(*s->residues));
+    if (!s->residues)
+        return AVERROR(ENOMEM);
+    memset(s->residues, 0, s->residues_buf_size);
+
+    for (i = 0; i < filter_order; i++)
+        s->residues[i] = *decoded++ >> dshift;
+
+    for (i = 0; i < subframe_size - filter_order; i++) {
+        int v = 1 << (filter_quant - 1);
+
+        v += s->dsp.scalarproduct_int16(&s->residues[i], filter,
+                                        FFALIGN(filter_order, 16));
+
+        v = (av_clip(v >> filter_quant, -8192, 8191) << dshift) - *decoded;
+        *decoded++ = v;
+        s->residues[filter_order + i] = v >> dshift;
+    }
+
+    emms_c();
+
+    return 0;
+}
+
+static int decode_channel(TAKDecContext *s, int chan)
+{
+    AVCodecContext *avctx = s->avctx;
+    GetBitContext *gb     = &s->gb;
+    int32_t *decoded      = s->decoded[chan];
+    int left              = s->nb_samples - 1;
+    int i, prev, ret, nb_subframes;
+    int subframe_len[MAX_SUBFRAMES];
+
+    s->sample_shift[chan] = get_bits_esc4(gb);
+    if (s->sample_shift[chan] >= avctx->bits_per_coded_sample)
+        return AVERROR_INVALIDDATA;
+
+    /* NOTE: TAK 2.2.0 appears to set the sample value to 0 if
+     *       bits_per_coded_sample - sample_shift is 1, but this produces
+     *       non-bit-exact output. Reading the 1 bit using get_sbits() instead
+     *       of skipping it produces bit-exact output. This has been reported
+     *       to the TAK author. */
+    *decoded++        = get_sbits(gb,
+                                  avctx->bits_per_coded_sample -
+                                  s->sample_shift[chan]);
+    s->lpc_mode[chan] = get_bits(gb, 2);
+    nb_subframes      = get_bits(gb, 3) + 1;
+
+    i = 0;
+    if (nb_subframes > 1) {
+        if (get_bits_left(gb) < (nb_subframes - 1) * 6)
+            return AVERROR_INVALIDDATA;
+
+        prev = 0;
+        for (; i < nb_subframes - 1; i++) {
+            int subframe_end = get_bits(gb, 6) * s->subframe_scale;
+            if (subframe_end <= prev)
+                return AVERROR_INVALIDDATA;
+            subframe_len[i] = subframe_end - prev;
+            left           -= subframe_len[i];
+            prev            = subframe_end;
+        }
+
+        if (left <= 0)
+            return AVERROR_INVALIDDATA;
+    }
+    subframe_len[i] = left;
+
+    prev = 0;
+    for (i = 0; i < nb_subframes; i++) {
+        if ((ret = decode_subframe(s, decoded, subframe_len[i], prev)) < 0)
+            return ret;
+        decoded += subframe_len[i];
+        prev     = subframe_len[i];
+    }
+
+    return 0;
+}
+
+static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
+{
+    GetBitContext *gb = &s->gb;
+    int32_t *p1       = s->decoded[c1] + 1;
+    int32_t *p2       = s->decoded[c2] + 1;
+    int i;
+    int dshift, dfactor;
+
+    switch (s->dmode) {
+    case 1: /* left/side */
+        for (i = 0; i < length; i++) {
+            int32_t a = p1[i];
+            int32_t b = p2[i];
+            p2[i]     = a + b;
+        }
+        break;
+    case 2: /* side/right */
+        for (i = 0; i < length; i++) {
+            int32_t a = p1[i];
+            int32_t b = p2[i];
+            p1[i]     = b - a;
+        }
+        break;
+    case 3: /* side/mid */
+        for (i = 0; i < length; i++) {
+            int32_t a = p1[i];
+            int32_t b = p2[i];
+            a        -= b >> 1;
+            p1[i]     = a;
+            p2[i]     = a + b;
+        }
+        break;
+    case 4: /* side/left with scale factor */
+        FFSWAP(int32_t*, p1, p2);
+    case 5: /* side/right with scale factor */
+        dshift  = get_bits_esc4(gb);
+        dfactor = get_sbits(gb, 10);
