X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fflacenc.c;h=d93bfb5be4be0294d52bcdfdaf6ff05099354554;hb=9f5ddd1e460a7e25b838286d36bf3fbe17baaa76;hp=7eabd9c994154fc5f4240858438dc5bff1ae1024;hpb=93d65e3d92623500b69fe7ccba43f6b6266c3250;p=ffmpeg

diff --git a/libavcodec/flacenc.c b/libavcodec/flacenc.c
index 7eabd9c9941..d93bfb5be4b 100644
--- a/libavcodec/flacenc.c
+++ b/libavcodec/flacenc.c
@@ -2,20 +2,20 @@
  * FLAC audio encoder
  * Copyright (c) 2006  Justin Ruggles <justin.ruggles@gmail.com>
  *
- * This file is part of FFmpeg.
+ * This file is part of Libav.
  *
- * FFmpeg is free software; you can redistribute it and/or
+ * 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.
  *
- * FFmpeg is distributed in the hope that it will be useful,
+ * 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 FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
@@ -23,7 +23,6 @@
 #include "libavutil/md5.h"
 #include "avcodec.h"
 #include "get_bits.h"
-#include "dsputil.h"
 #include "golomb.h"
 #include "lpc.h"
 #include "flac.h"
@@ -77,6 +76,7 @@ typedef struct FlacFrame {
     int bs_code[2];
     uint8_t crc8;
     int ch_mode;
+    int verbatim_only;
 } FlacFrame;
 
 typedef struct FlacEncodeContext {
@@ -94,7 +94,7 @@ typedef struct FlacEncodeContext {
     FlacFrame frame;
     CompressionOptions options;
     AVCodecContext *avctx;
-    DSPContext dsp;
+    LPCContext lpc_ctx;
     struct AVMD5 *md5ctx;
 } FlacEncodeContext;
 
@@ -211,14 +211,12 @@ static av_cold int flac_encode_init(AVCodecContext *avctx)
     int freq = avctx->sample_rate;
     int channels = avctx->channels;
     FlacEncodeContext *s = avctx->priv_data;
-    int i, level;
+    int i, level, ret;
     uint8_t *streaminfo;
 
     s->avctx = avctx;
 
-    dsputil_init(&s->dsp, avctx);
-
-    if (avctx->sample_fmt != SAMPLE_FMT_S16)
+    if (avctx->sample_fmt != AV_SAMPLE_FMT_S16)
         return -1;
 
     if (channels < 1 || channels > FLAC_MAX_CHANNELS)
@@ -287,7 +285,7 @@ static av_cold int flac_encode_init(AVCodecContext *avctx)
     s->options.max_partition_order = ((int[]){  2,  2,  3,  3,  3,  8,  8,  8,  8,  8,  8,  8,  8})[level];
 
     /* set compression option overrides from AVCodecContext */
-#if LIBAVCODEC_VERSION_MAJOR < 53
+#if FF_API_USE_LPC
     /* for compatibility with deprecated AVCodecContext.use_lpc */
     if (avctx->use_lpc == 0) {
         s->options.lpc_type = AV_LPC_TYPE_FIXED;
@@ -438,9 +436,12 @@ static av_cold int flac_encode_init(AVCodecContext *avctx)
     if (!avctx->coded_frame)
         return AVERROR(ENOMEM);
 
+    ret = ff_lpc_init(&s->lpc_ctx, avctx->frame_size,
+                      s->options.max_prediction_order, AV_LPC_TYPE_LEVINSON);
+
     dprint_compression_options(s);
 
-    return 0;
+    return ret;
 }
 
 
@@ -472,6 +473,8 @@ static void init_frame(FlacEncodeContext *s)
 
     for (ch = 0; ch < s->channels; ch++)
         frame->subframes[ch].obits = 16;
+
+    frame->verbatim_only = 0;
 }
 
 
@@ -490,6 +493,67 @@ static void copy_samples(FlacEncodeContext *s, const int16_t *samples)
 }
 
