You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1066 lines
38 KiB

/*
* Ut Video decoder
* Copyright (c) 2011 Konstantin Shishkov
*
* 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
* Ut Video decoder
*/
#include <inttypes.h>
#include <stdlib.h>
#define CACHED_BITSTREAM_READER !ARCH_X86_32
#define UNCHECKED_BITSTREAM_READER 1
#include "libavutil/intreadwrite.h"
#include "libavutil/pixdesc.h"
#include "avcodec.h"
#include "bswapdsp.h"
#include "bytestream.h"
#include "get_bits.h"
#include "internal.h"
#include "thread.h"
#include "utvideo.h"
avcodec/utvideodec: Avoid implicit qsort when creating Huffman tables The Huffman trees used by Ut Video have two important characteristics: (i) Longer codes are on the left of the tree and (ii) for codes of the same length, the symbol is descending from left to right in the tree. Therefore all the information that needs to be transmitted is how long the code corresponding to a given symbol is; and this is also all that is transmitted. Before 341914495e5c2f60bc920b0c6f660e5948a47f5a, the decoder used qsort to sort the (length, symbol) pairs by ascending length and for equal lengths by ascending symbol. Since said commit, the decoder uses a first pass over the lengths table to count how many symbols of each length there are; with (i) one can then easily calculate the code of the left-most code with a given length in the tree and from there one can calculate the codes for all entries, using one running counter for each possible length. This eliminated the explicit qsort in build_huff(). Yet ff_init_vlc_sparse() sorts the table itself as it has to ensure that all the entries that will be placed in the same subtable are contiguous. The tables created now are non-contiguous (they are ordered by symbol and codes of different length aren't ordered at all; only codes of the same length are ordered according to (ii)). This commit therefore modifies the algorithm used to automatically create tables whose codes are sorted from left to right in the tree. The key to do so is the observation that the counts obtained in the first pass can be used to contain the range of the codes of each length in the second pass: If counts[i] is the count of codes with length i, then the first counts[32] codes are of length 32, the next counts[31] codes are of length 31 etc. So one knows the index of the lowest symbol whose code has length 32 (if any): It is counts[32] - 1 due to (ii), whereas the index of the lowest symbol whose code has length 31 (if any) is counts[32] + counts[31] - 1; the index of the second-to-lowest symbol of length 32 (if existing) is counts[32] - 2 etc. If one follows the algorithm outlined above, one can switch to ff_init_vlc_from_lengths() which has no implicit qsort; it also means that one can offload the computation of the codes. This turned out to be beneficial for performance: For the sample from ticket #4044 it decreased the decicycles spent on one call to build_huff() from 508480 to 340688 (GCC 9.3, looping 10 times over the file to get enough runs and then repeating this ten times); for another sample (YUV420p, natural content, 5500 frames, also ten iterations) the time went down from 382346 to 275533 decicycles. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
typedef struct HuffEntry {
uint8_t len;
uint16_t sym;
} HuffEntry;
static int build_huff(UtvideoContext *c, const uint8_t *src, VLC *vlc,
int *fsym, unsigned nb_elems)
{
int i;
avcodec/utvideodec: Avoid implicit qsort when creating Huffman tables The Huffman trees used by Ut Video have two important characteristics: (i) Longer codes are on the left of the tree and (ii) for codes of the same length, the symbol is descending from left to right in the tree. Therefore all the information that needs to be transmitted is how long the code corresponding to a given symbol is; and this is also all that is transmitted. Before 341914495e5c2f60bc920b0c6f660e5948a47f5a, the decoder used qsort to sort the (length, symbol) pairs by ascending length and for equal lengths by ascending symbol. Since said commit, the decoder uses a first pass over the lengths table to count how many symbols of each length there are; with (i) one can then easily calculate the code of the left-most code with a given length in the tree and from there one can calculate the codes for all entries, using one running counter for each possible length. This eliminated the explicit qsort in build_huff(). Yet ff_init_vlc_sparse() sorts the table itself as it has to ensure that all the entries that will be placed in the same subtable are contiguous. The tables created now are non-contiguous (they are ordered by symbol and codes of different length aren't ordered at all; only codes of the same length are ordered according to (ii)). This commit therefore modifies the algorithm used to automatically create tables whose codes are sorted from left to right in the tree. The key to do so is the observation that the counts obtained in the first pass can be used to contain the range of the codes of each length in the second pass: If