+        for (i = 0; i < length; i++) {
+            int32_t a = p1[i];
+            int32_t b = p2[i];
+            b         = dfactor * (b >> dshift) + 128 >> 8 << dshift;
+            p1[i]     = b - a;
+        }
+        break;
+    case 6:
+        FFSWAP(int32_t*, p1, p2);
+    case 7: {
+        LOCAL_ALIGNED_16(int16_t, filter, [MAX_PREDICTORS]) = { 0 };
+        int length2, order_half, filter_order, dval1, dval2;
+        int av_uninit(code_size);
+
+        if (length < 256)
+            return AVERROR_INVALIDDATA;
+
+        dshift       = get_bits_esc4(gb);
+        filter_order = 8 << get_bits1(gb);
+        dval1        = get_bits1(gb);
+        dval2        = get_bits1(gb);
+
+        for (i = 0; i < filter_order; i++) {
+            if (!(i & 3))
+                code_size = 14 - get_bits(gb, 3);
+            filter[i] = get_sbits(gb, code_size);
+        }
+
+        order_half = filter_order / 2;
+        length2    = length - (filter_order - 1);
+
+        /* decorrelate beginning samples */
+        if (dval1) {
+            for (i = 0; i < order_half; i++) {
+                int32_t a = p1[i];
+                int32_t b = p2[i];
+                p1[i]     = a + b;
+            }
+        }
+
+        /* decorrelate ending samples */
+        if (dval2) {
+            for (i = length2 + order_half; i < length; i++) {
+                int32_t a = p1[i];
+                int32_t b = p2[i];
+                p1[i]     = a + b;
+            }
+        }
+
+        av_fast_malloc(&s->residues, &s->residues_buf_size,
+                       FFALIGN(length + 16, 16) * sizeof(*s->residues));
+        if (!s->residues)
+            return AVERROR(ENOMEM);
+        memset(s->residues, 0, s->residues_buf_size);
+
+        for (i = 0; i < length; i++)
+            s->residues[i] = p2[i] >> dshift;
+
+        p1 += order_half;
+
+        for (i = 0; i < length2; i++) {
+            int v = 1 << 9;
+
+            v += s->dsp.scalarproduct_int16(&s->residues[i], filter,
+                                            FFALIGN(filter_order, 16));
+
+            p1[i] = (av_clip(v >> 10, -8192, 8191) << dshift) - p1[i];
+        }
+
+        emms_c();
+        break;
+    }
+    }
+
+    return 0;
+}
+
+static int tak_decode_frame(AVCodecContext *avctx, void *data,
+                            int *got_frame_ptr, AVPacket *pkt)
+{
+    TAKDecContext *s  = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    int chan, i, ret, hsize;
+
+    if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
+        return AVERROR_INVALIDDATA;
+
+    init_get_bits(gb, pkt->data, pkt->size * 8);
+
+    if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
+        return ret;
+
+    if (s->ti.flags & TAK_FRAME_FLAG_HAS_METADATA) {
+        av_log_missing_feature(avctx, "frame metadata", 1);
+        return AVERROR_PATCHWELCOME;
+    }
+
+    hsize = get_bits_count(gb) / 8;
+    if (avctx->err_recognition & AV_EF_CRCCHECK) {
+        if (ff_tak_check_crc(pkt->data, hsize)) {
+            av_log(avctx, AV_LOG_ERROR, "CRC error\n");
+            return AVERROR_INVALIDDATA;
+        }
+    }
+
+    if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
+        s->ti.codec != TAK_CODEC_MULTICHANNEL) {
+        av_log(avctx, AV_LOG_ERROR, "unsupported codec: %d\n", s->ti.codec);
+        return AVERROR_PATCHWELCOME;
+    }
+    if (s->ti.data_type) {
+        av_log(avctx, AV_LOG_ERROR,
+               "unsupported data type: %d\n", s->ti.data_type);
+        return AVERROR_INVALIDDATA;
+    }
+    if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
+        av_log(avctx, AV_LOG_ERROR,
+               "invalid number of channels: %d\n", s->ti.channels);
