* TwinVQ decoder
* Copyright (c) 2009 Vitor Sessak
*
- * 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
*/
+#include "libavutil/float_dsp.h"
#include "avcodec.h"
#include "get_bits.h"
#include "dsputil.h"
+#include "fft.h"
+#include "lsp.h"
+#include "sinewin.h"
#include <math.h>
#include <stdint.h>
typedef struct TwinContext {
AVCodecContext *avctx;
+ AVFrame frame;
DSPContext dsp;
+ AVFloatDSPContext fdsp;
FFTContext mdct_ctx[3];
const ModeTab *mtab;
float *curr_frame; ///< non-interleaved output
float *prev_frame; ///< non-interleaved previous frame
int last_block_pos[2];
+ int discarded_packets;
float *cos_tabs[3];
} TwinContext;
#define PPC_SHAPE_CB_SIZE 64
+#define PPC_SHAPE_LEN_MAX 60
#define SUB_AMP_MAX 4500.0
#define MULAW_MU 100.0
#define GAIN_BITS 8
#define SUB_GAIN_BITS 5
#define WINDOW_TYPE_BITS 4
#define PGAIN_MU 200
+#define LSP_COEFS_MAX 20
+#define LSP_SPLIT_MAX 4
+#define CHANNELS_MAX 2
+#define SUBBLOCKS_MAX 16
+#define BARK_N_COEF_MAX 4
/** @note not speed critical, hence not optimized */
static void memset_float(float *buf, float val, int size)
* be a multiple of four.
* @return the LPC value
*
- * @todo reuse code from vorbis_dec.c: vorbis_floor0_decode
+ * @todo reuse code from Vorbis decoder: vorbis_floor0_decode
*/
static float eval_lpc_spectrum(const float *lsp, float cos_val, int order)
{
}
/**
- * Evaluates the LPC amplitude spectrum envelope from the line spectrum pairs.
+ * Evaluate the LPC amplitude spectrum envelope from the line spectrum pairs.
*/
static void eval_lpcenv(TwinContext *tctx, const float *cos_vals, float *lpc)
{
}
/**
- * Evaluates the LPC amplitude spectrum envelope from the line spectrum pairs.
+ * Evaluate the LPC amplitude spectrum envelope from the line spectrum pairs.
* Probably for speed reasons, the coefficients are evaluated as
* siiiibiiiisiiiibiiiisiiiibiiiisiiiibiiiis ...
* where s is an evaluated value, i is a value interpolated from the others
* a*b == 200 and the nearest integer is ill-defined, use a table to emulate
* the following broken float-based implementation used by the binary decoder:
*
- * \code
+ * @code
* static int very_broken_op(int a, int b)
* {
* static float test; // Ugh, force gcc to do the division first...
* test = a/400.;
* return b * test + 0.5;
* }
- * \endcode
+ * @endcode
*
* @note if this function is replaced by just ROUNDED_DIV(a*b,400.), the stddev
* between the original file (before encoding with Yamaha encoder) and the
}
}
-static void bubblesort(float *lsp, int lp_order)
-{
- int i,j;
-
- /* sort lsp in ascending order. float bubble agorithm,
- O(n) if data already sorted, O(n^2) - otherwise */
- for (i = 0; i < lp_order - 1; i++)
- for (j = i; j >= 0 && lsp[j] > lsp[j+1]; j--)
- FFSWAP(float, lsp[j], lsp[j+1]);