counts[i] is the count of codes with length i, then the first counts[32] codes are of length 32, the next counts[31] codes are of length 31 etc. So one knows the index of the lowest symbol whose code has length 32 (if any): It is counts[32] - 1 due to (ii), whereas the index of the lowest symbol whose code has length 31 (if any) is counts[32] + counts[31] - 1; the index of the second-to-lowest symbol of length 32 (if existing) is counts[32] - 2 etc. If one follows the algorithm outlined above, one can switch to ff_init_vlc_from_lengths() which has no implicit qsort; it also means that one can offload the computation of the codes. This turned out to be beneficial for performance: For the sample from ticket #4044 it decreased the decicycles spent on one call to build_huff() from 508480 to 340688 (GCC 9.3, looping 10 times over the file to get enough runs and then repeating this ten times); for another sample (YUV420p, natural content, 5500 frames, also ten iterations) the time went down from 382346 to 275533 decicycles. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
HuffEntry he[1024];
uint8_t bits[1024];
avcodec/utvideodec: Avoid qsort when creating Huffman tables The Ut video format uses Huffman trees which are only implicitly coded in the bitstream: Only the lengths of the codes are coded, the rest has to be inferred by the decoder according to the rule that the longer codes are to the left of shorter codes in the tree and on each level the symbols are descending from left to right. Because longer codes are to the left of shorter codes, one needs to know how many non-leaf nodes there are on each level in order to know the code of the next left-most leaf (which belongs to the highest symbol on that level). The current code does this by sorting the entries to be ascending according to length and (for entries with the same length) ascending according to their symbols. This array is then traversed in reverse order, so that the lowest level is dealt with first, so that the number of non-leaf nodes of the next higher level is known when processing said level. But this can also be calculated without sorting: Simply count how many leaf nodes there are on each level. Then one can calculate the number of non-leaf nodes on each level iteratively from the lowest level upwards: It is just half the number of nodes of the level below. This improves performance: For the sample from ticket #4044 the amount of decicycles for one call to build_huff() decreased from 1055489 to 446310 for Clang 10 and from 1080306 to 535155 for GCC 9. Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
uint16_t codes_count[33] = { 0 };
*fsym = -1;
for (i = 0; i < nb_elems; i++) {
avcodec/utvideodec: Avoid qsort when creating Huffman tables The Ut video format uses Huffman trees which are only implicitly coded in the bitstream: Only the lengths of the codes are coded, the rest has to be inferred by the decoder according to the rule that the longer codes are to the left of shorter codes in the tree and on each level the symbols are descending from left to right. Because longer codes are to the left of shorter codes, one needs to know how many non-leaf nodes there are on each level in order to know the code of the next left-most leaf (which belongs to the highest symbol on that level). The current code does this by sorting the entries to be ascending according to length and (for entries with the same length) ascending according to their symbols. This array is then traversed in reverse order, so that the lowest level is dealt with first, so that the number of non-leaf nodes of the next higher level is known when processing said level. But this can also be calculated without sorting: Simply count how many leaf nodes there are on each level. Then one can calculate the number of non-leaf nodes on each level iteratively from the lowest level upwards: It is just half the number of nodes of the level below. This improves performance: For the sample from ticket #4044 the amount of decicycles for one call to build_huff() decreased from 1055489 to 446310 for Clang 10 and from 1080306 to 535155 for GCC 9. Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
if (src[i] == 0) {
*fsym = i;
return 0;
} else if (src[i] == 255) {
bits[i] = 0;
} else if (src[i] <= 32) {
bits[i] = src[i];
} else
return AVERROR_INVALIDDATA;