+        return AVERROR_INVALIDDATA;
+    }
+    if (s->ti.channels > 6) {
+        av_log(avctx, AV_LOG_ERROR,
+               "unsupported number of channels: %d\n", s->ti.channels);
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (s->ti.frame_samples <= 0) {
+        av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (s->ti.bps != avctx->bits_per_coded_sample) {
+        avctx->bits_per_coded_sample = s->ti.bps;
+        if ((ret = set_bps_params(avctx)) < 0)
+            return ret;
+    }
+    if (s->ti.sample_rate != avctx->sample_rate) {
+        avctx->sample_rate = s->ti.sample_rate;
+        set_sample_rate_params(avctx);
+    }
+    if (s->ti.ch_layout)
+        avctx->channel_layout = s->ti.ch_layout;
+    avctx->channels = s->ti.channels;
+
+    s->nb_samples = s->ti.last_frame_samples ? s->ti.last_frame_samples
+                                             : s->ti.frame_samples;
+
+    s->frame.nb_samples = s->nb_samples;
+    if ((ret = ff_get_buffer(avctx, &s->frame)) < 0)
+        return ret;
+
+    if (avctx->bits_per_coded_sample <= 16) {
+        int buf_size = av_samples_get_buffer_size(NULL, avctx->channels,
+                                                  s->nb_samples,
+                                                  AV_SAMPLE_FMT_S32P, 0);
+        av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
+        if (!s->decode_buffer)
+            return AVERROR(ENOMEM);
+        ret = av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
+                                     s->decode_buffer, avctx->channels,
+                                     s->nb_samples, AV_SAMPLE_FMT_S32P, 0);
+        if (ret < 0)
+            return ret;
+    } else {
+        for (chan = 0; chan < avctx->channels; chan++)
+            s->decoded[chan] = (int32_t *)s->frame.extended_data[chan];
+    }
+
+    if (s->nb_samples < 16) {
+        for (chan = 0; chan < avctx->channels; chan++) {
+            int32_t *decoded = s->decoded[chan];
+            for (i = 0; i < s->nb_samples; i++)
+                decoded[i] = get_sbits(gb, avctx->bits_per_coded_sample);
+        }
+    } else {
+        if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
+            for (chan = 0; chan < avctx->channels; chan++)
+                if (ret = decode_channel(s, chan))
+                    return ret;
+
+            if (avctx->channels == 2) {
+                if (get_bits1(gb)) {
+                    // some kind of subframe length, but it seems to be unused
+                    skip_bits(gb, 6);
+                }
+
+                s->dmode = get_bits(gb, 3);
+                if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
+                    return ret;
+            }
+        } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
+            if (get_bits1(gb)) {
+                int ch_mask = 0;
+
+                chan = get_bits(gb, 4) + 1;
+                if (chan > avctx->channels)
+                    return AVERROR_INVALIDDATA;
+
+                for (i = 0; i < chan; i++) {
+                    int nbit = get_bits(gb, 4);
+
+                    if (nbit >= avctx->channels)
+                        return AVERROR_INVALIDDATA;
+
+                    if (ch_mask & 1 << nbit)
+                        return AVERROR_INVALIDDATA;
+
+                    s->mcdparams[i].present = get_bits1(gb);
+                    if (s->mcdparams[i].present) {
+                        s->mcdparams[i].index = get_bits(gb, 2);
+                        s->mcdparams[i].chan2 = get_bits(gb, 4);
+                        if (s->mcdparams[i].index == 1) {
+                            if ((nbit == s->mcdparams[i].chan2) ||