-}
-
static void decode_lsp(TwinContext *tctx, int lpc_idx1, uint8_t *lpc_idx2,
int lpc_hist_idx, float *lsp, float *hist)
{
rearrange_lsp(mtab->n_lsp, lsp, 0.0001);
rearrange_lsp(mtab->n_lsp, lsp, 0.000095);
- bubblesort(lsp, mtab->n_lsp);
+ ff_sort_nearly_sorted_floats(lsp, mtab->n_lsp);
}
static void dec_lpc_spectrum_inv(TwinContext *tctx, float *lsp,
static void imdct_and_window(TwinContext *tctx, enum FrameType ftype, int wtype,
float *in, float *prev, int ch)
{
+ FFTContext *mdct = &tctx->mdct_ctx[ftype];
const ModeTab *mtab = tctx->mtab;
int bsize = mtab->size / mtab->fmode[ftype].sub;
int size = mtab->size;
wsize = types_sizes[wtype_to_wsize[sub_wtype]];
- ff_imdct_half(&tctx->mdct_ctx[ftype], buf1 + bsize*j, in + bsize*j);
+ mdct->imdct_half(mdct, buf1 + bsize*j, in + bsize*j);
tctx->dsp.vector_fmul_window(out2,
prev_buf + (bsize-wsize)/2,
buf1 + bsize*j,
ff_sine_windows[av_log2(wsize)],
- 0.0,
wsize/2);
out2 += wsize;
float *out)
{
const ModeTab *mtab = tctx->mtab;
+ int size1, size2;
float *prev_buf = tctx->prev_frame + tctx->last_block_pos[0];
- int i, j;
+ int i;
for (i = 0; i < tctx->avctx->channels; i++) {
imdct_and_window(tctx, ftype, wtype,
i);
}
+ if (!out)
+ return;
+
+ size2 = tctx->last_block_pos[0];
+ size1 = mtab->size - size2;
if (tctx->avctx->channels == 2) {
- for (i = 0; i < mtab->size - tctx->last_block_pos[0]; i++) {
- float f1 = prev_buf[ i];
- float f2 = prev_buf[2*mtab->size + i];
- out[2*i ] = f1 + f2;
- out[2*i + 1] = f1 - f2;
- }
- for (j = 0; i < mtab->size; j++,i++) {
- float f1 = tctx->curr_frame[ j];
- float f2 = tctx->curr_frame[2*mtab->size + j];
- out[2*i ] = f1 + f2;
- out[2*i + 1] = f1 - f2;
- }
+ tctx->dsp.butterflies_float_interleave(out, prev_buf,
+ &prev_buf[2*mtab->size],
+ size1);
+
+ out += 2 * size1;
+
+ tctx->dsp.butterflies_float_interleave(out, tctx->curr_frame,
+ &tctx->curr_frame[2*mtab->size],
+ size2);
} else {
- memcpy(out, prev_buf,
- (mtab->size - tctx->last_block_pos[0]) * sizeof(*out));
+ memcpy(out, prev_buf, size1 * sizeof(*out));
- out += mtab->size - tctx->last_block_pos[0];
+ out += size1;
- memcpy(out, tctx->curr_frame,
- (tctx->last_block_pos[0]) * sizeof(*out));
+ memcpy(out, tctx->curr_frame, size2 * sizeof(*out));
}
}
int channels = tctx->avctx->channels;
int sub = mtab->fmode[ftype].sub;
int block_size = mtab->size / sub;
- float gain[channels*sub];
- float ppc_shape[mtab->ppc_shape_len * channels * 4];
- uint8_t bark1[channels][sub][mtab->fmode[ftype].bark_n_coef];
- uint8_t bark_use_hist[channels][sub];
+ float gain[CHANNELS_MAX*SUBBLOCKS_MAX];
+ float ppc_shape[PPC_SHAPE_LEN_MAX * CHANNELS_MAX * 4];
+ uint8_t bark1[CHANNELS_MAX][SUBBLOCKS_MAX][BARK_N_COEF_MAX];
+ uint8_t bark_use_hist[CHANNELS_MAX][SUBBLOCKS_MAX];