avcodec/utvideodec: Avoid qsort when creating Huffman tables The Ut video format uses Huffman trees which are only implicitly coded in the bitstream: Only the lengths of the codes are coded, the rest has to be inferred by the decoder according to the rule that the longer codes are to the left of shorter codes in the tree and on each level the symbols are descending from left to right. Because longer codes are to the left of shorter codes, one needs to know how many non-leaf nodes there are on each level in order to know the code of the next left-most leaf (which belongs to the highest symbol on that level). The current code does this by sorting the entries to be ascending according to length and (for entries with the same length) ascending according to their symbols. This array is then traversed in reverse order, so that the lowest level is dealt with first, so that the number of non-leaf nodes of the next higher level is known when processing said level. But this can also be calculated without sorting: Simply count how many leaf nodes there are on each level. Then one can calculate the number of non-leaf nodes on each level iteratively from the lowest level upwards: It is just half the number of nodes of the level below. This improves performance: For the sample from ticket #4044 the amount of decicycles for one call to build_huff() decreased from 1055489 to 446310 for Clang 10 and from 1080306 to 535155 for GCC 9. Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
codes_count[bits[i]]++;
}
avcodec/utvideodec: Avoid qsort when creating Huffman tables The Ut video format uses Huffman trees which are only implicitly coded in the bitstream: Only the lengths of the codes are coded, the rest has to be inferred by the decoder according to the rule that the longer codes are to the left of shorter codes in the tree and on each level the symbols are descending from left to right. Because longer codes are to the left of shorter codes, one needs to know how many non-leaf nodes there are on each level in order to know the code of the next left-most leaf (which belongs to the highest symbol on that level). The current code does this by sorting the entries to be ascending according to length and (for entries with the same length) ascending according to their symbols. This array is then traversed in reverse order, so that the lowest level is dealt with first, so that the number of non-leaf nodes of the next higher level is known when processing said level. But this can also be calculated without sorting: Simply count how many leaf nodes there are on each level. Then one can calculate the number of non-leaf nodes on each level iteratively from the lowest level upwards: It is just half the number of nodes of the level below. This improves performance: For the sample from ticket #4044 the amount of decicycles for one call to build_huff() decreased from 1055489 to 446310 for Clang 10 and from 1080306 to 535155 for GCC 9. Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
if (codes_count[0] == nb_elems)
return AVERROR_INVALIDDATA;
avcodec/utvideodec: Avoid implicit qsort when creating Huffman tables The Huffman trees used by Ut Video have two important characteristics: (i) Longer codes are on the left of the tree and (ii) for codes of the same length, the symbol is descending from left to right in the tree. Therefore all the information that needs to be transmitted is how long the code corresponding to a given symbol is; and this is also all that is transmitted. Before 341914495e5c2f60bc920b0c6f660e5948a47f5a, the decoder used qsort to sort the (length, symbol) pairs by ascending length and for equal lengths by ascending symbol. Since said commit, the decoder uses a first pass over the lengths table to count how many symbols of each length there are; with (i) one can then easily calculate the code of the left-most code with a given length in the tree and from there one can calculate the codes for all entries, using one running counter for each possible length. This eliminated the explicit qsort in build_huff(). Yet ff_init_vlc_sparse() sorts the table itself as it has to ensure that all the entries that will be placed in the same subtable are contiguous. The tables created now are non-contiguous (they are ordered by symbol and codes of different length aren't ordered at all; only codes of the same length are ordered according to (ii)). This commit therefore modifies the algorithm used to automatically create tables whose codes are sorted from left to right in the tree. The key to do so is the observation that the counts obtained in the first pass can be used to contain the range of the codes of each length in the second pass: If counts[i] is the count of codes with length i, then the first counts[32] codes are of length 32, the next counts[31] codes are of length 31 etc. So one knows the index of the lowest symbol whose code has length 32 (if any): It is counts[32] - 1 due to (ii), whereas the index of the lowest symbol whose code has length 31 (if any) is counts[32] + counts[31] - 1; the index of the second-to-lowest symbol of length 32 (if existing) is counts[32] - 2 etc. If one follows the algorithm outlined above, one can switch to ff_init_vlc_from_lengths() which has no implicit qsort; it also means that one can offload the computation of the codes. This turned out to be beneficial for performance: For the sample from ticket #4044 it decreased the decicycles spent on one call to build_huff() from 508480 to 340688 (GCC 9.3, looping 10 times over the file to get enough runs and then repeating this ten times); for another sample (YUV420p, natural content, 5500 frames, also ten iterations) the time went down from 382346 to 275533 decicycles. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
/* For Ut Video, longer codes are to the left of the tree and
