You've already forked FFmpeg
mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2025-11-29 05:57:37 +02:00
This patch adds a fully-featured level 3 and 4 decoder for FFv1, supporting Golomb and all Range coding variants, all pixel formats, and all features, except for the newly added floating-point formats. On a 6000 Ada, for 3840x2160 bgr0 content at 50Mbps (standard desktop recording), it is able to do 400fps. An Alder Lake with 24 threads can barely do 100fps.
160 lines
3.6 KiB
Plaintext
160 lines
3.6 KiB
Plaintext
/*
|
|
* FFv1 codec
|
|
*
|
|
* Copyright (c) 2024 Lynne <dev@lynne.ee>
|
|
*
|
|
* 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
|
|
*/
|
|
|
|
#define VLC_STATE_SIZE 8
|
|
layout(buffer_reference, buffer_reference_align = VLC_STATE_SIZE) buffer VlcState {
|
|
uint32_t error_sum;
|
|
int16_t drift;
|
|
int8_t bias;
|
|
uint8_t count;
|
|
};
|
|
|
|
void update_vlc_state(inout VlcState state, const int v)
|
|
{
|
|
int drift = state.drift;
|
|
int count = state.count;
|
|
int bias = state.bias;
|
|
state.error_sum += uint16_t(abs(v));
|
|
drift += v;
|
|
|
|
if (count == 128) { // FIXME: variable
|
|
count >>= 1;
|
|
drift >>= 1;
|
|
state.error_sum >>= 1;
|
|
}
|
|
count++;
|
|
|
|
if (drift <= -count) {
|
|
bias = max(bias - 1, -128);
|
|
drift = max(drift + count, -count + 1);
|
|
} else if (drift > 0) {
|
|
bias = min(bias + 1, 127);
|
|
drift = min(drift - count, 0);
|
|
}
|
|
|
|
state.bias = int8_t(bias);
|
|
state.drift = int16_t(drift);
|
|
state.count = uint8_t(count);
|
|
}
|
|
|
|
struct Symbol {
|
|
uint32_t bits;
|
|
uint32_t val;
|
|
};
|
|
|
|
Symbol set_ur_golomb(int i, int k, int limit, int esc_len)
|
|
{
|
|
int e;
|
|
Symbol sym;
|
|
|
|
#ifdef DEBUG
|
|
if (i < 0)
|
|
debugPrintfEXT("Error: i is zero!");
|
|
#endif
|
|
|
|
e = i >> k;
|
|
if (e < limit) {
|
|
sym.bits = e + k + 1;
|
|
sym.val = (1 << k) + zero_extend(i, k);
|
|
} else {
|
|
sym.bits = limit + esc_len;
|
|
sym.val = i - limit + 1;
|
|
}
|
|
|
|
return sym;
|
|
}
|
|
|
|
/**
|
|
* write signed golomb rice code (ffv1).
|
|
*/
|
|
Symbol set_sr_golomb(int i, int k, int limit, int esc_len)
|
|
{
|
|
int v;
|
|
|
|
v = -2 * i - 1;
|
|
v ^= (v >> 31);
|
|
|
|
return set_ur_golomb(v, k, limit, esc_len);
|
|
}
|
|
|
|
Symbol get_vlc_symbol(inout VlcState state, int v, int bits)
|
|
{
|
|
int i, k, code;
|
|
Symbol sym;
|
|
v = fold(v - int(state.bias), bits);
|
|
|
|
i = state.count;
|
|
k = 0;
|
|
while (i < state.error_sum) { // FIXME: optimize
|
|
k++;
|
|
i += i;
|
|
}
|
|
|
|
#ifdef DEBUG
|
|
if (k > 16)
|
|
debugPrintfEXT("Error: k > 16!");
|
|
#endif
|
|
|
|
code = v ^ ((2 * state.drift + state.count) >> 31);
|
|
|
|
update_vlc_state(state, v);
|
|
|
|
return set_sr_golomb(code, k, 12, bits);
|
|
}
|
|
|
|
uint get_ur_golomb(inout GetBitContext gb, uint k, int limit, int esc_len)
|
|
{
|
|
for (uint i = 0; i < 12; i++)
|
|
if (get_bit(gb))
|
|
return get_bits(gb, k) + (i << k);
|
|
|
|
return get_bits(gb, esc_len) + 11;
|
|
}
|
|
|
|
int get_sr_golomb(inout GetBitContext gb, uint k, int limit, int esc_len)
|
|
{
|
|
int v = int(get_ur_golomb(gb, k, limit, esc_len));
|
|
return (v >> 1) ^ -(v & 1);
|
|
}
|
|
|
|
int read_vlc_symbol(inout GetBitContext gb, inout VlcState state, int bits)
|
|
{
|
|
int k, i, v, ret;
|
|
|
|
i = state.count;
|
|
k = 0;
|
|
while (i < state.error_sum) { // FIXME: optimize
|
|
k++;
|
|
i += i;
|
|
}
|
|
|
|
v = get_sr_golomb(gb, k, 12, bits);
|
|
|
|
v ^= ((2 * state.drift + state.count) >> 31);
|
|
|
|
ret = fold(v + state.bias, bits);
|
|
|
|
update_vlc_state(state, v);
|
|
|
|
return ret;
|
|
}
|