- uint8_t lpc_idx1[channels];
- uint8_t lpc_idx2[channels][tctx->mtab->lsp_split];
- uint8_t lpc_hist_idx[channels];
+ uint8_t lpc_idx1[CHANNELS_MAX];
+ uint8_t lpc_idx2[CHANNELS_MAX][LSP_SPLIT_MAX];
+ uint8_t lpc_hist_idx[CHANNELS_MAX];
int i, j, k;
for (i = 0; i < channels; i++) {
float *chunk = out + mtab->size * i;
- float lsp[tctx->mtab->n_lsp];
+ float lsp[LSP_COEFS_MAX];
for (j = 0; j < sub; j++) {
dec_bark_env(tctx, bark1[i][j], bark_use_hist[i][j], i,
tctx->tmp_buf, gain[sub*i+j], ftype);
- tctx->dsp.vector_fmul(chunk + block_size*j, tctx->tmp_buf,
- block_size);
+ tctx->fdsp.vector_fmul(chunk + block_size*j, chunk + block_size*j,
+ tctx->tmp_buf, block_size);
}
dec_lpc_spectrum_inv(tctx, lsp, ftype, tctx->tmp_buf);
for (j = 0; j < mtab->fmode[ftype].sub; j++) {
- tctx->dsp.vector_fmul(chunk, tctx->tmp_buf, block_size);
+ tctx->fdsp.vector_fmul(chunk, chunk, tctx->tmp_buf, block_size);
chunk += block_size;
}
}
}
static int twin_decode_frame(AVCodecContext * avctx, void *data,
- int *data_size, AVPacket *avpkt)
+ int *got_frame_ptr, AVPacket *avpkt)
{
const uint8_t *buf = avpkt->data;
int buf_size = avpkt->size;
TwinContext *tctx = avctx->priv_data;
GetBitContext gb;
const ModeTab *mtab = tctx->mtab;
- float *out = data;
+ float *out = NULL;
enum FrameType ftype;
- int window_type;
+ int window_type, ret;
static const enum FrameType wtype_to_ftype_table[] = {
FT_LONG, FT_LONG, FT_SHORT, FT_LONG,
FT_MEDIUM, FT_LONG, FT_LONG, FT_MEDIUM, FT_MEDIUM
if (buf_size*8 < avctx->bit_rate*mtab->size/avctx->sample_rate + 8) {
av_log(avctx, AV_LOG_ERROR,
"Frame too small (%d bytes). Truncated file?\n", buf_size);
- *data_size = 0;
- return buf_size;
+ return AVERROR(EINVAL);
+ }
+
+ /* get output buffer */
+ if (tctx->discarded_packets >= 2) {
+ tctx->frame.nb_samples = mtab->size;
+ if ((ret = avctx->get_buffer(avctx, &tctx->frame)) < 0) {
+ av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+ return ret;
+ }
+ out = (float *)tctx->frame.data[0];
}
init_get_bits(&gb, buf, buf_size * 8);
FFSWAP(float*, tctx->curr_frame, tctx->prev_frame);
- if (tctx->avctx->frame_number < 2) {
- *data_size=0;
+ if (tctx->discarded_packets < 2) {
+ tctx->discarded_packets++;
+ *got_frame_ptr = 0;
return buf_size;
}
- tctx->dsp.vector_clipf(out, out, -32700./(1<<15), 32700./(1<<15),
- avctx->channels * mtab->size);
-
- *data_size = mtab->size*avctx->channels*4;
+ *got_frame_ptr = 1;
+ *(AVFrame *)data = tctx->frame;;
return buf_size;
}
/**
* Init IMDCT and windowing tables
*/
-static av_cold void init_mdct_win(TwinContext *tctx)
+static av_cold int init_mdct_win(TwinContext *tctx)
{
- int i,j;
+ int i, j, ret;
const ModeTab *mtab = tctx->mtab;
int size_s = mtab->size / mtab->fmode[FT_SHORT].sub;