* for codes with the same length the symbol is descending from
* left to right. So after the next loop --codes_count[i] will
* be the index of the first (lowest) symbol of length i when
* indexed by the position in the tree with left nodes being first. */
for (int i = 31; i >= 0; i--)
codes_count[i] += codes_count[i + 1];
for (unsigned i = 0; i < nb_elems; i++)
he[--codes_count[bits[i]]] = (HuffEntry) { bits[i], i };
#define VLC_BITS 11
avcodec/utvideodec: Avoid implicit qsort when creating Huffman tables The Huffman trees used by Ut Video have two important characteristics: (i) Longer codes are on the left of the tree and (ii) for codes of the same length, the symbol is descending from left to right in the tree. Therefore all the information that needs to be transmitted is how long the code corresponding to a given symbol is; and this is also all that is transmitted. Before 341914495e5c2f60bc920b0c6f660e5948a47f5a, the decoder used qsort to sort the (length, symbol) pairs by ascending length and for equal lengths by ascending symbol. Since said commit, the decoder uses a first pass over the lengths table to count how many symbols of each length there are; with (i) one can then easily calculate the code of the left-most code with a given length in the tree and from there one can calculate the codes for all entries, using one running counter for each possible length. This eliminated the explicit qsort in build_huff(). Yet ff_init_vlc_sparse() sorts the table itself as it has to ensure that all the entries that will be placed in the same subtable are contiguous. The tables created now are non-contiguous (they are ordered by symbol and codes of different length aren't ordered at all; only codes of the same length are ordered according to (ii)). This commit therefore modifies the algorithm used to automatically create tables whose codes are sorted from left to right in the tree. The key to do so is the observation that the counts obtained in the first pass can be used to contain the range of the codes of each length in the second pass: If counts[i] is the count of codes with length i, then the first counts[32] codes are of length 32, the next counts[31] codes are of length 31 etc. So one knows the index of the lowest symbol whose code has length 32 (if any): It is counts[32] - 1 due to (ii), whereas the index of the lowest symbol whose code has length 31 (if any) is counts[32] + counts[31] - 1; the index of the second-to-lowest symbol of length 32 (if existing) is counts[32] - 2 etc. If one follows the algorithm outlined above, one can switch to ff_init_vlc_from_lengths() which has no implicit qsort; it also means that one can offload the computation of the codes. This turned out to be beneficial for performance: For the sample from ticket #4044 it decreased the decicycles spent on one call to build_huff() from 508480 to 340688 (GCC 9.3, looping 10 times over the file to get enough runs and then repeating this ten times); for another sample (YUV420p, natural content, 5500 frames, also ten iterations) the time went down from 382346 to 275533 decicycles. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
return ff_init_vlc_from_lengths(vlc, VLC_BITS, codes_count[0],
&he[0].len, sizeof(*he),
&he[0].sym, sizeof(*he), 2, 0, 0, c->avctx);
}
static int decode_plane10(UtvideoContext *c, int plane_no,
uint16_t *dst, ptrdiff_t stride,
int width, int height,
const uint8_t *src, const uint8_t *huff,
int use_pred)
{
int i, j, slice, pix, ret;
int sstart, send;
VLC vlc;
GetBitContext gb;
int prev, fsym;
avcodec/utvideodec: Avoid implicit qsort when creating Huffman tables The Huffman trees used by Ut Video have two important characteristics: (i) Longer codes are on the left of the tree and (ii) for codes of the same length, the symbol is descending from left to right in the tree. Therefore all the information that needs to be transmitted is how long the code corresponding to a given symbol is; and this is also all that is transmitted. Before 341914495e5c2f60bc920b0c6f660e5948a47f5a, the decoder used qsort to sort the (length, symbol) pairs by ascending length and for equal lengths by ascending symbol. Since said commit, the decoder uses a first pass over the lengths table to count how many symbols of each length there are; with (i) one can then easily calculate the code of the left-most code with a given length in the tree and from there one can calculate the codes for all entries, using one running counter for each possible length. This eliminated the explicit qsort in build_huff(). Yet ff_init_vlc_sparse() sorts the table itself as it has to ensure that all the entries that will be placed in the same subtable are contiguous. The tables created now are non-contiguous (they are ordered by symbol and codes of different length aren't ordered at all; only codes of the same length are ordered according to (ii)). This commit therefore modifies the algorithm used to automatically create tables whose codes are sorted from left to right in the tree. The key to do so is the observation that the counts