]> git.sesse.net Git - ffmpeg/blobdiff - libavcodec/binkaudio.c
lavf/qsvvpp: bypass vpp if not needed.
[ffmpeg] / libavcodec / binkaudio.c
index d917e7a12c5279124dce89a08d39d549832891da..51fb6c83cee0c8a11b52020fccc26386700ce28d 100644 (file)
  *  http://wiki.multimedia.cx/index.php?title=Bink_Audio
  */
 
+#include "libavutil/channel_layout.h"
+#include "libavutil/intfloat.h"
+
+#define BITSTREAM_READER_LE
 #include "avcodec.h"
-#define ALT_BITSTREAM_READER_LE
-#include "get_bits.h"
-#include "dsputil.h"
+#include "bitstream.h"
 #include "dct.h"
+#include "decode.h"
+#include "internal.h"
 #include "rdft.h"
-#include "fmtconvert.h"
-#include "libavutil/intfloat_readwrite.h"
-
-extern const uint16_t ff_wma_critical_freqs[25];
+#include "wma_freqs.h"
 
 static float quant_table[96];
 
 #define MAX_CHANNELS 2
 #define BINK_BLOCK_MAX_SIZE (MAX_CHANNELS << 11)
 
-typedef struct {
-    GetBitContext gb;
-    DSPContext dsp;
-    FmtConvertContext fmt_conv;
+typedef struct BinkAudioContext {
+    BitstreamContext bc;
     int version_b;          ///< Bink version 'b'
     int first;
     int channels;
@@ -58,11 +57,8 @@ typedef struct {
     unsigned int *bands;
     float root;
     DECLARE_ALIGNED(32, FFTSample, coeffs)[BINK_BLOCK_MAX_SIZE];
-    DECLARE_ALIGNED(16, int16_t, previous)[BINK_BLOCK_MAX_SIZE / 16];  ///< coeffs from previous audio block
-    DECLARE_ALIGNED(16, int16_t, current)[BINK_BLOCK_MAX_SIZE / 16];
-    float *coeffs_ptr[MAX_CHANNELS]; ///< pointers to the coeffs arrays for float_to_int16_interleave
-    float *prev_ptr[MAX_CHANNELS];   ///< pointers to the overlap points in the coeffs array
-    uint8_t *packet_buffer;
+    float previous[MAX_CHANNELS][BINK_BLOCK_MAX_SIZE / 16];  ///< coeffs from previous audio block
+    AVPacket *pkt;
     union {
         RDFTContext rdft;
         DCTContext dct;
@@ -78,9 +74,6 @@ static av_cold int decode_init(AVCodecContext *avctx)
     int i;
     int frame_len_bits;
 
-    dsputil_init(&s->dsp, avctx);
-    ff_fmt_convert_init(&s->fmt_conv, avctx);
-
     /* determine frame length */
     if (avctx->sample_rate < 22050) {
         frame_len_bits = 9;
@@ -94,24 +87,31 @@ static av_cold int decode_init(AVCodecContext *avctx)
         av_log(avctx, AV_LOG_ERROR, "too many channels: %d\n", avctx->channels);
         return -1;
     }
+    avctx->channel_layout = avctx->channels == 1 ? AV_CH_LAYOUT_MONO :
+                                                   AV_CH_LAYOUT_STEREO;
 
     s->version_b = avctx->extradata && avctx->extradata[3] == 'b';
 
-    if (avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT) {
+    if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT) {
         // audio is already interleaved for the RDFT format variant
+        avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
         sample_rate  *= avctx->channels;
         s->channels = 1;
         if (!s->version_b)
             frame_len_bits += av_log2(avctx->channels);
     } else {
         s->channels = avctx->channels;
+        avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
     }
 
     s->frame_len     = 1 << frame_len_bits;
     s->overlap_len   = s->frame_len / 16;
     s->block_size    = (s->frame_len - s->overlap_len) * s->channels;
     sample_rate_half = (sample_rate + 1) / 2;
-    s->root          = 2.0 / sqrt(s->frame_len);
+    if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
+        s->root = 2.0 / (sqrt(s->frame_len) * 32768.0);
+    else
+        s->root = s->frame_len / (sqrt(s->frame_len) * 32768.0);
     for (i = 0; i < 96; i++) {
         /* constant is result of 0.066399999/log10(M_E) */
         quant_table[i] = expf(i * 0.15289164787221953823f) * s->root;
@@ -133,28 +133,26 @@ static av_cold int decode_init(AVCodecContext *avctx)
     s->bands[s->num_bands] = s->frame_len;
 