+                                (ch_mask & 1 << s->mcdparams[i].chan2))
+                                return AVERROR_INVALIDDATA;
+
+                            ch_mask |= 1 << s->mcdparams[i].chan2;
+                        } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
+                            return AVERROR_INVALIDDATA;
+                        }
+                    }
+                    s->mcdparams[i].chan1 = nbit;
+
+                    ch_mask |= 1 << nbit;
+                }
+            } else {
+                chan = avctx->channels;
+                for (i = 0; i < chan; i++) {
+                    s->mcdparams[i].present = 0;
+                    s->mcdparams[i].chan1   = i;
+                }
+            }
+
+            for (i = 0; i < chan; i++) {
+                if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
+                    if (ret = decode_channel(s, s->mcdparams[i].chan2))
+                        return ret;
+
+                if (ret = decode_channel(s, s->mcdparams[i].chan1))
+                    return ret;
+
+                if (s->mcdparams[i].present) {
+                    s->dmode = mc_dmodes[s->mcdparams[i].index];
+                    if (ret = decorrelate(s,
+                                          s->mcdparams[i].chan2,
+                                          s->mcdparams[i].chan1,
+                                          s->nb_samples - 1))
+                        return ret;
+                }
+            }
+        }
+
+        for (chan = 0; chan < avctx->channels; chan++) {
+            int32_t *decoded = s->decoded[chan];
+
+            if (s->lpc_mode[chan])
+                decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
+
+            if (s->sample_shift[chan] > 0)
+                for (i = 0; i < s->nb_samples; i++)
+                    decoded[i] <<= s->sample_shift[chan];
+        }
+    }
+
+    align_get_bits(gb);
+    skip_bits(gb, 24);
+    if (get_bits_left(gb) < 0)
+        av_log(avctx, AV_LOG_DEBUG, "overread\n");
+    else if (get_bits_left(gb) > 0)
+        av_log(avctx, AV_LOG_DEBUG, "underread\n");
+
+    if (avctx->err_recognition & AV_EF_CRCCHECK) {
+        if (ff_tak_check_crc(pkt->data + hsize,
+                             get_bits_count(gb) / 8 - hsize)) {
+            av_log(avctx, AV_LOG_ERROR, "CRC error\n");
+            return AVERROR_INVALIDDATA;
+        }
+    }
+
+    /* convert to output buffer */
+    switch (avctx->sample_fmt) {
+    case AV_SAMPLE_FMT_U8P:
+        for (chan = 0; chan < avctx->channels; chan++) {
+            uint8_t *samples = (uint8_t *)s->frame.extended_data[chan];
+            int32_t *decoded = s->decoded[chan];
+            for (i = 0; i < s->nb_samples; i++)
+                samples[i] = decoded[i] + 0x80;
+        }
+        break;
+    case AV_SAMPLE_FMT_S16P:
+        for (chan = 0; chan < avctx->channels; chan++) {
+            int16_t *samples = (int16_t *)s->frame.extended_data[chan];
+            int32_t *decoded = s->decoded[chan];
+            for (i = 0; i < s->nb_samples; i++)
+                samples[i] = decoded[i];
+        }
+        break;
+    case AV_SAMPLE_FMT_S32P:
+        for (chan = 0; chan < avctx->channels; chan++) {
+            int32_t *samples = (int32_t *)s->frame.extended_data[chan];
+            for (i = 0; i < s->nb_samples; i++)
+                samples[i] <<= 8;
+        }
+        break;
+    }
+
+    *got_frame_ptr   = 1;
+    *(AVFrame *)data = s->frame;
+
+    return pkt->size;
+}
+
+static av_cold int tak_decode_close(AVCodecContext *avctx)
+{
+    TAKDecContext *s = avctx->priv_data;
+
+    av_freep(&s->decode_buffer);