obtained in the first pass can be used to contain the range of the codes of each length in the second pass: If counts[i] is the count of codes with length i, then the first counts[32] codes are of length 32, the next counts[31] codes are of length 31 etc. So one knows the index of the lowest symbol whose code has length 32 (if any): It is counts[32] - 1 due to (ii), whereas the index of the lowest symbol whose code has length 31 (if any) is counts[32] + counts[31] - 1; the index of the second-to-lowest symbol of length 32 (if existing) is counts[32] - 2 etc. If one follows the algorithm outlined above, one can switch to ff_init_vlc_from_lengths() which has no implicit qsort; it also means that one can offload the computation of the codes. This turned out to be beneficial for performance: For the sample from ticket #4044 it decreased the decicycles spent on one call to build_huff() from 508480 to 340688 (GCC 9.3, looping 10 times over the file to get enough runs and then repeating this ten times); for another sample (YUV420p, natural content, 5500 frames, also ten iterations) the time went down from 382346 to 275533 decicycles. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
if ((ret = build_huff(c, huff, &vlc, &fsym, 1024)) < 0) {
av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
return ret;
}
if (fsym >= 0) { // build_huff reported a symbol to fill slices with
send = 0;
for (slice = 0; slice < c->slices; slice++) {
uint16_t *dest;
sstart = send;
send = (height * (slice + 1) / c->slices);
dest = dst + sstart * stride;
prev = 0x200;
for (j = sstart; j < send; j++) {
for (i = 0; i < width; i++) {
pix = fsym;
if (use_pred) {
prev += pix;
prev &= 0x3FF;
pix = prev;
}
dest[i] = pix;
}
dest += stride;
}
}
return 0;
}
send = 0;
for (slice = 0; slice < c->slices; slice++) {
uint16_t *dest;
int slice_data_start, slice_data_end, slice_size;
sstart = send;
send = (height * (slice + 1) / c->slices);
dest = dst + sstart * stride;
// slice offset and size validation was done earlier
slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
slice_data_end = AV_RL32(src + slice * 4);
slice_size = slice_data_end - slice_data_start;
if (!slice_size) {
av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
"yet a slice has a length of zero.\n");
goto fail;
}
memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
(uint32_t *)(src + slice_data_start + c->slices * 4),
(slice_data_end - slice_data_start + 3) >> 2);
init_get_bits(&gb, c->slice_bits, slice_size * 8);
prev = 0x200;
for (j = sstart; j < send; j++) {
for (i = 0; i < width; i++) {
pix = get_vlc2(&gb, vlc.table, VLC_BITS, 3);
if (pix < 0) {
av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
goto fail;
}
if (use_pred) {
prev += pix;
prev &= 0x3FF;
pix = prev;
}
dest[i] = pix;
}
dest += stride;
if (get_bits_left(&gb) < 0) {
av_log(c->avctx, AV_LOG_ERROR,
"Slice decoding ran out of bits\n");
goto fail;
}
}
if (get_bits_left(&gb) > 32)
av_log(c->avctx, AV_LOG_WARNING,
"%d bits left after decoding slice\n", get_bits_left(&gb));
}
ff_free_vlc(&vlc);
return 0;
fail:
ff_free_vlc(&vlc);
return AVERROR_INVALIDDATA;
}
static int compute_cmask(int plane_no, int interlaced, enum AVPixelFormat pix_fmt)
{
const int is_luma = (pix_fmt == AV_PIX_FMT_YUV420P) && !plane_no;
if (interlaced)
return ~(1 + 2 * is_luma);
return ~is_luma;
}
static int decode_plane(UtvideoContext *c, int plane_no,
uint8_t *dst, ptrdiff_t stride,
int width, int height,
const uint8_t *src, int use_pred)
{
int i, j, slice, pix;
int sstart, send;
VLC vlc;
GetBitContext gb;
int ret, prev, fsym;
const int cmask = compute_cmask(plane_no, c->interlaced, c->avctx->pix_fmt);
if (c->pack) {
send = 0;
for (slice = 0; slice < c->slices; slice++) {
GetBitContext cbit, pbit;
uint8_t *dest, *p;
ret = init_get_bits8_le(&cbit, c->control_stream[plane_no][slice], c->control_stream_size[plane_no][slice]);
if (ret < 0)
return ret;
ret = init_get_bits8_le(&pbit, c->packed_stream[plane_no][slice], c->packed_stream_size[plane_no][slice]);
if (ret < 0)
return ret;
sstart = send;
send = (height * (slice + 1) / c->slices) & cmask;
dest = dst + sstart * stride;
if (3 * ((dst + send * stride - dest + 7)/8) > get_bits_left(&cbit))
return AVERROR_INVALIDDATA;
for (p = dest; p < dst + send * stride; p += 8) {
int bits = get_bits_le(&cbit, 3);
if (bits == 0) {
*(uint64_t *) p = 0;
} else {
uint32_t sub = 0x80 >> (8 - (bits + 1)), add;
int k;
if ((bits + 1) * 8 > get_bits_left(&pbit))
return AVERROR_INVALIDDATA;
for (k = 0; k < 8; k++) {
p[k] = get_bits_le(&pbit, bits + 1);
add = (~p[k] & sub) << (8 - bits);
p[k] -= sub;
p[k] += add;
}
}
}
}
return 0;
}