     s->first = 1;
-    avctx->sample_fmt = AV_SAMPLE_FMT_S16;
-
-    for (i = 0; i < s->channels; i++) {
-        s->coeffs_ptr[i] = s->coeffs + i * s->frame_len;
-        s->prev_ptr[i]   = s->coeffs_ptr[i] + s->frame_len - s->overlap_len;
-    }
 
-    if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT)
+    if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
         ff_rdft_init(&s->trans.rdft, frame_len_bits, DFT_C2R);
     else if (CONFIG_BINKAUDIO_DCT_DECODER)
         ff_dct_init(&s->trans.dct, frame_len_bits, DCT_III);
     else
         return -1;
 
+    s->pkt = av_packet_alloc();
+    if (!s->pkt)
+        return AVERROR(ENOMEM);
+
     return 0;
 }
 
-static float get_float(GetBitContext *gb)
+static float get_float(BitstreamContext *bc)
 {
-    int power = get_bits(gb, 5);
-    float f = ldexpf(get_bits_long(gb, 23), power - 23);
-    if (get_bits1(gb))
+    int power = bitstream_read(bc, 5);
+    float f = ldexpf(bitstream_read(bc, 23), power - 23);
+    if (bitstream_read_bit(bc))
         f = -f;
     return f;
 }
@@ -163,45 +161,40 @@ static const uint8_t rle_length_tab[16] = {
     2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 32, 64
 };
 
-#define GET_BITS_SAFE(out, nbits) do {  \
-    if (get_bits_left(gb) < nbits)      \
-        return AVERROR_INVALIDDATA;     \
-    out = get_bits(gb, nbits);          \
-} while (0)
-
 /**
  * Decode Bink Audio block
  * @param[out] out Output buffer (must contain s->block_size elements)
  * @return 0 on success, negative error code on failure
  */
-static int decode_block(BinkAudioContext *s, int16_t *out, int use_dct)
+static int decode_block(BinkAudioContext *s, float **out, int use_dct)
 {
     int ch, i, j, k;
     float q, quant[25];
     int width, coeff;
-    GetBitContext *gb = &s->gb;
+    BitstreamContext *bc = &s->bc;
 
     if (use_dct)
-        skip_bits(gb, 2);
+        bitstream_skip(bc, 2);
 
     for (ch = 0; ch < s->channels; ch++) {
-        FFTSample *coeffs = s->coeffs_ptr[ch];
+        FFTSample *coeffs = out[ch];
+
         if (s->version_b) {
-            if (get_bits_left(gb) < 64)
+            if (bitstream_bits_left(bc) < 64)
                 return AVERROR_INVALIDDATA;
-            coeffs[0] = av_int2flt(get_bits(gb, 32)) * s->root;
-            coeffs[1] = av_int2flt(get_bits(gb, 32)) * s->root;
+            coeffs[0] = av_int2float(bitstream_read(bc, 32)) * s->root;
+            coeffs[1] = av_int2float(bitstream_read(bc, 32)) * s->root;
         } else {
-            if (get_bits_left(gb) < 58)
+            if (bitstream_bits_left(bc) < 58)
                 return AVERROR_INVALIDDATA;
-            coeffs[0] = get_float(gb) * s->root;
-            coeffs[1] = get_float(gb) * s->root;
+            coeffs[0] = get_float(bc) * s->root;
+            coeffs[1] = get_float(bc) * s->root;
         }
 
-        if (get_bits_left(gb) < s->num_bands * 8)
+        if (bitstream_bits_left(bc) < s->num_bands * 8)
             return AVERROR_INVALIDDATA;
         for (i = 0; i < s->num_bands; i++) {
-            int value = get_bits(gb, 8);
+            int value = bitstream_read(bc, 8);
             quant[i]  = quant_table[FFMIN(value, 95)];
         }
 