 
+static int rice_count_exact(int32_t *res, int n, int k)
+{
+    int i;
+    int count = 0;
+
+    for (i = 0; i < n; i++) {
+        int32_t v = -2 * res[i] - 1;
+        v ^= v >> 31;
+        count += (v >> k) + 1 + k;
+    }
+    return count;
+}
+
+
+static int subframe_count_exact(FlacEncodeContext *s, FlacSubframe *sub,
+                                int pred_order)
+{
+    int p, porder, psize;
+    int i, part_end;
+    int count = 0;
+
+    /* subframe header */
+    count += 8;
+
+    /* subframe */
+    if (sub->type == FLAC_SUBFRAME_CONSTANT) {
+        count += sub->obits;
+    } else if (sub->type == FLAC_SUBFRAME_VERBATIM) {
+        count += s->frame.blocksize * sub->obits;
+    } else {
+        /* warm-up samples */
+        count += pred_order * sub->obits;
+
+        /* LPC coefficients */
+        if (sub->type == FLAC_SUBFRAME_LPC)
+            count += 4 + 5 + pred_order * s->options.lpc_coeff_precision;
+
+        /* rice-encoded block */
+        count += 2;
+
+        /* partition order */
+        porder = sub->rc.porder;
+        psize  = s->frame.blocksize >> porder;
+        count += 4;
+
+        /* residual */
+        i        = pred_order;
+        part_end = psize;
+        for (p = 0; p < 1 << porder; p++) {
+            int k = sub->rc.params[p];
+            count += 4;
+            count += rice_count_exact(&sub->residual[i], part_end - i, k);
+            i = part_end;
+            part_end = FFMIN(s->frame.blocksize, part_end + psize);
+        }
+    }
+
+    return count;
+}
+
+
 #define rice_encode_count(sum, n, k) (((n)*((k)+1))+((sum-(n>>1))>>(k)))
 
 /**
@@ -603,39 +667,23 @@ static int get_max_p_order(int max_porder, int n, int order)
 }
 
 
-static uint32_t calc_rice_params_fixed(RiceContext *rc, int pmin, int pmax,
-                                       int32_t *data, int n, int pred_order,
-                                       int bps)
-{
-    uint32_t bits;
-    pmin  = get_max_p_order(pmin, n, pred_order);
-    pmax  = get_max_p_order(pmax, n, pred_order);
-    bits  = pred_order * bps + 6;
-    bits += calc_rice_params(rc, pmin, pmax, data, n, pred_order);
-    return bits;
-}
-
-
-static uint32_t calc_rice_params_lpc(RiceContext *rc, int pmin, int pmax,
-                                     int32_t *data, int n, int pred_order,
-                                     int bps, int precision)
+static uint32_t find_subframe_rice_params(FlacEncodeContext *s,
+                                          FlacSubframe *sub, int pred_order)
 {
-    uint32_t bits;
-    pmin  = get_max_p_order(pmin, n, pred_order);
-    pmax  = get_max_p_order(pmax, n, pred_order);
-    bits  = pred_order*bps + 4 + 5 + pred_order*precision + 6;
-    bits += calc_rice_params(rc, pmin, pmax, data, n, pred_order);
+    int pmin = get_max_p_order(s->options.min_partition_order,
+                               s->frame.blocksize, pred_order);
+    int pmax = get_max_p_order(s->options.max_partition_order,
+                               s->frame.blocksize, pred_order);
+
+    uint32_t bits = 8 + pred_order * sub->obits + 2 + 4;
+    if (sub->type == FLAC_SUBFRAME_LPC)
+        bits += 4 + 5 + pred_order * s->options.lpc_coeff_precision;
+    bits += calc_rice_params(&sub->rc, pmin, pmax, sub->residual,
+                             s->frame.blocksize, pred_order);
     return bits;
 }
 