avcodec/utvideodec: Avoid implicit qsort when creating Huffman tables The Huffman trees used by Ut Video have two important characteristics: (i) Longer codes are on the left of the tree and (ii) for codes of the same length, the symbol is descending from left to right in the tree. Therefore all the information that needs to be transmitted is how long the code corresponding to a given symbol is; and this is also all that is transmitted. Before 341914495e5c2f60bc920b0c6f660e5948a47f5a, the decoder used qsort to sort the (length, symbol) pairs by ascending length and for equal lengths by ascending symbol. Since said commit, the decoder uses a first pass over the lengths table to count how many symbols of each length there are; with (i) one can then easily calculate the code of the left-most code with a given length in the tree and from there one can calculate the codes for all entries, using one running counter for each possible length. This eliminated the explicit qsort in build_huff(). Yet ff_init_vlc_sparse() sorts the table itself as it has to ensure that all the entries that will be placed in the same subtable are contiguous. The tables created now are non-contiguous (they are ordered by symbol and codes of different length aren't ordered at all; only codes of the same length are ordered according to (ii)). This commit therefore modifies the algorithm used to automatically create tables whose codes are sorted from left to right in the tree. The key to do so is the observation that the counts obtained in the first pass can be used to contain the range of the codes of each length in the second pass: If counts[i] is the count of codes with length i, then the first counts[32] codes are of length 32, the next counts[31] codes are of length 31 etc. So one knows the index of the lowest symbol whose code has length 32 (if any): It is counts[32] - 1 due to (ii), whereas the index of the lowest symbol whose code has length 31 (if any) is counts[32] + counts[31] - 1; the index of the second-to-lowest symbol of length 32 (if existing) is counts[32] - 2 etc. If one follows the algorithm outlined above, one can switch to ff_init_vlc_from_lengths() which has no implicit qsort; it also means that one can offload the computation of the codes. This turned out to be beneficial for performance: For the sample from ticket #4044 it decreased the decicycles spent on one call to build_huff() from 508480 to 340688 (GCC 9.3, looping 10 times over the file to get enough runs and then repeating this ten times); for another sample (YUV420p, natural content, 5500 frames, also ten iterations) the time went down from 382346 to 275533 decicycles. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
if (build_huff(c, src, &vlc, &fsym, 256)) {
av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
return AVERROR_INVALIDDATA;
}
if (fsym >= 0) { // build_huff reported a symbol to fill slices with
send = 0;
for (slice = 0; slice < c->slices; slice++) {
uint8_t *dest;
sstart = send;
send = (height * (slice + 1) / c->slices) & cmask;
dest = dst + sstart * stride;
prev = 0x80;
for (j = sstart; j < send; j++) {
for (i = 0; i < width; i++) {
pix = fsym;
if (use_pred) {
prev += (unsigned)pix;
pix = prev;
}
dest[i] = pix;
}
dest += stride;
}
}
return 0;
}
src += 256;
send = 0;
for (slice = 0; slice < c->slices; slice++) {
uint8_t *dest;
int slice_data_start, slice_data_end, slice_size;
sstart = send;
send = (height * (slice + 1) / c->slices) & cmask;
dest = dst + sstart * stride;
// slice offset and size validation was done earlier
slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
slice_data_end = AV_RL32(src + slice * 4);
slice_size = slice_data_end - slice_data_start;
if (!slice_size) {
av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
"yet a slice has a length of zero.\n");
goto fail;
}
memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
(uint32_t *)(src + slice_data_start + c->slices * 4),
(slice_data_end - slice_data_start + 3) >> 2);
init_get_bits(&gb, c->slice_bits, slice_size * 8);
prev = 0x80;
for (j = sstart; j < send; j++) {
for (i = 0; i < width; i++) {
pix = get_vlc2(&gb, vlc.table, VLC_BITS, 3);
if (pix < 0) {
av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
goto fail;
}
if (use_pred) {
prev += pix;
pix = prev;
}
dest[i] = pix;
}
if (get_bits_left(&gb) < 0) {
av_log(c->avctx, AV_LOG_ERROR,
"Slice decoding ran out of bits\n");
goto fail;
}
dest += stride;
}
if (get_bits_left(&gb) > 32)
av_log(c->avctx, AV_LOG_WARNING,
"%d bits left after decoding slice\n", get_bits_left(&gb));
}
ff_free_vlc(&vlc);
return 0;
fail:
ff_free_vlc(&vlc);
return AVERROR_INVALIDDATA;
}
#undef A
#undef B
#undef C
static void restore_median_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
int width, int height, int slices, int rmode)
{
int i, j, slice;
int A, B, C;
uint8_t *bsrc;
int slice_start, slice_height;
const int cmask = ~rmode;
for (slice = 0; slice < slices; slice++) {
slice_start = ((slice * height) / slices) & cmask;
slice_height = ((((slice + 1) * height) / slices) & cmask) -
slice_start;
if (!slice_height)
continue;
bsrc = src + slice_start * stride;
// first line - left neighbour prediction
bsrc[0] += 0x80;
c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
bsrc += stride;
if (slice_height <= 1)
continue;
// second line - first element has top prediction, the rest uses median
C = bsrc[-stride];
bsrc[0] += C;
A = bsrc[0];
for (i = 1; i < FFMIN(width, 16); i++) { /* scalar loop (DSP need align 16) */
B = bsrc[i - stride];
bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