+    av_freep(&s->residues);
+
+    return 0;
+}
+
+AVCodec ff_tak_decoder = {
+    .name             = "tak",
+    .type             = AVMEDIA_TYPE_AUDIO,
+    .id               = AV_CODEC_ID_TAK,
+    .priv_data_size   = sizeof(TAKDecContext),
+    .init             = tak_decode_init,
+    .init_static_data = tak_init_static_data,
+    .close            = tak_decode_close,
+    .decode           = tak_decode_frame,
+    .capabilities     = CODEC_CAP_DR1,
+    .long_name        = NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
+    .sample_fmts      = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
+                                                        AV_SAMPLE_FMT_S16P,
+                                                        AV_SAMPLE_FMT_S32P,
+                                                        AV_SAMPLE_FMT_NONE },
+};
index 45ff507..d170f67 100644 (file)
@@ -27,7 +27,7 @@
  */
 
 #define LIBAVCODEC_VERSION_MAJOR 54
-#define LIBAVCODEC_VERSION_MINOR 33
+#define LIBAVCODEC_VERSION_MINOR 34
 #define LIBAVCODEC_VERSION_MICRO  0
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
index 346045a..fa9366d 100644 (file)
@@ -304,6 +304,7 @@ OBJS-$(CONFIG_SRT_MUXER)                 += rawenc.o
 OBJS-$(CONFIG_STR_DEMUXER)               += psxstr.o
 OBJS-$(CONFIG_SWF_DEMUXER)               += swfdec.o swf.o
 OBJS-$(CONFIG_SWF_MUXER)                 += swfenc.o swf.o
+OBJS-$(CONFIG_TAK_DEMUXER)               += takdec.o apetag.o img2.o rawdec.o
 OBJS-$(CONFIG_THP_DEMUXER)               += thp.o
 OBJS-$(CONFIG_TIERTEXSEQ_DEMUXER)        += tiertexseq.o
 OBJS-$(CONFIG_TMV_DEMUXER)               += tmv.o
index c7c4be9..042d3b9 100644 (file)
@@ -214,6 +214,7 @@ void av_register_all(void)
     REGISTER_MUXDEMUX (SRT, srt);
     REGISTER_DEMUXER  (STR, str);
     REGISTER_MUXDEMUX (SWF, swf);
+    REGISTER_DEMUXER  (TAK, tak);
     REGISTER_MUXER    (TG2, tg2);
     REGISTER_MUXER    (TGP, tgp);
     REGISTER_DEMUXER  (THP, thp);
diff --git a/libavformat/takdec.c b/libavformat/takdec.c
new file mode 100644 (file)
index 0000000..13bc49b
--- /dev/null
@@ -0,0 +1,185 @@
+/*
+ * Raw TAK demuxer
+ * Copyright (c) 2012 Paul B Mahol
+ *
+ * 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 "libavcodec/tak.h"
+#include "avformat.h"
+#include "internal.h"
+#include "rawdec.h"
+#include "apetag.h"
+
+typedef struct TAKDemuxContext {
+    int     mlast_frame;
+    int64_t data_end;
+} TAKDemuxContext;
+
+static int tak_probe(AVProbeData *p)
+{
+    if (!memcmp(p->buf, "tBaK", 4))
+        return AVPROBE_SCORE_MAX / 2;
+    return 0;
+}
+
+static int tak_read_header(AVFormatContext *s)
+{
+    TAKDemuxContext *tc = s->priv_data;
+    AVIOContext *pb     = s->pb;
+    GetBitContext gb;
+    AVStream *st;
+    uint8_t *buffer = NULL;
+    int ret;
+
+    st = avformat_new_stream(s, 0);
+    if (!st)
+        return AVERROR(ENOMEM);
+
+    st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
+    st->codec->codec_id   = AV_CODEC_ID_TAK;
+    st->need_parsing      = AVSTREAM_PARSE_FULL;
+
+    tc->mlast_frame = 0;
+    if (avio_rl32(pb) != MKTAG('t', 'B', 'a', 'K')) {
+        avio_seek(pb, -4, SEEK_CUR);
+        return 0;
+    }
+
+    while (!pb->eof_reached) {
+        enum TAKMetaDataType type;
+        int size;
+
+        type = avio_r8(pb) & 0x7f;
+        size = avio_rl24(pb);
+
+        switch (type) {
+        case TAK_METADATA_STREAMINFO:
+        case TAK_METADATA_LAST_FRAME:
+        case TAK_METADATA_ENCODER:
+            buffer = av_malloc(size + FF_INPUT_BUFFER_PADDING_SIZE);
+            if (!buffer)
+                return AVERROR(ENOMEM);
+
+            if (avio_read(pb, buffer, size) != size) {
