1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-26 19:01:44 +02:00
FFmpeg/libavcodec/opus_rc.c
Rostislav Pehlivanov 3aca16e857 opus_rc: fix CDF encoding of zeroes
0th CDF value is where probabilities are.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
2017-07-20 08:24:46 +01:00

412 lines
12 KiB
C

/*
* Copyright (c) 2012 Andrew D'Addesio
* Copyright (c) 2013-2014 Mozilla Corporation
* Copyright (c) 2017 Rostislav Pehlivanov <atomnuker@gmail.com>
*
* 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
*/
#include "opus_rc.h"
#define OPUS_RC_BITS 32
#define OPUS_RC_SYM 8
#define OPUS_RC_CEIL ((1 << OPUS_RC_SYM) - 1)
#define OPUS_RC_TOP (1u << 31)
#define OPUS_RC_BOT (OPUS_RC_TOP >> OPUS_RC_SYM)
#define OPUS_RC_SHIFT (OPUS_RC_BITS - OPUS_RC_SYM - 1)
static av_always_inline void opus_rc_enc_carryout(OpusRangeCoder *rc, int cbuf)
{
const int cb = cbuf >> OPUS_RC_SYM, mb = (OPUS_RC_CEIL + cb) & OPUS_RC_CEIL;
if (cbuf == OPUS_RC_CEIL) {
rc->ext++;
return;
}
rc->rng_cur[0] = rc->rem + cb;
rc->rng_cur += (rc->rem >= 0);
for (; rc->ext > 0; rc->ext--)
*rc->rng_cur++ = mb;
av_assert0(rc->rng_cur < rc->rb.position);
rc->rem = cbuf & OPUS_RC_CEIL; /* Propagate */
}
static av_always_inline void opus_rc_dec_normalize(OpusRangeCoder *rc)
{
while (rc->range <= OPUS_RC_BOT) {
rc->value = ((rc->value << OPUS_RC_SYM) | (get_bits(&rc->gb, OPUS_RC_SYM) ^ OPUS_RC_CEIL)) & (OPUS_RC_TOP - 1);
rc->range <<= OPUS_RC_SYM;
rc->total_bits += OPUS_RC_SYM;
}
}
static av_always_inline void opus_rc_enc_normalize(OpusRangeCoder *rc)
{
while (rc->range <= OPUS_RC_BOT) {
opus_rc_enc_carryout(rc, rc->value >> OPUS_RC_SHIFT);
rc->value = (rc->value << OPUS_RC_SYM) & (OPUS_RC_TOP - 1);
rc->range <<= OPUS_RC_SYM;
rc->total_bits += OPUS_RC_SYM;
}
}
static av_always_inline void opus_rc_dec_update(OpusRangeCoder *rc, uint32_t scale,
uint32_t low, uint32_t high,
uint32_t total)
{
rc->value -= scale * (total - high);
rc->range = low ? scale * (high - low)
: rc->range - scale * (total - high);
opus_rc_dec_normalize(rc);
}
/* Main encoding function, this needs to go fast */
static av_always_inline void opus_rc_enc_update(OpusRangeCoder *rc, uint32_t b, uint32_t p,
uint32_t p_tot, const int ptwo)
{
uint32_t rscaled, cnd = !!b;
if (ptwo) /* Whole function is inlined so hopefully branch is optimized out */
rscaled = rc->range >> ff_log2(p_tot);
else
rscaled = rc->range/p_tot;
rc->value += cnd*(rc->range - rscaled*(p_tot - b));
rc->range = (!cnd)*(rc->range - rscaled*(p_tot - p)) + cnd*rscaled*(p - b);
opus_rc_enc_normalize(rc);
}
uint32_t ff_opus_rc_dec_cdf(OpusRangeCoder *rc, const uint16_t *cdf)
{
unsigned int k, scale, total, symbol, low, high;
total = *cdf++;
scale = rc->range / total;
symbol = rc->value / scale + 1;
symbol = total - FFMIN(symbol, total);
for (k = 0; cdf[k] <= symbol; k++);
high = cdf[k];
low = k ? cdf[k-1] : 0;
opus_rc_dec_update(rc, scale, low, high, total);
return k;
}
void ff_opus_rc_enc_cdf(OpusRangeCoder *rc, int val, const uint16_t *cdf)
{
opus_rc_enc_update(rc, (!!val)*cdf[val], cdf[val + 1], cdf[0], 1);
}
uint32_t ff_opus_rc_dec_log(OpusRangeCoder *rc, uint32_t bits)
{
uint32_t k, scale;
scale = rc->range >> bits; // in this case, scale = symbol
if (rc->value >= scale) {
rc->value -= scale;
rc->range -= scale;
k = 0;
} else {
rc->range = scale;
k = 1;
}
opus_rc_dec_normalize(rc);
return k;
}
void ff_opus_rc_enc_log(OpusRangeCoder *rc, int val, uint32_t bits)