int size_m = mtab->size / mtab->fmode[FT_MEDIUM].sub;
for (i = 0; i < 3; i++) {
int bsize = tctx->mtab->size/tctx->mtab->fmode[i].sub;
- ff_mdct_init(&tctx->mdct_ctx[i], av_log2(bsize) + 1, 1,
- -sqrt(norm/bsize) / (1<<15));
+ if ((ret = ff_mdct_init(&tctx->mdct_ctx[i], av_log2(bsize) + 1, 1,
+ -sqrt(norm/bsize) / (1<<15))))
+ return ret;
}
- tctx->tmp_buf = av_malloc(mtab->size * sizeof(*tctx->tmp_buf));
+ FF_ALLOC_OR_GOTO(tctx->avctx, tctx->tmp_buf,
+ mtab->size * sizeof(*tctx->tmp_buf), alloc_fail);
- tctx->spectrum = av_malloc(2*mtab->size*channels*sizeof(float));
- tctx->curr_frame = av_malloc(2*mtab->size*channels*sizeof(float));
- tctx->prev_frame = av_malloc(2*mtab->size*channels*sizeof(float));
+ FF_ALLOC_OR_GOTO(tctx->avctx, tctx->spectrum,
+ 2 * mtab->size * channels * sizeof(*tctx->spectrum),
+ alloc_fail);
+ FF_ALLOC_OR_GOTO(tctx->avctx, tctx->curr_frame,
+ 2 * mtab->size * channels * sizeof(*tctx->curr_frame),
+ alloc_fail);
+ FF_ALLOC_OR_GOTO(tctx->avctx, tctx->prev_frame,
+ 2 * mtab->size * channels * sizeof(*tctx->prev_frame),
+ alloc_fail);
for (i = 0; i < 3; i++) {
int m = 4*mtab->size/mtab->fmode[i].sub;
double freq = 2*M_PI/m;
- tctx->cos_tabs[i] = av_malloc((m/4)*sizeof(*tctx->cos_tabs));
+ FF_ALLOC_OR_GOTO(tctx->avctx, tctx->cos_tabs[i],
+ (m / 4) * sizeof(*tctx->cos_tabs[i]), alloc_fail);
for (j = 0; j <= m/8; j++)
tctx->cos_tabs[i][j] = cos((2*j + 1)*freq);
}
- ff_sine_window_init(ff_sine_windows[av_log2(size_m) ], size_m );
- ff_sine_window_init(ff_sine_windows[av_log2(size_s/2) ], size_s/2);
- ff_sine_window_init(ff_sine_windows[av_log2(mtab->size)], mtab->size);
+ ff_init_ff_sine_windows(av_log2(size_m));
+ ff_init_ff_sine_windows(av_log2(size_s/2));
+ ff_init_ff_sine_windows(av_log2(mtab->size));
+
+ return 0;
+alloc_fail:
+ return AVERROR(ENOMEM);
}
/**
/**
* Interpret the input data as in the following table:
*
- * \verbatim
+ * @verbatim
*
* abcdefgh
* ijklmnop
* qrstuvw
* x123456
*
- * \endverbatim
+ * @endverbatim
*
* and transpose it, giving the output
* aiqxbjr1cks2dlt3emu4fvn5gow6hp
{
int block_size;
const ModeTab *mtab = tctx->mtab;
- int size = tctx->avctx->channels*mtab->fmode[ftype].sub;
+ int size;
int16_t *tmp_perm = (int16_t *) tctx->tmp_buf;
if (ftype == FT_PPC) {
size = tctx->avctx->channels;
block_size = mtab->ppc_shape_len;
- } else
+ } else {
+ size = tctx->avctx->channels * mtab->fmode[ftype].sub;
block_size = mtab->size / mtab->fmode[ftype].sub;
+ }
permutate_in_line(tmp_perm, tctx->n_div[ftype], size,
block_size, tctx->length[ftype],
construct_perm_table(tctx, frametype);
}
+static av_cold int twin_decode_close(AVCodecContext *avctx)
+{
+ TwinContext *tctx = avctx->priv_data;
+ int i;
+
+ for (i = 0; i < 3; i++) {
+ ff_mdct_end(&tctx->mdct_ctx[i]);
+ av_free(tctx->cos_tabs[i]);
