har0ke/FFmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
FFmpeg/libavcodec/cook.c
Andreas Rheinhardt 790f793844 avutil/common: Don't auto-include mem.h 
There are lots of files that don't need it: The number of object
files that actually need it went down from 2011 to 884 here.

Keep it for external users in order to not cause breakages.

Also improve the other headers a bit while just at it.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-03-31 00:08:43 +01:00

1315 lines
45 KiB
C
Raw Blame History

/*
* COOK compatible decoder
* Copyright (c) 2003 Sascha Sommer
* Copyright (c) 2005 Benjamin Larsson
*
* This file is part of FFmpeg.
*
* FFmpeg 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,
* 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
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* Cook compatible decoder. Bastardization of the G.722.1 standard.
* This decoder handles RealNetworks, RealAudio G2 data.
* Cook is identified by the codec name cook in RM files.
*
* To use this decoder, a calling application must supply the extradata
* bytes provided from the RM container; 8+ bytes for mono streams and
* 16+ for stereo streams (maybe more).
*
* Codec technicalities (all this assume a buffer length of 1024):
* Cook works with several different techniques to achieve its compression.
* In the timedomain the buffer is divided into 8 pieces and quantized. If
* two neighboring pieces have different quantization index a smooth
* quantization curve is used to get a smooth overlap between the different
* pieces.
* To get to the transformdomain Cook uses a modulated lapped transform.
* The transform domain has 50 subbands with 20 elements each. This
* means only a maximum of 50*20=1000 coefficients are used out of the 1024
* available.
*/
#include "libavutil/channel_layout.h"
#include "libavutil/lfg.h"
#include "libavutil/mem.h"
#include "libavutil/mem_internal.h"
#include "libavutil/thread.h"
#include "libavutil/tx.h"
#include "audiodsp.h"
#include "avcodec.h"
#include "get_bits.h"
#include "bytestream.h"
#include "codec_internal.h"
#include "decode.h"
#include "sinewin.h"
#include "unary.h"
#include "cookdata.h"
/* the different Cook versions */
#define MONO 0x1000001
#define STEREO 0x1000002
#define JOINT_STEREO 0x1000003
#define MC_COOK 0x2000000
#define SUBBAND_SIZE 20
#define MAX_SUBPACKETS 5
#define QUANT_VLC_BITS 9
#define COUPLING_VLC_BITS 6
typedef struct cook_gains {
int *now;
int *previous;
} cook_gains;
typedef struct COOKSubpacket {
int ch_idx;
int size;
int num_channels;
int cookversion;
int subbands;
int js_subband_start;
int js_vlc_bits;
int samples_per_channel;
int log2_numvector_size;
unsigned int channel_mask;
VLC channel_coupling;
int joint_stereo;
int bits_per_subpacket;
int bits_per_subpdiv;
int total_subbands;
int numvector_size; // 1 << log2_numvector_size;
float mono_previous_buffer1[1024];
float mono_previous_buffer2[1024];
cook_gains gains1;
cook_gains gains2;
int gain_1[9];
int gain_2[9];
int gain_3[9];
int gain_4[9];
} COOKSubpacket;
typedef struct cook {
/*
* The following 5 functions provide the lowlevel arithmetic on
* the internal audio buffers.
*/
void (*scalar_dequant)(struct cook *q, int index, int quant_index,
int *subband_coef_index, int *subband_coef_sign,
float *mlt_p);
void (*decouple)(struct cook *q,
COOKSubpacket *p,
int subband,
float f1, float f2,
float *decode_buffer,
float *mlt_buffer1, float *mlt_buffer2);
void (*imlt_window)(struct cook *q, float *buffer1,
cook_gains *gains_ptr, float *previous_buffer);
void (*interpolate)(struct cook *q, float *buffer,
int gain_index, int gain_index_next);
void (*saturate_output)(struct cook *q, float *out);
AVCodecContext* avctx;
AudioDSPContext adsp;
GetBitContext gb;
/* stream data */
int num_vectors;
int samples_per_channel;
/* states */
AVLFG random_state;
int discarded_packets;
/* transform data */
AVTXContext *mdct_ctx;
av_tx_fn mdct_fn;
float* mlt_window;
/* VLC data */
VLC envelope_quant_index[13];
VLC sqvh[7]; // scalar quantization
/* generate tables and related variables */
int gain_size_factor;
float gain_table[31];
/* data buffers */
uint8_t* decoded_bytes_buffer;
DECLARE_ALIGNED(32, float, mono_mdct_output)[2048];
float decode_buffer_1[1024];
float decode_buffer_2[1024];
float decode_buffer_0[1060]; /* static allocation for joint decode */
const float *cplscales[5];
int num_subpackets;
COOKSubpacket subpacket[MAX_SUBPACKETS];
} COOKContext;
static float pow2tab[127];
static float rootpow2tab[127];
/*************** init functions ***************/
/* table generator */
static av_cold void init_pow2table(void)
{
/* fast way of computing 2^i and 2^(0.5*i) for -63 <= i < 64 */
int i;
static const float exp2_tab[2] = {1, M_SQRT2};