C = B;
A = bsrc[i];
}
if (width > 16)
c->llviddsp.add_median_pred(bsrc + 16, bsrc - stride + 16,
bsrc + 16, width - 16, &A, &B);
bsrc += stride;
// the rest of lines use continuous median prediction
for (j = 2; j < slice_height; j++) {
c->llviddsp.add_median_pred(bsrc, bsrc - stride,
bsrc, width, &A, &B);
bsrc += stride;
}
}
}
/* UtVideo interlaced mode treats every two lines as a single one,
* so restoring function should take care of possible padding between
* two parts of the same "line".
*/
static void restore_median_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
int width, int height, int slices, int rmode)
{
int i, j, slice;
int A, B, C;
uint8_t *bsrc;
int slice_start, slice_height;
const int cmask = ~(rmode ? 3 : 1);
const ptrdiff_t stride2 = stride << 1;
for (slice = 0; slice < slices; slice++) {
slice_start = ((slice * height) / slices) & cmask;
slice_height = ((((slice + 1) * height) / slices) & cmask) -
slice_start;
slice_height >>= 1;
if (!slice_height)
continue;
bsrc = src + slice_start * stride;
// first line - left neighbour prediction
bsrc[0] += 0x80;
A = c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
c->llviddsp.add_left_pred(bsrc + stride, bsrc + stride, width, A);
bsrc += stride2;
if (slice_height <= 1)
continue;
// second line - first element has top prediction, the rest uses median
C = bsrc[-stride2];
bsrc[0] += C;
A = bsrc[0];
for (i = 1; i < FFMIN(width, 16); i++) { /* scalar loop (DSP need align 16) */
B = bsrc[i - stride2];
bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
C = B;
A = bsrc[i];
}
if (width > 16)
c->llviddsp.add_median_pred(bsrc + 16, bsrc - stride2 + 16,
bsrc + 16, width - 16, &A, &B);
c->llviddsp.add_median_pred(bsrc + stride, bsrc - stride,
bsrc + stride, width, &A, &B);
bsrc += stride2;
// the rest of lines use continuous median prediction
for (j = 2; j < slice_height; j++) {
c->llviddsp.add_median_pred(bsrc, bsrc - stride2,
bsrc, width, &A, &B);
c->llviddsp.add_median_pred(bsrc + stride, bsrc - stride,
bsrc + stride, width, &A, &B);
bsrc += stride2;
}
}
}
static void restore_gradient_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
int width, int height, int slices, int rmode)
{
int i, j, slice;
int A, B, C;
uint8_t *bsrc;
int slice_start, slice_height;
const int cmask = ~rmode;
int min_width = FFMIN(width, 32);
for (slice = 0; slice < slices; slice++) {
slice_start = ((slice * height) / slices) & cmask;
slice_height = ((((slice + 1) * height) / slices) & cmask) -
slice_start;
if (!slice_height)
continue;
bsrc = src + slice_start * stride;
// first line - left neighbour prediction
bsrc[0] += 0x80;
c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
bsrc += stride;
if (slice_height <= 1)
continue;
for (j = 1; j < slice_height; j++) {
// second line - first element has top prediction, the rest uses gradient
bsrc[0] = (bsrc[0] + bsrc[-stride]) & 0xFF;
for (i = 1; i < min_width; i++) { /* dsp need align 32 */
A = bsrc[i - stride];
B = bsrc[i - (stride + 1)];
C = bsrc[i - 1];
bsrc[i] = (A - B + C + bsrc[i]) & 0xFF;
}
if (width > 32)
c->llviddsp.add_gradient_pred(bsrc + 32, stride, width - 32);
bsrc += stride;
}
}
}
static void restore_gradient_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
int width, int height, int slices, int rmode)
{
int i, j, slice;
int A, B, C;
uint8_t *bsrc;
int slice_start, slice_height;
const int cmask = ~(rmode ? 3 : 1);
const ptrdiff_t stride2 = stride << 1;
int min_width = FFMIN(width, 32);
for (slice = 0; slice < slices; slice++) {
slice_start = ((slice * height) / slices) & cmask;
slice_height = ((((slice + 1) * height) / slices) & cmask) -
slice_start;
slice_height >>= 1;
if (!slice_height)
continue;
bsrc = src + slice_start * stride;
// first line - left neighbour prediction
bsrc[0] += 0x80;
A = c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
c->llviddsp.add_left_pred(bsrc + stride, bsrc + stride, width, A);
bsrc += stride2;
if (slice_height <= 1)
continue;
for (j = 1; j < slice_height; j++) {
// second line - first element has top prediction, the rest uses gradient
bsrc[0] = (bsrc[0] + bsrc[-stride2]) & 0xFF;
for (i = 1; i < min_width; i++) { /* dsp need align 32 */
A = bsrc[i - stride2];
B = bsrc[i - (stride2 + 1)];
C = bsrc[i - 1];
bsrc[i] = (A - B + C + bsrc[i]) & 0xFF;
}
if (width > 32)
c->llviddsp.add_gradient_pred(bsrc + 32, stride2, width - 32);
A = bsrc[-stride];
B = bsrc[-(1 + stride + stride - width)];
C = bsrc[width - 1];
bsrc[stride] = (A - B + C + bsrc[stride]) & 0xFF;
for (i = 1; i < width; i++) {
A = bsrc[i - stride];
B = bsrc[i - (1 + stride)];
C = bsrc[i - 1 + stride];
bsrc[i + stride] = (A - B + C + bsrc[i + stride]) & 0xFF;
}
bsrc += stride2;
}
}
}
static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
AVPacket *avpkt)
{
const uint8_t *buf = avpkt->data;
int buf_size = avpkt->size;
UtvideoContext *c = avctx->priv_data;
int i, j;
const uint8_t *plane_start[5];
int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
int ret;
GetByteContext gb;
ThreadFrame frame = { .f = data };
if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0)
return ret;