+ }
+
+
+ av_free(tctx->curr_frame);
+ av_free(tctx->spectrum);
+ av_free(tctx->prev_frame);
+ av_free(tctx->tmp_buf);
+
+ return 0;
+}
+
static av_cold int twin_decode_init(AVCodecContext *avctx)
{
+ int ret;
TwinContext *tctx = avctx->priv_data;
- int isampf = avctx->sample_rate/1000;
- int ibps = avctx->bit_rate/(1000 * avctx->channels);
+ int isampf, ibps;
tctx->avctx = avctx;
- avctx->sample_fmt = SAMPLE_FMT_FLT;
+ avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
+
+ if (!avctx->extradata || avctx->extradata_size < 12) {
+ av_log(avctx, AV_LOG_ERROR, "Missing or incomplete extradata\n");
+ return AVERROR_INVALIDDATA;
+ }
+ avctx->channels = AV_RB32(avctx->extradata ) + 1;
+ avctx->bit_rate = AV_RB32(avctx->extradata + 4) * 1000;
+ isampf = AV_RB32(avctx->extradata + 8);
+ switch (isampf) {
+ case 44: avctx->sample_rate = 44100; break;
+ case 22: avctx->sample_rate = 22050; break;
+ case 11: avctx->sample_rate = 11025; break;
+ default: avctx->sample_rate = isampf * 1000; break;
+ }
- if (avctx->channels > 2) {
+ if (avctx->channels > CHANNELS_MAX) {
av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %i\n",
avctx->channels);
return -1;
}
+ ibps = avctx->bit_rate / (1000 * avctx->channels);
switch ((isampf << 8) + ibps) {
case (8 <<8) + 8: tctx->mtab = &mode_08_08; break;
return -1;
}
- dsputil_init(&tctx->dsp, avctx);
- init_mdct_win(tctx);
+ ff_dsputil_init(&tctx->dsp, avctx);
+ avpriv_float_dsp_init(&tctx->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
+ if ((ret = init_mdct_win(tctx))) {
+ av_log(avctx, AV_LOG_ERROR, "Error initializing MDCT\n");
+ twin_decode_close(avctx);
+ return ret;
+ }
init_bitstream_params(tctx);
memset_float(tctx->bark_hist[0][0], 0.1, FF_ARRAY_ELEMS(tctx->bark_hist));
- return 0;
-}
-
-static av_cold int twin_decode_close(AVCodecContext *avctx)
-{
- TwinContext *tctx = avctx->priv_data;
- int i;
-
- for (i = 0; i < 3; i++) {
- ff_mdct_end(&tctx->mdct_ctx[i]);
- av_free(tctx->cos_tabs[i]);
- }
-
-
- av_free(tctx->curr_frame);
- av_free(tctx->spectrum);
- av_free(tctx->prev_frame);
- av_free(tctx->tmp_buf);
+ avcodec_get_frame_defaults(&tctx->frame);
+ avctx->coded_frame = &tctx->frame;
return 0;
}
-AVCodec twinvq_decoder =
-{
- "twinvq",
- CODEC_TYPE_AUDIO,
- CODEC_ID_TWINVQ,
- sizeof(TwinContext),
- twin_decode_init,
- NULL,
- twin_decode_close,
- twin_decode_frame,
- .long_name = NULL_IF_CONFIG_SMALL("VQF TwinVQ"),
+AVCodec ff_twinvq_decoder = {
+ .name = "twinvq",
+ .type = AVMEDIA_TYPE_AUDIO,
+ .id = AV_CODEC_ID_TWINVQ,
+ .priv_data_size = sizeof(TwinContext),
+ .init = twin_decode_init,
+ .close = twin_decode_close,
+ .decode = twin_decode_frame,
+ .capabilities = CODEC_CAP_DR1,
+ .long_name = NULL_IF_CONFIG_SMALL("VQF TwinVQ"),
};