float exp2_val = powf(2, -63);
float root_val = powf(2, -32);
for (i = -63; i < 64; i++) {
if (!(i & 1))
root_val *= 2;
pow2tab[63 + i] = exp2_val;
rootpow2tab[63 + i] = root_val * exp2_tab[i & 1];
exp2_val *= 2;
}
}
/* table generator */
static av_cold void init_gain_table(COOKContext *q)
{
int i;
q->gain_size_factor = q->samples_per_channel / 8;
for (i = 0; i < 31; i++)
q->gain_table[i] = pow(pow2tab[i + 48],
(1.0 / (double) q->gain_size_factor));
}
static av_cold int build_vlc(VLC *vlc, int nb_bits, const uint8_t counts[16],
const void *syms, int symbol_size, int offset,
void *logctx)
{
uint8_t lens[MAX_COOK_VLC_ENTRIES];
unsigned num = 0;
for (int i = 0; i < 16; i++)
for (unsigned count = num + counts[i]; num < count; num++)
lens[num] = i + 1;
return ff_vlc_init_from_lengths(vlc, nb_bits, num, lens, 1,
syms, symbol_size, symbol_size,
offset, 0, logctx);
}
static av_cold int init_cook_vlc_tables(COOKContext *q)
{
int i, result;
result = 0;
for (i = 0; i < 13; i++) {
result |= build_vlc(&q->envelope_quant_index[i], QUANT_VLC_BITS,
envelope_quant_index_huffcounts[i],
envelope_quant_index_huffsyms[i], 1, -12, q->avctx);
}
av_log(q->avctx, AV_LOG_DEBUG, "sqvh VLC init\n");
for (i = 0; i < 7; i++) {
int sym_size = 1 + (i == 3);
result |= build_vlc(&q->sqvh[i], vhvlcsize_tab[i],
cvh_huffcounts[i],
cvh_huffsyms[i], sym_size, 0, q->avctx);
}
for (i = 0; i < q->num_subpackets; i++) {
if (q->subpacket[i].joint_stereo == 1) {
result |= build_vlc(&q->subpacket[i].channel_coupling, COUPLING_VLC_BITS,
ccpl_huffcounts[q->subpacket[i].js_vlc_bits - 2],
ccpl_huffsyms[q->subpacket[i].js_vlc_bits - 2], 1,
0, q->avctx);
av_log(q->avctx, AV_LOG_DEBUG, "subpacket %i Joint-stereo VLC used.\n", i);
}
}
av_log(q->avctx, AV_LOG_DEBUG, "VLC tables initialized.\n");
return result;
}
static av_cold int init_cook_mlt(COOKContext *q)
{
int j, ret;
int mlt_size = q->samples_per_channel;
const float scale = 1.0 / 32768.0;
if (!(q->mlt_window = av_malloc_array(mlt_size, sizeof(*q->mlt_window))))
return AVERROR(ENOMEM);
/* Initialize the MLT window: simple sine window. */
ff_sine_window_init(q->mlt_window, mlt_size);
for (j = 0; j < mlt_size; j++)
q->mlt_window[j] *= sqrt(2.0 / q->samples_per_channel);
/* Initialize the MDCT. */
ret = av_tx_init(&q->mdct_ctx, &q->mdct_fn, AV_TX_FLOAT_MDCT,
1, mlt_size, &scale, AV_TX_FULL_IMDCT);
if (ret < 0)
return ret;
return 0;
}
static av_cold void init_cplscales_table(COOKContext *q)
{
int i;
for (i = 0; i < 5; i++)
q->cplscales[i] = cplscales[i];
}
/*************** init functions end ***********/
#define DECODE_BYTES_PAD1(bytes) (3 - ((bytes) + 3) % 4)
#define DECODE_BYTES_PAD2(bytes) ((bytes) % 4 + DECODE_BYTES_PAD1(2 * (bytes)))
/**
* Cook indata decoding, every 32 bits are XORed with 0x37c511f2.
* Why? No idea, some checksum/error detection method maybe.
*
* Out buffer size: extra bytes are needed to cope with
* padding/misalignment.
* Subpackets passed to the decoder can contain two, consecutive
* half-subpackets, of identical but arbitrary size.
* 1234 1234 1234 1234 extraA extraB
* Case 1: AAAA BBBB 0 0
* Case 2: AAAA ABBB BB-- 3 3
* Case 3: AAAA AABB BBBB 2 2
* Case 4: AAAA AAAB BBBB BB-- 1 5
*
* Nice way to waste CPU cycles.
*
* @param inbuffer pointer to byte array of indata
* @param out pointer to byte array of outdata
* @param bytes number of bytes
*/
static inline int decode_bytes(const uint8_t *inbuffer, uint8_t *out, int bytes)
{
static const uint32_t tab[4] = {
AV_BE2NE32C(0x37c511f2u), AV_BE2NE32C(0xf237c511u),
AV_BE2NE32C(0x11f237c5u), AV_BE2NE32C(0xc511f237u),
};
int i, off;
uint32_t c;
const uint32_t *buf;
uint32_t *obuf = (uint32_t *) out;
/* FIXME: 64 bit platforms would be able to do 64 bits at a time.
* I'm too lazy though, should be something like
* for (i = 0; i < bitamount / 64; i++)
* (int64_t) out[i] = 0x37c511f237c511f2 ^ av_be2ne64(int64_t) in[i]);
* Buffer alignment needs to be checked. */
off = (intptr_t) inbuffer & 3;
buf = (const uint32_t *) (inbuffer - off);
c = tab[off];
bytes += 3 + off;
for (i = 0; i < bytes / 4; i++)
obuf[i] = c ^ buf[i];
return off;
}
static av_cold int cook_decode_close(AVCodecContext *avctx)
{
int i;
COOKContext *q = avctx->priv_data;
av_log(avctx, AV_LOG_DEBUG, "Deallocating memory.\n");
/* Free allocated memory buffers. */
av_freep(&q->mlt_window);
av_freep(&q->decoded_bytes_buffer);
/* Free the transform. */
av_tx_uninit(&q->mdct_ctx);
/* Free the VLC tables. */
for (i = 0; i < 13; i++)
ff_vlc_free(&q->envelope_quant_index[i]);
for (i = 0; i < 7; i++)
ff_vlc_free(&q->sqvh[i]);
for (i = 0; i < q->num_subpackets; i++)
ff_vlc_free(&q->subpacket[i].channel_coupling);
av_log(avctx, AV_LOG_DEBUG, "Memory deallocated.\n");
return 0;
}
/**
* Fill the gain array for the timedomain quantization.