 
-static void encode_residual_verbatim(int32_t *res, int32_t *smp, int n)
-{
-    assert(n > 0);
-    memcpy(res, smp, n * sizeof(int32_t));
-}
-
-
 static void encode_residual_fixed(int32_t *res, const int32_t *smp, int n,
                                   int order)
 {
@@ -794,8 +842,7 @@ static void encode_residual_lpc(int32_t *res, const int32_t *smp, int n,
 static int encode_residual_ch(FlacEncodeContext *s, int ch)
 {
     int i, n;
-    int min_order, max_order, opt_order, precision, omethod;
-    int min_porder, max_porder;
+    int min_order, max_order, opt_order, omethod;
     FlacFrame *frame;
     FlacSubframe *sub;
     int32_t coefs[MAX_LPC_ORDER][MAX_LPC_ORDER];
@@ -815,24 +862,22 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch)
     if (i == n) {
         sub->type = sub->type_code = FLAC_SUBFRAME_CONSTANT;
         res[0] = smp[0];
-        return sub->obits;
+        return subframe_count_exact(s, sub, 0);
     }
 
     /* VERBATIM */
-    if (n < 5) {
+    if (frame->verbatim_only || n < 5) {
         sub->type = sub->type_code = FLAC_SUBFRAME_VERBATIM;
-        encode_residual_verbatim(res, smp, n);
-        return sub->obits * n;
+        memcpy(res, smp, n * sizeof(int32_t));
+        return subframe_count_exact(s, sub, 0);
     }
 
     min_order  = s->options.min_prediction_order;
     max_order  = s->options.max_prediction_order;
-    min_porder = s->options.min_partition_order;
-    max_porder = s->options.max_partition_order;
-    precision  = s->options.lpc_coeff_precision;
     omethod    = s->options.prediction_order_method;
 
     /* FIXED */
+    sub->type = FLAC_SUBFRAME_FIXED;
     if (s->options.lpc_type == AV_LPC_TYPE_NONE  ||
         s->options.lpc_type == AV_LPC_TYPE_FIXED || n <= max_order) {
         uint32_t bits[MAX_FIXED_ORDER+1];
@@ -842,25 +887,23 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch)
         bits[0]   = UINT32_MAX;
         for (i = min_order; i <= max_order; i++) {
             encode_residual_fixed(res, smp, n, i);
-            bits[i] = calc_rice_params_fixed(&sub->rc, min_porder, max_porder, res,
-                                             n, i, sub->obits);
+            bits[i] = find_subframe_rice_params(s, sub, i);
             if (bits[i] < bits[opt_order])
                 opt_order = i;
         }
         sub->order     = opt_order;
-        sub->type      = FLAC_SUBFRAME_FIXED;
         sub->type_code = sub->type | sub->order;
         if (sub->order != max_order) {
             encode_residual_fixed(res, smp, n, sub->order);
-            return calc_rice_params_fixed(&sub->rc, min_porder, max_porder, res, n,
-                                          sub->order, sub->obits);
+            find_subframe_rice_params(s, sub, sub->order);
         }
-        return bits[sub->order];
+        return subframe_count_exact(s, sub, sub->order);
     }
 
     /* LPC */
-    opt_order = ff_lpc_calc_coefs(&s->dsp, smp, n, min_order, max_order,
-                                  precision, coefs, shift, s->options.lpc_type,
+    sub->type = FLAC_SUBFRAME_LPC;
+    opt_order = ff_lpc_calc_coefs(&s->lpc_ctx, smp, n, min_order, max_order,
+                                  s->options.lpc_coeff_precision, coefs, shift, s->options.lpc_type,
                                   s->options.lpc_passes, omethod,
                                   MAX_LPC_SHIFT, 0);
 
@@ -878,8 +921,7 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch)
             if (order < 0)
                 order = 0;
             encode_residual_lpc(res, smp, n, order+1, coefs[order], shift[order]);
-            bits[i] = calc_rice_params_lpc(&sub->rc, min_porder, max_porder,
-                                           res, n, order+1, sub->obits, precision);
+            bits[i] = find_subframe_rice_params(s, sub, order+1);
             if (bits[i] < bits[opt_index]) {
                 opt_index = i;
                 opt_order = order;
@@ -893,8 +935,7 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch)
         bits[0]   = UINT32_MAX;
         for (i = min_order-1; i < max_order; i++) {
             encode_residual_lpc(res, smp, n, i+1, coefs[i], shift[i]);
-            bits[i] = calc_rice_params_lpc(&sub->rc, min_porder, max_porder,
-                                           res, n, i+1, sub->obits, precision);
+            bits[i] = find_subframe_rice_params(s, sub, i+1);
             if (bits[i] < bits[opt_order])
                 opt_order = i;
         }
@@ -912,9 +953,7 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch)
                 if (i < min_order-1 || i >= max_order || bits[i] < UINT32_MAX)
                     continue;
                 encode_residual_lpc(res, smp, n, i+1, coefs[i], shift[i]);
-                bits[i] = calc_rice_params_lpc(&sub->rc, min_porder, max_porder,
-                                               res, n, i+1, sub->obits,
-                                               precision);
+                bits[i] = find_subframe_rice_params(s, sub, i+1);
                 if (bits[i] < bits[opt_order])
                     opt_order = i;
             }
@@ -923,7 +962,6 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch)
     }
 