{
bits = (1 << bits) - 1;
opus_rc_enc_update(rc, (!!val)*bits, bits + !!val, bits + 1, 1);
}
/**
* CELT: read 1-25 raw bits at the end of the frame, backwards byte-wise
*/
uint32_t ff_opus_rc_get_raw(OpusRangeCoder *rc, uint32_t count)
{
uint32_t value = 0;
while (rc->rb.bytes && rc->rb.cachelen < count) {
rc->rb.cacheval |= *--rc->rb.position << rc->rb.cachelen;
rc->rb.cachelen += 8;
rc->rb.bytes--;
}
value = av_mod_uintp2(rc->rb.cacheval, count);
rc->rb.cacheval >>= count;
rc->rb.cachelen -= count;
rc->total_bits += count;
return value;
}
/**
* CELT: write 0 - 31 bits to the rawbits buffer
*/
void ff_opus_rc_put_raw(OpusRangeCoder *rc, uint32_t val, uint32_t count)
{
const int to_write = FFMIN(32 - rc->rb.cachelen, count);
rc->total_bits += count;
rc->rb.cacheval |= av_mod_uintp2(val, to_write) << rc->rb.cachelen;
rc->rb.cachelen = (rc->rb.cachelen + to_write) % 32;
if (!rc->rb.cachelen && count) {
AV_WB32(rc->rb.position, rc->rb.cacheval);
rc->rb.bytes += 4;
rc->rb.position -= 4;
rc->rb.cachelen = count - to_write;
rc->rb.cacheval = av_mod_uintp2(val >> to_write, rc->rb.cachelen);
av_assert0(rc->rng_cur < rc->rb.position);
}
}
/**
* CELT: read a uniform distribution
*/
uint32_t ff_opus_rc_dec_uint(OpusRangeCoder *rc, uint32_t size)
{
uint32_t bits, k, scale, total;
bits = opus_ilog(size - 1);
total = (bits > 8) ? ((size - 1) >> (bits - 8)) + 1 : size;
scale = rc->range / total;
k = rc->value / scale + 1;
k = total - FFMIN(k, total);
opus_rc_dec_update(rc, scale, k, k + 1, total);
if (bits > 8) {
k = k << (bits - 8) | ff_opus_rc_get_raw(rc, bits - 8);
return FFMIN(k, size - 1);
} else
return k;
}
/**
* CELT: write a uniformly distributed integer
*/
void ff_opus_rc_enc_uint(OpusRangeCoder *rc, uint32_t val, uint32_t size)
{
const int ps = FFMAX(opus_ilog(size - 1) - 8, 0);
opus_rc_enc_update(rc, val >> ps, (val >> ps) + 1, ((size - 1) >> ps) + 1, 0);
ff_opus_rc_put_raw(rc, val, ps);
}
uint32_t ff_opus_rc_dec_uint_step(OpusRangeCoder *rc, int k0)
{
/* Use a probability of 3 up to itheta=8192 and then use 1 after */
uint32_t k, scale, symbol, total = (k0+1)*3 + k0;
scale = rc->range / total;
symbol = rc->value / scale + 1;
symbol = total - FFMIN(symbol, total);
k = (symbol < (k0+1)*3) ? symbol/3 : symbol - (k0+1)*2;
opus_rc_dec_update(rc, scale, (k <= k0) ? 3*(k+0) : (k-1-k0) + 3*(k0+1),
(k <= k0) ? 3*(k+1) : (k-0-k0) + 3*(k0+1), total);
return k;
}
void ff_opus_rc_enc_uint_step(OpusRangeCoder *rc, uint32_t val, int k0)
{
const uint32_t a = val <= k0, b = 2*a + 1;
k0 = (k0 + 1) << 1;
val = b*(val + k0) - 3*a*k0;
opus_rc_enc_update(rc, val, val + b, (k0 << 1) - 1, 0);
}
uint32_t ff_opus_rc_dec_uint_tri(OpusRangeCoder *rc, int qn)
{
uint32_t k, scale, symbol, total, low, center;
total = ((qn>>1) + 1) * ((qn>>1) + 1);
scale = rc->range / total;
center = rc->value / scale + 1;
center = total - FFMIN(center, total);
if (center < total >> 1) {
k = (ff_sqrt(8 * center + 1) - 1) >> 1;
low = k * (k + 1) >> 1;
symbol = k + 1;
} else {
k = (2*(qn + 1) - ff_sqrt(8*(total - center - 1) + 1)) >> 1;
low = total - ((qn + 1 - k) * (qn + 2 - k) >> 1);
symbol = qn + 1 - k;
}
opus_rc_dec_update(rc, scale, low, low + symbol, total);
return k;
}
void ff_opus_rc_enc_uint_tri(OpusRangeCoder *rc, uint32_t k, int qn)
{
uint32_t symbol, low, total;
total = ((qn>>1) + 1) * ((qn>>1) + 1);
if (k <= qn >> 1) {
low = k * (k + 1) >> 1;
symbol = k + 1;
} else {