*
* @param gb pointer to the GetBitContext
* @param gaininfo array[9] of gain indexes
*/
static void decode_gain_info(GetBitContext *gb, int *gaininfo)
{
int i, n;
n = get_unary(gb, 0, get_bits_left(gb)); // amount of elements*2 to update
i = 0;
while (n--) {
int index = get_bits(gb, 3);
int gain = get_bits1(gb) ? get_bits(gb, 4) - 7 : -1;
while (i <= index)
gaininfo[i++] = gain;
}
while (i <= 8)
gaininfo[i++] = 0;
}
/**
* Create the quant index table needed for the envelope.
*
* @param q pointer to the COOKContext
* @param quant_index_table pointer to the array
*/
static int decode_envelope(COOKContext *q, COOKSubpacket *p,
int *quant_index_table)
{
int i, j, vlc_index;
quant_index_table[0] = get_bits(&q->gb, 6) - 6; // This is used later in categorize
for (i = 1; i < p->total_subbands; i++) {
vlc_index = i;
if (i >= p->js_subband_start * 2) {
vlc_index -= p->js_subband_start;
} else {
vlc_index /= 2;
if (vlc_index < 1)
vlc_index = 1;
}
if (vlc_index > 13)
vlc_index = 13; // the VLC tables >13 are identical to No. 13
j = get_vlc2(&q->gb, q->envelope_quant_index[vlc_index - 1].table,
QUANT_VLC_BITS, 2);
quant_index_table[i] = quant_index_table[i - 1] + j; // differential encoding
if (quant_index_table[i] > 63 || quant_index_table[i] < -63) {
av_log(q->avctx, AV_LOG_ERROR,
"Invalid quantizer %d at position %d, outside [-63, 63] range\n",
quant_index_table[i], i);
return AVERROR_INVALIDDATA;
}
}
return 0;
}
/**
* Calculate the category and category_index vector.
*
* @param q pointer to the COOKContext
* @param quant_index_table pointer to the array
* @param category pointer to the category array
* @param category_index pointer to the category_index array
*/
static void categorize(COOKContext *q, COOKSubpacket *p, const int *quant_index_table,
int *category, int *category_index)
{
int exp_idx, bias, tmpbias1, tmpbias2, bits_left, num_bits, index, v, i, j;
int exp_index2[102] = { 0 };
int exp_index1[102] = { 0 };
int tmp_categorize_array[128 * 2] = { 0 };
int tmp_categorize_array1_idx = p->numvector_size;
int tmp_categorize_array2_idx = p->numvector_size;
bits_left = p->bits_per_subpacket - get_bits_count(&q->gb);
if (bits_left > q->samples_per_channel)
bits_left = q->samples_per_channel +
((bits_left - q->samples_per_channel) * 5) / 8;
bias = -32;
/* Estimate bias. */
for (i = 32; i > 0; i = i / 2) {
num_bits = 0;
index = 0;
for (j = p->total_subbands; j > 0; j--) {
exp_idx = av_clip_uintp2((i - quant_index_table[index] + bias) / 2, 3);
index++;
num_bits += expbits_tab[exp_idx];
}
if (num_bits >= bits_left - 32)
bias += i;
}
/* Calculate total number of bits. */
num_bits = 0;
for (i = 0; i < p->total_subbands; i++) {
exp_idx = av_clip_uintp2((bias - quant_index_table[i]) / 2, 3);
num_bits += expbits_tab[exp_idx];
exp_index1[i] = exp_idx;
exp_index2[i] = exp_idx;
}
tmpbias1 = tmpbias2 = num_bits;
for (j = 1; j < p->numvector_size; j++) {
if (tmpbias1 + tmpbias2 > 2 * bits_left) { /* ---> */
int max = -999999;
index = -1;
for (i = 0; i < p->total_subbands; i++) {
if (exp_index1[i] < 7) {
v = (-2 * exp_index1[i]) - quant_index_table[i] + bias;
if (v >= max) {
max = v;
index = i;
}
}
}
if (index == -1)
break;
tmp_categorize_array[tmp_categorize_array1_idx++] = index;
tmpbias1 -= expbits_tab[exp_index1[index]] -
expbits_tab[exp_index1[index] + 1];
++exp_index1[index];
} else { /* <--- */
int min = 999999;
index = -1;
for (i = 0; i < p->total_subbands; i++) {
if (exp_index2[i] > 0) {
v = (-2 * exp_index2[i]) - quant_index_table[i] + bias;
if (v < min) {
min = v;
index = i;
}
}
}
if (index == -1)
break;
tmp_categorize_array[--tmp_categorize_array2_idx] = index;
tmpbias2 -= expbits_tab[exp_index2[index]] -
expbits_tab[exp_index2[index] - 1];
--exp_index2[index];
}
}
for (i = 0; i < p->total_subbands; i++)
category[i] = exp_index2[i];
for (i = 0; i < p->numvector_size - 1; i++)
category_index[i] = tmp_categorize_array[tmp_categorize_array2_idx++];
}
/**
* Expand the category vector.