     sub->order     = opt_order;
-    sub->type      = FLAC_SUBFRAME_LPC;
     sub->type_code = sub->type | (sub->order-1);
     sub->shift     = shift[sub->order-1];
     for (i = 0; i < sub->order; i++)
@@ -931,8 +969,9 @@ static int encode_residual_ch(FlacEncodeContext *s, int ch)
 
     encode_residual_lpc(res, smp, n, sub->order, sub->coefs, sub->shift);
 
-    return calc_rice_params_lpc(&sub->rc, min_porder, max_porder, res, n,
-                                sub->order, sub->obits, precision);
+    find_subframe_rice_params(s, sub, sub->order);
+
+    return subframe_count_exact(s, sub, sub->order);
 }
 
 
@@ -957,7 +996,10 @@ static int count_frame_header(FlacEncodeContext *s)
     PUT_UTF8(s->frame_count, tmp, count += 8;)
 
     /* explicit block size */
-    count += FFMAX(0, s->frame.bs_code[0] - 5) * 8;
+    if (s->frame.bs_code[0] == 6)
+        count += 8;
+    else if (s->frame.bs_code[0] == 7)
+        count += 16;
 
     /* explicit sample rate */
     count += ((s->sr_code[0] == 12) + (s->sr_code[0] > 12)) * 8;
@@ -985,36 +1027,6 @@ static int encode_frame(FlacEncodeContext *s)
 }
 
 
-static int encode_residual_v(FlacEncodeContext *s, int ch)
-{
-    int i, n;
-    FlacFrame *frame;
-    FlacSubframe *sub;
-    int32_t *res, *smp;
-
-    frame = &s->frame;
-    sub   = &frame->subframes[ch];
-    res   = sub->residual;
-    smp   = sub->samples;
-    n     = frame->blocksize;
-
-    /* CONSTANT */
-    for (i = 1; i < n; i++)
-        if (smp[i] != smp[0])
-            break;
-    if (i == n) {
-        sub->type = sub->type_code = FLAC_SUBFRAME_CONSTANT;
-        res[0]    = smp[0];
-        return sub->obits;
-    }
-
-    /* VERBATIM */
-    sub->type = sub->type_code = FLAC_SUBFRAME_VERBATIM;
-    encode_residual_verbatim(res, smp, n);
-    return sub->obits * n;
-}
-
-
 static int estimate_stereo_mode(int32_t *left_ch, int32_t *right_ch, int n)
 {
     int i, best;
@@ -1113,7 +1125,7 @@ static void write_utf8(PutBitContext *pb, uint32_t val)
 }
 
 
-static void output_frame_header(FlacEncodeContext *s)
+static void write_frame_header(FlacEncodeContext *s)
 {
     FlacFrame *frame;
     int crc;
@@ -1150,7 +1162,7 @@ static void output_frame_header(FlacEncodeContext *s)
 }
 
 
-static void output_subframes(FlacEncodeContext *s)
+static void write_subframes(FlacEncodeContext *s)
 {
     int ch;
 
@@ -1208,7 +1220,7 @@ static void output_subframes(FlacEncodeContext *s)
 }
 
 
-static void output_frame_footer(FlacEncodeContext *s)
+static void write_frame_footer(FlacEncodeContext *s)
 {
     int crc;
     flush_put_bits(&s->pb);
@@ -1219,6 +1231,16 @@ static void output_frame_footer(FlacEncodeContext *s)
 }
 