+                av_freep(&buffer);
+                return AVERROR(EIO);
+            }
+
+            init_get_bits(&gb, buffer, size * 8);
+            break;
+        case TAK_METADATA_MD5: {
+            uint8_t md5[16];
+            int i;
+
+            if (size != 19)
+                return AVERROR_INVALIDDATA;
+            avio_read(pb, md5, 16);
+            avio_skip(pb, 3);
+            av_log(s, AV_LOG_VERBOSE, "MD5=");
+            for (i = 0; i < 16; i++)
+                av_log(s, AV_LOG_VERBOSE, "%02x", md5[i]);
+            av_log(s, AV_LOG_VERBOSE, "\n");
+            break;
+        }
+        case TAK_METADATA_END: {
+            int64_t curpos = avio_tell(pb);
+
+            if (pb->seekable) {
+                ff_ape_parse_tag(s);
+                avio_seek(pb, curpos, SEEK_SET);
+            }
+
+            tc->data_end += curpos;
+            return 0;
+        }
+        default:
+            ret = avio_skip(pb, size);
+            if (ret < 0)
+                return ret;
+        }
+
+        if (type == TAK_METADATA_STREAMINFO) {
+            TAKStreamInfo ti;
+
+            avpriv_tak_parse_streaminfo(&gb, &ti);
+            if (ti.samples > 0)
+                st->duration = ti.samples;
+            st->codec->bits_per_coded_sample = ti.bps;
+            if (ti.ch_layout)
+                st->codec->channel_layout = ti.ch_layout;
+            st->codec->sample_rate           = ti.sample_rate;
+            st->codec->channels              = ti.channels;
+            st->start_time                   = 0;
+            avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
+            st->codec->extradata             = buffer;
+            st->codec->extradata_size        = size;
+            buffer                           = NULL;
+        } else if (type == TAK_METADATA_LAST_FRAME) {
+            if (size != 11)
+                return AVERROR_INVALIDDATA;
+            tc->mlast_frame = 1;
+            tc->data_end    = get_bits64(&gb, TAK_LAST_FRAME_POS_BITS) +
+                              get_bits(&gb, TAK_LAST_FRAME_SIZE_BITS);
+            av_freep(&buffer);
+        } else if (type == TAK_METADATA_ENCODER) {
+            av_log(s, AV_LOG_VERBOSE, "encoder version: %0X\n",
+                   get_bits_long(&gb, TAK_ENCODER_VERSION_BITS));
+            av_freep(&buffer);
+        }
+    }
+
+    return AVERROR_EOF;
+}
+
+static int raw_read_packet(AVFormatContext *s, AVPacket *pkt)
+{
+    TAKDemuxContext *tc = s->priv_data;
+    int ret;
+
+    if (tc->mlast_frame) {
+        AVIOContext *pb = s->pb;
+        int64_t size, left;
+
+        left = tc->data_end - avio_tell(s->pb);
+        size = FFMIN(left, 1024);
+        if (size <= 0)
+            return AVERROR_EOF;
+
+        ret = av_get_packet(pb, pkt, size);
+        if (ret < 0)
+            return ret;
+
+        pkt->stream_index = 0;
+    } else {
+        ret = ff_raw_read_partial_packet(s, pkt);
+    }
+
+    return ret;
+}
+
+AVInputFormat ff_tak_demuxer = {
+    .name           = "tak",
+    .long_name      = NULL_IF_CONFIG_SMALL("raw TAK"),
+    .priv_data_size = sizeof(TAKDemuxContext),
+    .read_probe     = tak_probe,
+    .read_header    = tak_read_header,
+    .read_packet    = raw_read_packet,
+    .flags          = AVFMT_GENERIC_INDEX,
+    .extensions     = "tak",
+    .raw_codec_id   = AV_CODEC_ID_TAK,
+};
index 034fa0a..349ba80 100644 (file)
@@ -30,7 +30,7 @@
 #include "libavutil/avutil.h"
 
 #define LIBAVFORMAT_VERSION_MAJOR 54
-#define LIBAVFORMAT_VERSION_MINOR 19
+#define LIBAVFORMAT_VERSION_MINOR 20
 #define LIBAVFORMAT_VERSION_MICRO  0
 
 #define LIBAVFORMAT_VERSION_INT AV_VERSION_INT(LIBAVFORMAT_VERSION_MAJOR, \