*
* @param q pointer to the COOKContext
* @param category pointer to the category array
* @param category_index pointer to the category_index array
*/
static inline void expand_category(COOKContext *q, int *category,
int *category_index)
{
int i;
for (i = 0; i < q->num_vectors; i++)
{
int idx = category_index[i];
if (++category[idx] >= FF_ARRAY_ELEMS(dither_tab))
--category[idx];
}
}
/**
* The real requantization of the mltcoefs
*
* @param q pointer to the COOKContext
* @param index index
* @param quant_index quantisation index
* @param subband_coef_index array of indexes to quant_centroid_tab
* @param subband_coef_sign signs of coefficients
* @param mlt_p pointer into the mlt buffer
*/
static void scalar_dequant_float(COOKContext *q, int index, int quant_index,
int *subband_coef_index, int *subband_coef_sign,
float *mlt_p)
{
int i;
float f1;
for (i = 0; i < SUBBAND_SIZE; i++) {
if (subband_coef_index[i]) {
f1 = quant_centroid_tab[index][subband_coef_index[i]];
if (subband_coef_sign[i])
f1 = -f1;
} else {
/* noise coding if subband_coef_index[i] == 0 */
f1 = dither_tab[index];
if (av_lfg_get(&q->random_state) < 0x80000000)
f1 = -f1;
}
mlt_p[i] = f1 * rootpow2tab[quant_index + 63];
}
}
/**
* Unpack the subband_coef_index and subband_coef_sign vectors.
*
* @param q pointer to the COOKContext
* @param category pointer to the category array
* @param subband_coef_index array of indexes to quant_centroid_tab
* @param subband_coef_sign signs of coefficients
*/
static int unpack_SQVH(COOKContext *q, COOKSubpacket *p, int category,
int *subband_coef_index, int *subband_coef_sign)
{
int i, j;
int vlc, vd, tmp, result;
vd = vd_tab[category];
result = 0;
for (i = 0; i < vpr_tab[category]; i++) {
vlc = get_vlc2(&q->gb, q->sqvh[category].table, q->sqvh[category].bits, 3);
if (p->bits_per_subpacket < get_bits_count(&q->gb)) {
vlc = 0;
result = 1;
}
for (j = vd - 1; j >= 0; j--) {
tmp = (vlc * invradix_tab[category]) / 0x100000;
subband_coef_index[vd * i + j] = vlc - tmp * (kmax_tab[category] + 1);
vlc = tmp;
}
for (j = 0; j < vd; j++) {
if (subband_coef_index[i * vd + j]) {
if (get_bits_count(&q->gb) < p->bits_per_subpacket) {
subband_coef_sign[i * vd + j] = get_bits1(&q->gb);
} else {
result = 1;
subband_coef_sign[i * vd + j] = 0;
}
} else {
subband_coef_sign[i * vd + j] = 0;
}
}
}
return result;
}
/**
* Fill the mlt_buffer with mlt coefficients.
*
* @param q pointer to the COOKContext
* @param category pointer to the category array
* @param quant_index_table pointer to the array
* @param mlt_buffer pointer to mlt coefficients
*/
static void decode_vectors(COOKContext *q, COOKSubpacket *p, int *category,
int *quant_index_table, float *mlt_buffer)
{
/* A zero in this table means that the subband coefficient is
random noise coded. */
int subband_coef_index[SUBBAND_SIZE];
/* A zero in this table means that the subband coefficient is a
positive multiplicator. */
int subband_coef_sign[SUBBAND_SIZE];
int band, j;
int index = 0;
for (band = 0; band < p->total_subbands; band++) {
index = category[band];
if (category[band] < 7) {
if (unpack_SQVH(q, p, category[band], subband_coef_index, subband_coef_sign)) {
index = 7;
for (j = 0; j < p->total_subbands; j++)
category[band + j] = 7;
}
}
if (index >= 7) {
memset(subband_coef_index, 0, sizeof(subband_coef_index));
memset(subband_coef_sign, 0, sizeof(subband_coef_sign));
}
q->scalar_dequant(q, index, quant_index_table[band],
subband_coef_index, subband_coef_sign,
&mlt_buffer[band * SUBBAND_SIZE]);
}
/* FIXME: should this be removed, or moved into loop above? */
if (p->total_subbands * SUBBAND_SIZE >= q->samples_per_channel)
return;
}
static int mono_decode(COOKContext *q, COOKSubpacket *p, float *mlt_buffer)
{
int category_index[128] = { 0 };
int category[128] = { 0 };
int quant_index_table[102];
int res, i;
if ((res = decode_envelope(q, p, quant_index_table)) < 0)
return res;
q->num_vectors = get_bits(&q->gb, p->log2_numvector_size);
categorize(q, p, quant_index_table, category, category_index);
expand_category(q, category, category_index);
for (i=0; i<p->total_subbands; i++) {
if (category[i] > 7)
return AVERROR_INVALIDDATA;
}
decode_vectors(q, p, category, quant_index_table, mlt_buffer);
return 0;
}
/**
* the actual requantization of the timedomain samples
*
* @param q pointer to the COOKContext
* @param buffer pointer to the timedomain buffer
* @param gain_index index for the block multiplier
* @param gain_index_next index for the next block multiplier
*/
static void interpolate_float(COOKContext *q, float *buffer,
int gain_index, int gain_index_next)
{
int i;
float fc1, fc2;
fc1 = pow2tab[gain_index + 63];
if (gain_index == gain_index_next) { // static gain
for (i = 0; i < q->gain_size_factor; i++)
buffer[i] *= fc1;
} else { // smooth gain
fc2 = q->gain_table[15 + (gain_index_next - gain_index)];
for (i = 0; i < q->gain_size_factor; i++) {
buffer[i] *= fc1;
fc1 *= fc2;
}
}
}
/**
* Apply transform window, overlap buffers.