@@ -214,10 +207,9 @@ static int decode_block(BinkAudioContext *s, int16_t *out, int use_dct)
             if (s->version_b) {
                 j = i + 16;
             } else {
-                int v;
-                GET_BITS_SAFE(v, 1);
+                int v = bitstream_read_bit(bc);
                 if (v) {
-                    GET_BITS_SAFE(v, 4);
+                    v = bitstream_read(bc, 4);
                     j = i + rle_length_tab[v] * 8;
                 } else {
                     j = i + 8;
@@ -226,7 +218,7 @@ static int decode_block(BinkAudioContext *s, int16_t *out, int use_dct)
 
             j = FFMIN(j, s->frame_len);
 
-            GET_BITS_SAFE(width, 4);
+            width = bitstream_read(bc, 4);
             if (width == 0) {
                 memset(coeffs + i, 0, (j - i) * sizeof(*coeffs));
                 i = j;
@@ -236,10 +228,10 @@ static int decode_block(BinkAudioContext *s, int16_t *out, int use_dct)
                 while (i < j) {
                     if (s->bands[k] == i)
                         q = quant[k++];
-                    GET_BITS_SAFE(coeff, width);
+                    coeff = bitstream_read(bc, width);
                     if (coeff) {
                         int v;
-                        GET_BITS_SAFE(v, 1);
+                        v = bitstream_read_bit(bc);
                         if (v)
                             coeffs[i] = -q * coeff;
                         else
@@ -255,30 +247,24 @@ static int decode_block(BinkAudioContext *s, int16_t *out, int use_dct)
         if (CONFIG_BINKAUDIO_DCT_DECODER && use_dct) {
             coeffs[0] /= 0.5;
             s->trans.dct.dct_calc(&s->trans.dct,  coeffs);
-            s->dsp.vector_fmul_scalar(coeffs, coeffs, s->frame_len / 2, s->frame_len);
         }
         else if (CONFIG_BINKAUDIO_RDFT_DECODER)
             s->trans.rdft.rdft_calc(&s->trans.rdft, coeffs);
     }
 
-    s->fmt_conv.float_to_int16_interleave(s->current,
-                                          (const float **)s->prev_ptr,
-                                          s->overlap_len, s->channels);
-    s->fmt_conv.float_to_int16_interleave(out, (const float **)s->coeffs_ptr,
-                                          s->frame_len - s->overlap_len,
-                                          s->channels);
-
-    if (!s->first) {
+    for (ch = 0; ch < s->channels; ch++) {
+        int j;
         int count = s->overlap_len * s->channels;
-        int shift = av_log2(count);
-        for (i = 0; i < count; i++) {
-            out[i] = (s->previous[i] * (count - i) + out[i] * i) >> shift;
+        if (!s->first) {
+            j = ch;
+            for (i = 0; i < s->overlap_len; i++, j += s->channels)
+                out[ch][i] = (s->previous[ch][i] * (count - j) +
+                                      out[ch][i] *          j) / count;
         }
+        memcpy(s->previous[ch], &out[ch][s->frame_len - s->overlap_len],
+               s->overlap_len * sizeof(*s->previous[ch]));
     }
 
-    memcpy(s->previous, s->current,
-           s->overlap_len * s->channels * sizeof(*s->previous));
-
     s->first = 0;
 
     return 0;
@@ -288,88 +274,94 @@ static av_cold int decode_end(AVCodecContext *avctx)
 {
     BinkAudioContext * s = avctx->priv_data;
     av_freep(&s->bands);
-    av_freep(&s->packet_buffer);
-    if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT)
+    if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
         ff_rdft_end(&s->trans.rdft);
     else if (CONFIG_BINKAUDIO_DCT_DECODER)
         ff_dct_end(&s->trans.dct);
+
+    av_packet_free(&s->pkt);
+
     return 0;
 }
 
-static void get_bits_align32(GetBitContext *s)
+static void get_bits_align32(BitstreamContext *s)
 {
-    int n = (-get_bits_count(s)) & 31;
-    if (n) skip_bits(s, n);
+    int n = (-bitstream_tell(s)) & 31;
+    if (n)
+        bitstream_skip(s, n);
 }
 