 
+static int write_frame(FlacEncodeContext *s, uint8_t *frame, int buf_size)
+{
+    init_put_bits(&s->pb, frame, buf_size);
+    write_frame_header(s);
+    write_subframes(s);
+    write_frame_footer(s);
+    return put_bits_count(&s->pb) >> 3;
+}
+
+
 static void update_md5_sum(FlacEncodeContext *s, const int16_t *samples)
 {
 #if HAVE_BIGENDIAN
@@ -1236,19 +1258,12 @@ static void update_md5_sum(FlacEncodeContext *s, const int16_t *samples)
 static int flac_encode_frame(AVCodecContext *avctx, uint8_t *frame,
                              int buf_size, void *data)
 {
-    int ch;
     FlacEncodeContext *s;
     const int16_t *samples = data;
-    int out_bytes;
-    int reencoded=0;
+    int frame_bytes, out_bytes;
 
     s = avctx->priv_data;
 
-    if (buf_size < s->max_framesize * 2) {
-        av_log(avctx, AV_LOG_ERROR, "output buffer too small\n");
-        return 0;
-    }
-
     /* when the last block is reached, update the header in extradata */
     if (!data) {
         s->max_framesize = s->max_encoded_framesize;
@@ -1257,34 +1272,32 @@ static int flac_encode_frame(AVCodecContext *avctx, uint8_t *frame,
         return 0;
     }
 
+    /* change max_framesize for small final frame */
+    if (avctx->frame_size < s->frame.blocksize) {
+        s->max_framesize = ff_flac_get_max_frame_size(avctx->frame_size,
+                                                      s->channels, 16);
+    }
+
     init_frame(s);
 
     copy_samples(s, samples);
 
     channel_decorrelation(s);
 
-    encode_frame(s);
+    frame_bytes = encode_frame(s);
 
-write_frame:
-    init_put_bits(&s->pb, frame, buf_size);
-    output_frame_header(s);
-    output_subframes(s);
-    output_frame_footer(s);
-    out_bytes = put_bits_count(&s->pb) >> 3;
-
-    if (out_bytes > s->max_framesize) {
-        if (reencoded) {
-            /* still too large. must be an error. */
-            av_log(avctx, AV_LOG_ERROR, "error encoding frame\n");
-            return -1;
-        }
+    /* fallback to verbatim mode if the compressed frame is larger than it
+       would be if encoded uncompressed. */
+    if (frame_bytes > s->max_framesize) {
+        s->frame.verbatim_only = 1;
+        frame_bytes = encode_frame(s);
+    }
 
-        /* frame too large. use verbatim mode */
-        for (ch = 0; ch < s->channels; ch++)
-            encode_residual_v(s, ch);
-        reencoded = 1;
-        goto write_frame;
+    if (buf_size < frame_bytes) {
+        av_log(avctx, AV_LOG_ERROR, "output buffer too small\n");
+        return 0;
     }
+    out_bytes = write_frame(s, frame, buf_size);
 
     s->frame_count++;
     avctx->coded_frame->pts = s->sample_count;
@@ -1304,6 +1317,7 @@ static av_cold int flac_encode_close(AVCodecContext *avctx)
     if (avctx->priv_data) {
         FlacEncodeContext *s = avctx->priv_data;
         av_freep(&s->md5ctx);
+        ff_lpc_end(&s->lpc_ctx);
     }
     av_freep(&avctx->extradata);
     avctx->extradata_size = 0;
@@ -1312,7 +1326,7 @@ static av_cold int flac_encode_close(AVCodecContext *avctx)
 }
 
 
-AVCodec flac_encoder = {
+AVCodec ff_flac_encoder = {
     "flac",
     AVMEDIA_TYPE_AUDIO,
     CODEC_ID_FLAC,
@@ -1322,6 +1336,6 @@ AVCodec flac_encoder = {
     flac_encode_close,
     NULL,
     .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY,
-    .sample_fmts = (const enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE},
+    .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
     .long_name = NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),
 };