*
* @param q pointer to the COOKContext
* @param inbuffer pointer to the mltcoefficients
* @param gains_ptr current and previous gains
* @param previous_buffer pointer to the previous buffer to be used for overlapping
*/
static void imlt_window_float(COOKContext *q, float *inbuffer,
cook_gains *gains_ptr, float *previous_buffer)
{
const float fc = pow2tab[gains_ptr->previous[0] + 63];
int i;
/* The weird thing here, is that the two halves of the time domain
* buffer are swapped. Also, the newest data, that we save away for
* next frame, has the wrong sign. Hence the subtraction below.
* Almost sounds like a complex conjugate/reverse data/FFT effect.
*/
/* Apply window and overlap */
for (i = 0; i < q->samples_per_channel; i++)
inbuffer[i] = inbuffer[i] * fc * q->mlt_window[i] -
previous_buffer[i] * q->mlt_window[q->samples_per_channel - 1 - i];
}
/**
* The modulated lapped transform, this takes transform coefficients
* and transforms them into timedomain samples.
* Apply transform window, overlap buffers, apply gain profile
* and buffer management.
*
* @param q pointer to the COOKContext
* @param inbuffer pointer to the mltcoefficients
* @param gains_ptr current and previous gains
* @param previous_buffer pointer to the previous buffer to be used for overlapping
*/
static void imlt_gain(COOKContext *q, float *inbuffer,
cook_gains *gains_ptr, float *previous_buffer)
{
float *buffer0 = q->mono_mdct_output;
float *buffer1 = q->mono_mdct_output + q->samples_per_channel;
int i;
/* Inverse modified discrete cosine transform */
q->mdct_fn(q->mdct_ctx, q->mono_mdct_output, inbuffer, sizeof(float));
q->imlt_window(q, buffer1, gains_ptr, previous_buffer);
/* Apply gain profile */
for (i = 0; i < 8; i++)
if (gains_ptr->now[i] || gains_ptr->now[i + 1])
q->interpolate(q, &buffer1[q->gain_size_factor * i],
gains_ptr->now[i], gains_ptr->now[i + 1]);
/* Save away the current to be previous block. */
memcpy(previous_buffer, buffer0,
q->samples_per_channel * sizeof(*previous_buffer));
}
/**
* function for getting the jointstereo coupling information
*
* @param q pointer to the COOKContext
* @param decouple_tab decoupling array
*/
static int decouple_info(COOKContext *q, COOKSubpacket *p, int *decouple_tab)
{
int i;
int vlc = get_bits1(&q->gb);
int start = cplband[p->js_subband_start];
int end = cplband[p->subbands - 1];
int length = end - start + 1;
if (start > end)
return 0;
if (vlc)
for (i = 0; i < length; i++)
decouple_tab[start + i] = get_vlc2(&q->gb,
p->channel_coupling.table,
COUPLING_VLC_BITS, 3);
else
for (i = 0; i < length; i++) {
int v = get_bits(&q->gb, p->js_vlc_bits);
if (v == (1<<p->js_vlc_bits)-1) {
av_log(q->avctx, AV_LOG_ERROR, "decouple value too large\n");
return AVERROR_INVALIDDATA;
}
decouple_tab[start + i] = v;
}
return 0;
}
/**
* function decouples a pair of signals from a single signal via multiplication.
*
* @param q pointer to the COOKContext
* @param subband index of the current subband
* @param f1 multiplier for channel 1 extraction
* @param f2 multiplier for channel 2 extraction
* @param decode_buffer input buffer
* @param mlt_buffer1 pointer to left channel mlt coefficients
* @param mlt_buffer2 pointer to right channel mlt coefficients
*/
static void decouple_float(COOKContext *q,
COOKSubpacket *p,
int subband,
float f1, float f2,
float *decode_buffer,
float *mlt_buffer1, float *mlt_buffer2)
{
int j, tmp_idx;
for (j = 0; j < SUBBAND_SIZE; j++) {
tmp_idx = ((p->js_subband_start + subband) * SUBBAND_SIZE) + j;
mlt_buffer1[SUBBAND_SIZE * subband + j] = f1 * decode_buffer[tmp_idx];
mlt_buffer2[SUBBAND_SIZE * subband + j] = f2 * decode_buffer[tmp_idx];
}
}
/**
* function for decoding joint stereo data
*
* @param q pointer to the COOKContext
* @param mlt_buffer1 pointer to left channel mlt coefficients
* @param mlt_buffer2 pointer to right channel mlt coefficients
*/
static int joint_decode(COOKContext *q, COOKSubpacket *p,
float *mlt_buffer_left, float *mlt_buffer_right)
{
int i, j, res;
int decouple_tab[SUBBAND_SIZE] = { 0 };
float *decode_buffer = q->decode_buffer_0;
int idx, cpl_tmp;
float f1, f2;
const float *cplscale;
memset(decode_buffer, 0, sizeof(q->decode_buffer_0));
/* Make sure the buffers are zeroed out. */
memset(mlt_buffer_left, 0, 1024 * sizeof(*mlt_buffer_left));
memset(mlt_buffer_right, 0, 1024 * sizeof(*mlt_buffer_right));
if ((res = decouple_info(q, p, decouple_tab)) < 0)
return res;
if ((res = mono_decode(q, p, decode_buffer)) < 0)
return res;
/* The two channels are stored interleaved in decode_buffer. */
for (i = 0; i < p->js_subband_start; i++) {
for (j = 0; j < SUBBAND_SIZE; j++) {
mlt_buffer_left[i * 20 + j] = decode_buffer[i * 40 + j];
mlt_buffer_right[i * 20 + j] = decode_buffer[i * 40 + 20 + j];
}
}
/* When we reach js_subband_start (the higher frequencies)
the coefficients are stored in a coupling scheme. */
idx = (1 << p->js_vlc_bits) - 1;
for (i = p->js_subband_start; i < p->subbands; i++) {
cpl_tmp = cplband[i];
idx -= decouple_tab[cpl_tmp];
cplscale = q->cplscales[p->js_vlc_bits - 2]; // choose decoupler table
f1 = cplscale[decouple_tab[cpl_tmp] + 1];
f2 = cplscale[idx];
q->decouple(q, p, i, f1, f2, decode_buffer,
mlt_buffer_left, mlt_buffer_right);
idx = (1 << p->js_vlc_bits) - 1;
}
return 0;
}
/**
* First part of subpacket decoding:
* decode raw stream bytes and read gain info.