-static int decode_frame(AVCodecContext *avctx,
-                        void *data, int *data_size,
-                        AVPacket *avpkt)
+static int binkaudio_receive_frame(AVCodecContext *avctx, AVFrame *frame)
 {
     BinkAudioContext *s = avctx->priv_data;
-    int16_t *samples      = data;
-    GetBitContext *gb = &s->gb;
-    int out_size, consumed = 0;
-
-    if (!get_bits_left(gb)) {
-        uint8_t *buf;
-        /* handle end-of-stream */
-        if (!avpkt->size) {
-            *data_size = 0;
-            return 0;
-        }
-        if (avpkt->size < 4) {
+    BitstreamContext *bc = &s->bc;
+    int ret;
+
+    if (!s->pkt->data) {
+        ret = ff_decode_get_packet(avctx, s->pkt);
+        if (ret < 0)
+            return ret;
+
+        if (s->pkt->size < 4) {
             av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
-            return AVERROR_INVALIDDATA;
+            ret = AVERROR_INVALIDDATA;
+            goto fail;
         }
-        buf = av_realloc(s->packet_buffer, avpkt->size + FF_INPUT_BUFFER_PADDING_SIZE);
-        if (!buf)
-            return AVERROR(ENOMEM);
-        s->packet_buffer = buf;
-        memcpy(s->packet_buffer, avpkt->data, avpkt->size);
-        init_get_bits(gb, s->packet_buffer, avpkt->size * 8);
-        consumed = avpkt->size;
+
+        ret = bitstream_init8(bc, s->pkt->data, s->pkt->size);
+        if (ret < 0)
+            goto fail;
 
         /* skip reported size */
-        skip_bits_long(gb, 32);
+        bitstream_skip(bc, 32);
     }
 
-    out_size = s->block_size * av_get_bytes_per_sample(avctx->sample_fmt);
-    if (*data_size < out_size) {
-        av_log(avctx, AV_LOG_ERROR, "Output buffer is too small\n");
-        return AVERROR(EINVAL);
+    /* get output buffer */
+    frame->nb_samples = s->frame_len;
+    if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
+        av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+        return ret;
     }
 
-    if (decode_block(s, samples, avctx->codec->id == CODEC_ID_BINKAUDIO_DCT)) {
+    if (decode_block(s, (float **)frame->extended_data,
+                     avctx->codec->id == AV_CODEC_ID_BINKAUDIO_DCT)) {
         av_log(avctx, AV_LOG_ERROR, "Incomplete packet\n");
         return AVERROR_INVALIDDATA;
     }
-    get_bits_align32(gb);
+    get_bits_align32(bc);
+    if (!bitstream_bits_left(bc)) {
+        memset(bc, 0, sizeof(*bc));
+        av_packet_unref(s->pkt);
+    }
 
-    *data_size = out_size;
-    return consumed;
+    frame->nb_samples = s->block_size / avctx->channels;
+
+    return 0;
+fail:
+    av_packet_unref(s->pkt);
+    return ret;
 }
 
 AVCodec ff_binkaudio_rdft_decoder = {
     .name           = "binkaudio_rdft",
+    .long_name      = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)"),
     .type           = AVMEDIA_TYPE_AUDIO,
-    .id             = CODEC_ID_BINKAUDIO_RDFT,
+    .id             = AV_CODEC_ID_BINKAUDIO_RDFT,
     .priv_data_size = sizeof(BinkAudioContext),
     .init           = decode_init,
     .close          = decode_end,
-    .decode         = decode_frame,
-    .capabilities   = CODEC_CAP_DELAY,
-    .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)")
+    .receive_frame  = binkaudio_receive_frame,
+    .capabilities   = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_DR1,
 };
 
 AVCodec ff_binkaudio_dct_decoder = {
     .name           = "binkaudio_dct",
+    .long_name      = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)"),
     .type           = AVMEDIA_TYPE_AUDIO,
-    .id             = CODEC_ID_BINKAUDIO_DCT,
+    .id             = AV_CODEC_ID_BINKAUDIO_DCT,
     .priv_data_size = sizeof(BinkAudioContext),
     .init           = decode_init,
     .close          = decode_end,
-    .decode         = decode_frame,
-    .capabilities   = CODEC_CAP_DELAY,
-    .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)")
+    .receive_frame  = binkaudio_receive_frame,
+    .capabilities   = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_DR1,
 };