low = total - ((qn + 1 - k) * (qn + 2 - k) >> 1);
symbol = qn + 1 - k;
}
opus_rc_enc_update(rc, low, low + symbol, total, 0);
}
int ff_opus_rc_dec_laplace(OpusRangeCoder *rc, uint32_t symbol, int decay)
{
/* extends the range coder to model a Laplace distribution */
int value = 0;
uint32_t scale, low = 0, center;
scale = rc->range >> 15;
center = rc->value / scale + 1;
center = (1 << 15) - FFMIN(center, 1 << 15);
if (center >= symbol) {
value++;
low = symbol;
symbol = 1 + ((32768 - 32 - symbol) * (16384-decay) >> 15);
while (symbol > 1 && center >= low + 2 * symbol) {
value++;
symbol *= 2;
low += symbol;
symbol = (((symbol - 2) * decay) >> 15) + 1;
}
if (symbol <= 1) {
int distance = (center - low) >> 1;
value += distance;
low += 2 * distance;
}
if (center < low + symbol)
value *= -1;
else
low += symbol;
}
opus_rc_dec_update(rc, scale, low, FFMIN(low + symbol, 32768), 32768);
return value;
}
void ff_opus_rc_enc_laplace(OpusRangeCoder *rc, int *value, uint32_t symbol, int decay)
{
uint32_t low = symbol;
int i = 1, val = FFABS(*value), pos = *value > 0;
if (!val) {
opus_rc_enc_update(rc, 0, symbol, 1 << 15, 1);
return;
}
symbol = ((32768 - 32 - symbol)*(16384 - decay)) >> 15;
for (; i < val && symbol; i++) {
low += (symbol << 1) + 2;
symbol = (symbol*decay) >> 14;
}
if (symbol) {
low += (++symbol)*pos;
} else {
const int distance = FFMIN(val - i, (((32768 - low) - !pos) >> 1) - 1);
low += pos + (distance << 1);
symbol = FFMIN(1, 32768 - low);
*value = FFSIGN(*value)*(distance + i);
}
opus_rc_enc_update(rc, low, low + symbol, 1 << 15, 1);
}
int ff_opus_rc_dec_init(OpusRangeCoder *rc, const uint8_t *data, int size)
{
int ret = init_get_bits8(&rc->gb, data, size);
if (ret < 0)
return ret;
rc->range = 128;
rc->value = 127 - get_bits(&rc->gb, 7);
rc->total_bits = 9;
opus_rc_dec_normalize(rc);
return 0;
}
void ff_opus_rc_dec_raw_init(OpusRangeCoder *rc, const uint8_t *rightend, uint32_t bytes)
{
rc->rb.position = rightend;
rc->rb.bytes = bytes;
rc->rb.cachelen = 0;
rc->rb.cacheval = 0;
}
void ff_opus_rc_enc_end(OpusRangeCoder *rc, uint8_t *dst, int size)
{
int rng_bytes, bits = OPUS_RC_BITS - opus_ilog(rc->range);
uint32_t mask = (OPUS_RC_TOP - 1) >> bits;
uint32_t end = (rc->value + mask) & ~mask;
if ((end | mask) >= rc->value + rc->range) {
bits++;
mask >>= 1;
end = (rc->value + mask) & ~mask;
}
/* Finish what's left */
while (bits > 0) {
opus_rc_enc_carryout(rc, end >> OPUS_RC_SHIFT);
end = (end << OPUS_RC_SYM) & (OPUS_RC_TOP - 1);
bits -= OPUS_RC_SYM;
}
/* Flush out anything left or marked */
if (rc->rem >= 0 || rc->ext > 0)
opus_rc_enc_carryout(rc, 0);
rng_bytes = rc->rng_cur - rc->buf;
memcpy(dst, rc->buf, rng_bytes);
rc->waste = size*8 - (rc->rb.bytes*8 + rc->rb.cachelen) - rng_bytes*8;
/* Put the rawbits part, if any */
if (rc->rb.bytes || rc->rb.cachelen) {
int i, lap;
uint8_t *rb_src, *rb_dst;
ff_opus_rc_put_raw(rc, 0, 32 - rc->rb.cachelen);
rb_src = rc->buf + OPUS_MAX_PACKET_SIZE + 12 - rc->rb.bytes;
rb_dst = dst + FFMAX(size - rc->rb.bytes, 0);
lap = &dst[rng_bytes] - rb_dst;
for (i = 0; i < lap; i++)
rb_dst[i] |= rb_src[i];
memcpy(&rb_dst[lap], &rb_src[lap], FFMAX(rc->rb.bytes - lap, 0));
}
}
void ff_opus_rc_enc_init(OpusRangeCoder *rc)
{
rc->value = 0;
rc->range = OPUS_RC_TOP;
rc->total_bits = OPUS_RC_BITS + 1;
rc->rem = -1;
rc->ext = 0;
rc->rng_cur = rc->buf;
ff_opus_rc_dec_raw_init(rc, rc->buf + OPUS_MAX_PACKET_SIZE + 8, 0);
}