*
* @param q pointer to the COOKContext
* @param inbuffer pointer to raw stream data
* @param gains_ptr array of current/prev gain pointers
*/
static inline void decode_bytes_and_gain(COOKContext *q, COOKSubpacket *p,
const uint8_t *inbuffer,
cook_gains *gains_ptr)
{
int offset;
offset = decode_bytes(inbuffer, q->decoded_bytes_buffer,
p->bits_per_subpacket / 8);
init_get_bits(&q->gb, q->decoded_bytes_buffer + offset,
p->bits_per_subpacket);
decode_gain_info(&q->gb, gains_ptr->now);
/* Swap current and previous gains */
FFSWAP(int *, gains_ptr->now, gains_ptr->previous);
}
/**
* Saturate the output signal and interleave.
*
* @param q pointer to the COOKContext
* @param out pointer to the output vector
*/
static void saturate_output_float(COOKContext *q, float *out)
{
q->adsp.vector_clipf(out, q->mono_mdct_output + q->samples_per_channel,
FFALIGN(q->samples_per_channel, 8), -1.0f, 1.0f);
}
/**
* Final part of subpacket decoding:
* Apply modulated lapped transform, gain compensation,
* clip and convert to integer.
*
* @param q pointer to the COOKContext
* @param decode_buffer pointer to the mlt coefficients
* @param gains_ptr array of current/prev gain pointers
* @param previous_buffer pointer to the previous buffer to be used for overlapping
* @param out pointer to the output buffer
*/
static inline void mlt_compensate_output(COOKContext *q, float *decode_buffer,
cook_gains *gains_ptr, float *previous_buffer,
float *out)
{
imlt_gain(q, decode_buffer, gains_ptr, previous_buffer);
if (out)
q->saturate_output(q, out);
}
/**
* Cook subpacket decoding. This function returns one decoded subpacket,
* usually 1024 samples per channel.
*
* @param q pointer to the COOKContext
* @param inbuffer pointer to the inbuffer
* @param outbuffer pointer to the outbuffer
*/
static int decode_subpacket(COOKContext *q, COOKSubpacket *p,
const uint8_t *inbuffer, float **outbuffer)
{
int sub_packet_size = p->size;
int res;
memset(q->decode_buffer_1, 0, sizeof(q->decode_buffer_1));
decode_bytes_and_gain(q, p, inbuffer, &p->gains1);
if (p->joint_stereo) {
if ((res = joint_decode(q, p, q->decode_buffer_1, q->decode_buffer_2)) < 0)
return res;
} else {
if ((res = mono_decode(q, p, q->decode_buffer_1)) < 0)
return res;
if (p->num_channels == 2) {
decode_bytes_and_gain(q, p, inbuffer + sub_packet_size / 2, &p->gains2);
if ((res = mono_decode(q, p, q->decode_buffer_2)) < 0)
return res;
}
}
mlt_compensate_output(q, q->decode_buffer_1, &p->gains1,
p->mono_previous_buffer1,
outbuffer ? outbuffer[p->ch_idx] : NULL);
if (p->num_channels == 2) {
if (p->joint_stereo)
mlt_compensate_output(q, q->decode_buffer_2, &p->gains1,
p->mono_previous_buffer2,
outbuffer ? outbuffer[p->ch_idx + 1] : NULL);
else
mlt_compensate_output(q, q->decode_buffer_2, &p->gains2,
p->mono_previous_buffer2,
outbuffer ? outbuffer[p->ch_idx + 1] : NULL);
}
return 0;
}
static int cook_decode_frame(AVCodecContext *avctx, AVFrame *frame,
int *got_frame_ptr, AVPacket *avpkt)
{
const uint8_t *buf = avpkt->data;
int buf_size = avpkt->size;
COOKContext *q = avctx->priv_data;
float **samples = NULL;
int i, ret;
int offset = 0;
int chidx = 0;
if (buf_size < avctx->block_align)
return buf_size;
/* get output buffer */
if (q->discarded_packets >= 2) {
frame->nb_samples = q->samples_per_channel;
if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
return ret;
samples = (float **)frame->extended_data;
}
/* estimate subpacket sizes */
q->subpacket[0].size = avctx->block_align;
for (i = 1; i < q->num_subpackets; i++) {
q->subpacket[i].size = 2 * buf[avctx->block_align - q->num_subpackets + i];
q->subpacket[0].size -= q->subpacket[i].size + 1;
if (q->subpacket[0].size < 0) {
av_log(avctx, AV_LOG_DEBUG,
"frame subpacket size total > avctx->block_align!\n");
return AVERROR_INVALIDDATA;
}
}
/* decode supbackets */
for (i = 0; i < q->num_subpackets; i++) {
q->subpacket[i].bits_per_subpacket = (q->subpacket[i].size * 8) >>
q->subpacket[i].bits_per_subpdiv;
q->subpacket[i].ch_idx = chidx;
av_log(avctx, AV_LOG_DEBUG,
"subpacket[%i] size %i js %i %i block_align %i\n",
i, q->subpacket[i].size, q->subpacket[i].joint_stereo, offset,
avctx->block_align);
if ((ret = decode_subpacket(q, &q->subpacket[i], buf + offset, samples)) < 0)
return ret;
offset += q->subpacket[i].size;
chidx += q->subpacket[i].num_channels;
av_log(avctx, AV_LOG_DEBUG, "subpacket[%i] %i %i\n",
