1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-11-23 21:54:53 +02:00
Files
FFmpeg/libavcodec/vulkan/rangecoder.comp
Lynne ed2391d341 ffv1enc: add a Vulkan encoder
This commit implements a standard, compliant, version 3 and version 4
FFv1 encoder, entirely in Vulkan. The encoder is written in standard
GLSL and requires a Vulkan 1.3 supporting GPU with the BDA extension.

The encoder can use any amount of slices, but nominally, should use
32x32 slices (1024 in total) to maximize parallelism.

All features are supported, as well as all pixel formats.
This includes:
 - Rice
 - Range coding with a custom quantization table
 - PCM encoding

CRC calculation is also massively parallelized on the GPU.

Encoding of unaligned dimensions on subsampled data requires
version 4, or requires oversizing the image to 64-pixel alignment
and cropping out the padding via container flags.

Performance-wise, this makes 1080p real-time screen capture possible
at 60fps on even modest GPUs.
2024-11-18 07:54:22 +01:00

191 lines
5.0 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
*/
struct RangeCoder {
u8buf bytestream_start;
u8buf bytestream;
uint low;
uint16_t range;
uint8_t outstanding_count;
uint8_t outstanding_byte;
};
/* Full renorm version that can handle outstanding_byte == 0xFF */
void renorm_encoder_full(inout RangeCoder c)
{
int bs_cnt = 0;
if (c.outstanding_byte == 0xFF) {
c.outstanding_byte = uint8_t(c.low >> 8);
} else if (c.low <= 0xFF00) {
c.bytestream[bs_cnt++].v = c.outstanding_byte;
uint8_t cnt = c.outstanding_count;
for (; cnt > 0; cnt--)
c.bytestream[bs_cnt++].v = uint8_t(0xFF);
c.outstanding_count = uint8_t(0);
c.outstanding_byte = uint8_t(c.low >> 8);
} else if (c.low >= 0x10000) {
c.bytestream[bs_cnt++].v = c.outstanding_byte + uint8_t(1);
uint8_t cnt = c.outstanding_count;
for (; cnt > 0; cnt--)
c.bytestream[bs_cnt++].v = uint8_t(0x00);
c.outstanding_count = uint8_t(0);
c.outstanding_byte = uint8_t(bitfieldExtract(c.low, 8, 8));
} else {
c.outstanding_count++;
}
c.bytestream = OFFBUF(u8buf, c.bytestream, bs_cnt);
c.range <<= 8;
c.low = bitfieldInsert(0, c.low, 8, 8);
}
/* Cannot deal with outstanding_byte == -1 in the name of speed */
void renorm_encoder(inout RangeCoder c)
{
uint8_t oc = c.outstanding_count + uint8_t(1);
uint low = c.low;
c.range <<= 8;
c.low = bitfieldInsert(0, low, 8, 8);
if (low > 0xFF00 && low < 0x10000) {
c.outstanding_count = oc;
return;
}
u8buf bs = c.bytestream;
uint8_t outstanding_byte = c.outstanding_byte;
c.bytestream = OFFBUF(u8buf, bs, oc);
c.outstanding_count = uint8_t(0);
c.outstanding_byte = uint8_t(low >> 8);
uint8_t obs = uint8_t(low > 0xFF00);
uint8_t fill = obs - uint8_t(1); /* unsigned underflow */
bs[0].v = outstanding_byte + obs;
for (int i = 1; i < oc; i++)
bs[i].v = fill;
}
void put_rac_norenorm(inout RangeCoder c, uint64_t state, bool bit)
{
u8buf sb = u8buf(state);
uint val = uint(sb.v);
uint16_t range1 = uint16_t((uint(c.range) * val) >> 8);
#ifdef DEBUG
if (val == 0)
debugPrintfEXT("Error: state is zero (addr: 0x%lx)", uint64_t(sb));
if (range1 >= c.range)
debugPrintfEXT("Error: range1 >= c.range");
if (range1 <= 0)
debugPrintfEXT("Error: range1 <= 0");
#endif
uint16_t diff = c.range - range1;
if (bit) {
c.low += diff;
c.range = range1;
} else {
c.range = diff;
}
sb.v = zero_one_state[(uint(bit) << 8) + val];
#ifdef DEBUG
if (sb.v == 0)
debugPrintfEXT("Error: inserted zero state from tab %i idx %i", bit, val);
#endif
}
/* Equiprobable bit */
void put_rac_equi(inout RangeCoder c, bool bit)
{
uint16_t range1 = c.range >> 1;
#ifdef DEBUG
if (range1 >= c.range)
debugPrintfEXT("Error: range1 >= c.range");
if (range1 <= 0)
debugPrintfEXT("Error: range1 <= 0");
#endif
if (bit) {
c.low += c.range - range1;
c.range = range1;
} else {
c.range -= range1;
}
if (c.range < 0x100)
renorm_encoder(c);
}
void put_rac_terminate(inout RangeCoder c)
{
uint16_t range1 = uint16_t((uint(c.range) * 129) >> 8);
#ifdef DEBUG
if (range1 >= c.range)
debugPrintfEXT("Error: range1 >= c.range");
if (range1 <= 0)
debugPrintfEXT("Error: range1 <= 0");
#endif
c.range -= range1;
if (c.range < 0x100)
renorm_encoder(c);
}
/* Return the number of bytes written. */
uint32_t rac_terminate(inout RangeCoder c)
{
put_rac_terminate(c);
c.range = uint16_t(0xFF);
c.low += 0xFF;
renorm_encoder(c);
c.range = uint16_t(0xFF);
renorm_encoder(c);
#ifdef DEBUG
if (c.low != 0)
debugPrintfEXT("Error: c.low != 0");
if (c.range < 0x100)
debugPrintfEXT("Error: range < 0x100");
#endif
return uint32_t(uint64_t(c.bytestream) - uint64_t(c.bytestream_start));
}
void rac_init(out RangeCoder r, u8buf data, uint64_t buf_size)
{
r.bytestream_start = data;
r.bytestream = data;
r.low = 0;
r.range = uint16_t(0xFF00);
r.outstanding_count = uint8_t(0);
r.outstanding_byte = uint8_t(0xFF);
}