i, q->subpacket[i].size * 8, get_bits_count(&q->gb));
}
/* Discard the first two frames: no valid audio. */
if (q->discarded_packets < 2) {
q->discarded_packets++;
*got_frame_ptr = 0;
return avctx->block_align;
}
*got_frame_ptr = 1;
return avctx->block_align;
}
static void dump_cook_context(COOKContext *q)
{
//int i=0;
#define PRINT(a, b) ff_dlog(q->avctx, " %s = %d\n", a, b);
ff_dlog(q->avctx, "COOKextradata\n");
ff_dlog(q->avctx, "cookversion=%x\n", q->subpacket[0].cookversion);
if (q->subpacket[0].cookversion > STEREO) {
PRINT("js_subband_start", q->subpacket[0].js_subband_start);
PRINT("js_vlc_bits", q->subpacket[0].js_vlc_bits);
}
ff_dlog(q->avctx, "COOKContext\n");
PRINT("nb_channels", q->avctx->ch_layout.nb_channels);
PRINT("bit_rate", (int)q->avctx->bit_rate);
PRINT("sample_rate", q->avctx->sample_rate);
PRINT("samples_per_channel", q->subpacket[0].samples_per_channel);
PRINT("subbands", q->subpacket[0].subbands);
PRINT("js_subband_start", q->subpacket[0].js_subband_start);
PRINT("log2_numvector_size", q->subpacket[0].log2_numvector_size);
PRINT("numvector_size", q->subpacket[0].numvector_size);
PRINT("total_subbands", q->subpacket[0].total_subbands);
}
/**
* Cook initialization
*
* @param avctx pointer to the AVCodecContext
*/
static av_cold int cook_decode_init(AVCodecContext *avctx)
{
static AVOnce init_static_once = AV_ONCE_INIT;
COOKContext *q = avctx->priv_data;
GetByteContext gb;
int s = 0;
unsigned int channel_mask = 0;
int samples_per_frame = 0;
int ret;
int channels = avctx->ch_layout.nb_channels;
q->avctx = avctx;
/* Take care of the codec specific extradata. */
if (avctx->extradata_size < 8) {
av_log(avctx, AV_LOG_ERROR, "Necessary extradata missing!\n");
return AVERROR_INVALIDDATA;
}
av_log(avctx, AV_LOG_DEBUG, "codecdata_length=%d\n", avctx->extradata_size);
bytestream2_init(&gb, avctx->extradata, avctx->extradata_size);
/* Take data from the AVCodecContext (RM container). */
if (!channels) {
av_log(avctx, AV_LOG_ERROR, "Invalid number of channels\n");
return AVERROR_INVALIDDATA;
}
if (avctx->block_align >= INT_MAX / 8)
return AVERROR(EINVAL);
/* Initialize RNG. */
av_lfg_init(&q->random_state, 0);
ff_audiodsp_init(&q->adsp);
while (bytestream2_get_bytes_left(&gb)) {
if (s >= FFMIN(MAX_SUBPACKETS, avctx->block_align)) {
avpriv_request_sample(avctx, "subpackets > %d", FFMIN(MAX_SUBPACKETS, avctx->block_align));
return AVERROR_PATCHWELCOME;
}
/* 8 for mono, 16 for stereo, ? for multichannel
Swap to right endianness so we don't need to care later on. */
q->subpacket[s].cookversion = bytestream2_get_be32(&gb);
samples_per_frame = bytestream2_get_be16(&gb);
q->subpacket[s].subbands = bytestream2_get_be16(&gb);
bytestream2_get_be32(&gb); // Unknown unused
q->subpacket[s].js_subband_start = bytestream2_get_be16(&gb);
if (q->subpacket[s].js_subband_start >= 51) {
av_log(avctx, AV_LOG_ERROR, "js_subband_start %d is too large\n", q->subpacket[s].js_subband_start);
return AVERROR_INVALIDDATA;
}
q->subpacket[s].js_vlc_bits = bytestream2_get_be16(&gb);
/* Initialize extradata related variables. */
q->subpacket[s].samples_per_channel = samples_per_frame / channels;
q->subpacket[s].bits_per_subpacket = avctx->block_align * 8;
/* Initialize default data states. */
q->subpacket[s].log2_numvector_size = 5;
q->subpacket[s].total_subbands = q->subpacket[s].subbands;
q->subpacket[s].num_channels = 1;
/* Initialize version-dependent variables */
av_log(avctx, AV_LOG_DEBUG, "subpacket[%i].cookversion=%x\n", s,
q->subpacket[s].cookversion);
q->subpacket[s].joint_stereo = 0;
switch (q->subpacket[s].cookversion) {
case MONO:
if (channels != 1) {
avpriv_request_sample(avctx, "Container channels != 1");
return AVERROR_PATCHWELCOME;
}
av_log(avctx, AV_LOG_DEBUG, "MONO\n");
break;
case STEREO:
if (channels != 1) {
q->subpacket[s].bits_per_subpdiv = 1;
q->subpacket[s].num_channels = 2;
}
av_log(avctx, AV_LOG_DEBUG, "STEREO\n");
break;
case JOINT_STEREO:
if (channels != 2) {
avpriv_request_sample(avctx, "Container channels != 2");
return AVERROR_PATCHWELCOME;
}
av_log(avctx, AV_LOG_DEBUG, "JOINT_STEREO\n");
if (avctx->extradata_size >= 16) {
q->subpacket[s].total_subbands = q->subpacket[s].subbands +
q->subpacket[s].js_subband_start;
q->subpacket[s].joint_stereo = 1;
q->subpacket[s].num_channels = 2;
}
if (q->subpacket[s].samples_per_channel > 256) {
q->subpacket[s].log2_numvector_size = 6;
}
if (q->subpacket[s].samples_per_channel > 512) {
q->subpacket[s].log2_numvector_size = 7;
}
break;
case MC_COOK:
av_log(avctx, AV_LOG_DEBUG, "MULTI_CHANNEL\n");
channel_mask |= q->subpacket[s].channel_mask = bytestream2_get_be32(&gb);
if (av_popcount64(q->subpacket[s].channel_mask) > 1) {
q->subpacket[s].total_subbands = q->subpacket[s].subbands +
q->subpacket[s].js_subband_start;
q->subpacket[s].joint_stereo = 1;
q->subpacket[s].num_channels = 2;
q->subpacket[s].samples_per_channel = samples_per_frame >> 1;
if (q->subpacket[s].samples_per_channel > 256) {
q->subpacket[s].log2_numvector_size = 6;
}
if (q->subpacket[s].samples_per_channel > 512) {
q->subpacket[s].log2_numvector_size = 7;
}
} else
q->subpacket[s].samples_per_channel = samples_per_frame;
break;
default:
avpriv_request_sample(avctx, "Cook version %d",
q->subpacket[s].cookversion);
return AVERROR_PATCHWELCOME;
}
if (s > 1 && q->subpacket[s].samples_per_channel != q->samples_per_channel) {
av_log(avctx, AV_LOG_ERROR, "different number of samples per channel!\n");
return AVERROR_INVALIDDATA;
} else
q->samples_per_channel = q->subpacket[0].samples_per_channel;
/* Initialize variable relations */
q->subpacket[s].numvector_size = (1 << q->subpacket[s].log2_numvector_size);
/* Try to catch some obviously faulty streams, otherwise it might be exploitable */
if (q->subpacket[s].total_subbands > 53) {
avpriv_request_sample(avctx, "total_subbands > 53");
return AVERROR_PATCHWELCOME;
}
if ((q->subpacket[s].js_vlc_bits > 6) ||
(q->subpacket[s].js_vlc_bits < 2 * q->subpacket[s].joint_stereo)) {
av_log(avctx, AV_LOG_ERROR, "js_vlc_bits = %d, only >= %d and <= 6 allowed!\n",
q->subpacket[s].js_vlc_bits, 2 * q->subpacket[s].joint_stereo);
return AVERROR_INVALIDDATA;
}
if (q->subpacket[s].subbands > 50) {
avpriv_request_sample(avctx, "subbands > 50");
return AVERROR_PATCHWELCOME;
}
if (q->subpacket[s].subbands == 0) {
avpriv_request_sample(avctx, "subbands = 0");
return AVERROR_PATCHWELCOME;
}
q->subpacket[s].gains1.now = q->subpacket[s].gain_1;
q->subpacket[s].gains1.previous = q->subpacket[s].gain_2;
q->subpacket[s].gains2.now = q->subpacket[s].gain_3;
q->subpacket[s].gains2.previous = q->subpacket[s].gain_4;
if (q->num_subpackets + q->subpacket[s].num_channels > channels) {
av_log(avctx, AV_LOG_ERROR, "Too many subpackets %d for channels %d\n", q->num_subpackets, channels);
return AVERROR_INVALIDDATA;
}
q->num_subpackets++;
s++;
}
/* Try to catch some obviously faulty streams, otherwise it might be exploitable */
if (q->samples_per_channel != 256 && q->samples_per_channel != 512 &&
q->samples_per_channel != 1024) {
avpriv_request_sample(avctx, "samples_per_channel = %d",
q->samples_per_channel);
return AVERROR_PATCHWELCOME;
}
/* Generate tables */
ff_thread_once(&init_static_once, init_pow2table);
init_gain_table(q);
init_cplscales_table(q);
if ((ret = init_cook_vlc_tables(q)))
return ret;
/* Pad the databuffer with:
DECODE_BYTES_PAD1 or DECODE_BYTES_PAD2 for decode_bytes(),
AV_INPUT_BUFFER_PADDING_SIZE, for the bitstreamreader. */
q->decoded_bytes_buffer =
av_mallocz(avctx->block_align
+ DECODE_BYTES_PAD1(avctx->block_align)
+ AV_INPUT_BUFFER_PADDING_SIZE);
if (!q->decoded_bytes_buffer)
return AVERROR(ENOMEM);
/* Initialize transform. */
if ((ret = init_cook_mlt(q)))
return ret;
/* Initialize COOK signal arithmetic handling */
if (1) {
q->scalar_dequant = scalar_dequant_float;
q->decouple = decouple_float;
q->imlt_window = imlt_window_float;
q->interpolate = interpolate_float;
q->saturate_output = saturate_output_float;
}
avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
av_channel_layout_uninit(&avctx->ch_layout);
if (channel_mask)
av_channel_layout_from_mask(&avctx->ch_layout, channel_mask);
else
av_channel_layout_default(&avctx->ch_layout, channels);
dump_cook_context(q);
return 0;
}
const FFCodec ff_cook_decoder = {
.p.name = "cook",
CODEC_LONG_NAME("Cook / Cooker / Gecko (RealAudio G2)"),
.p.type = AVMEDIA_TYPE_AUDIO,
.p.id = AV_CODEC_ID_COOK,
.priv_data_size = sizeof(COOKContext),
.init = cook_decode_init,
.close = cook_decode_close,
FF_CODEC_DECODE_CB(cook_decode_frame),
.p.capabilities = AV_CODEC_CAP_DR1,
.p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
AV_SAMPLE_FMT_NONE },
.caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
};
Reference in New Issue View Git Blame Copy Permalink