/* parse plane structure to get frame flags and validate slice offsets */
bytestream2_init(&gb, buf, buf_size);
if (c->pack) {
const uint8_t *packed_stream;
const uint8_t *control_stream;
GetByteContext pb;
uint32_t nb_cbs;
int left;
c->frame_info = PRED_GRADIENT << 8;
if (bytestream2_get_byte(&gb) != 1)
return AVERROR_INVALIDDATA;
bytestream2_skip(&gb, 3);
c->offset = bytestream2_get_le32(&gb);
if (buf_size <= c->offset + 8LL)
return AVERROR_INVALIDDATA;
bytestream2_init(&pb, buf + 8 + c->offset, buf_size - 8 - c->offset);
nb_cbs = bytestream2_get_le32(&pb);
if (nb_cbs > c->offset)
return AVERROR_INVALIDDATA;
packed_stream = buf + 8;
control_stream = packed_stream + (c->offset - nb_cbs);
left = control_stream - packed_stream;
for (i = 0; i < c->planes; i++) {
for (j = 0; j < c->slices; j++) {
c->packed_stream[i][j] = packed_stream;
c->packed_stream_size[i][j] = bytestream2_get_le32(&pb);
if (c->packed_stream_size[i][j] > left)
return AVERROR_INVALIDDATA;
left -= c->packed_stream_size[i][j];
packed_stream += c->packed_stream_size[i][j];
}
}
left = buf + buf_size - control_stream;
for (i = 0; i < c->planes; i++) {
for (j = 0; j < c->slices; j++) {
c->control_stream[i][j] = control_stream;
c->control_stream_size[i][j] = bytestream2_get_le32(&pb);
if (c->control_stream_size[i][j] > left)
return AVERROR_INVALIDDATA;
left -= c->control_stream_size[i][j];
control_stream += c->control_stream_size[i][j];
}
}
} else if (c->pro) {
if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
return AVERROR_INVALIDDATA;
}
c->frame_info = bytestream2_get_le32u(&gb);
c->slices = ((c->frame_info >> 16) & 0xff) + 1;
for (i = 0; i < c->planes; i++) {
plane_start[i] = gb.buffer;
if (bytestream2_get_bytes_left(&gb) < 1024 + 4 * c->slices) {
av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
return AVERROR_INVALIDDATA;
}
slice_start = 0;
slice_end = 0;
for (j = 0; j < c->slices; j++) {
slice_end = bytestream2_get_le32u(&gb);
if (slice_end < 0 || slice_end < slice_start ||
bytestream2_get_bytes_left(&gb) < slice_end + 1024LL) {
av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
return AVERROR_INVALIDDATA;
}
slice_size = slice_end - slice_start;
slice_start = slice_end;
max_slice_size = FFMAX(max_slice_size, slice_size);
}
plane_size = slice_end;
bytestream2_skipu(&gb, plane_size);
bytestream2_skipu(&gb, 1024);
}
plane_start[c->planes] = gb.buffer;
} else {
for (i = 0; i < c->planes; i++) {
plane_start[i] = gb.buffer;
if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
return AVERROR_INVALIDDATA;
}
bytestream2_skipu(&gb, 256);
slice_start = 0;
slice_end = 0;
for (j = 0; j < c->slices; j++) {
slice_end = bytestream2_get_le32u(&gb);
if (slice_end < 0 || slice_end < slice_start ||
bytestream2_get_bytes_left(&gb) < slice_end) {
av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
return AVERROR_INVALIDDATA;
}
slice_size = slice_end - slice_start;
slice_start = slice_end;
max_slice_size = FFMAX(max_slice_size, slice_size);
}
plane_size = slice_end;
bytestream2_skipu(&gb, plane_size);
}
plane_start[c->planes] = gb.buffer;
if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
return AVERROR_INVALIDDATA;
}
c->frame_info = bytestream2_get_le32u(&gb);
}
av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
c->frame_info);
c->frame_pred = (c->frame_info >> 8) & 3;
max_slice_size += 4*avctx->width;
if (!c->pack) {
av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
max_slice_size + AV_INPUT_BUFFER_PADDING_SIZE);
if (!c->slice_bits) {
av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
return AVERROR(ENOMEM);
}
}
switch (c->avctx->pix_fmt) {
case AV_PIX_FMT_GBRP:
case AV_PIX_FMT_GBRAP:
for (i = 0; i < c->planes; i++) {
ret = decode_plane(c, i, frame.f->data[i],
frame.f->linesize[i], avctx->width,
avctx->height, plane_start[i],
c->frame_pred == PRED_LEFT);
if (ret)
return ret;
if (c->frame_pred == PRED_MEDIAN) {
if (!c->interlaced) {
restore_median_planar(c, frame.f->data[i],
frame.f->linesize[i], avctx->width,
avctx->height, c->slices, 0);
} else {
restore_median_planar_il(c, frame.f->data[i],
frame.f->linesize[i],
avctx->width, avctx->height, c->slices,
0);
}
} else if (c->frame_pred == PRED_GRADIENT) {
if (!c->interlaced) {
restore_gradient_planar(c, frame.f->data[i],
frame.f->linesize[i], avctx->width,
avctx->height, c->slices, 0);
} else {
restore_gradient_planar_il(c, frame.f->data[i],
frame.f->linesize[i],
avctx->width, avctx->height, c->slices,
0);
}
}
}
c->utdsp.restore_rgb_planes(frame.f->data[2], frame.f->data[0], frame.f->data[1],
frame.f->linesize[2], frame.f->linesize[0], frame.f->linesize[1],
avctx->width, avctx->height);
break;
case AV_PIX_FMT_GBRAP10:
case AV_PIX_FMT_GBRP10:
for (i = 0; i < c->planes; i++) {
ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i],
frame.f->linesize[i] / 2, avctx->width,
avctx->height, plane_start[i],
plane_start[i + 1] - 1024,
c->frame_pred == PRED_LEFT);
if (ret)
return ret;
}
c->utdsp.restore_rgb_planes10((uint16_t *)frame.f->data[2], (uint16_t *)frame.f->data[0], (uint16_t *)frame.f->data[1],
frame.f->linesize[2] / 2, frame.f->linesize[0] / 2, frame.f->linesize[1] / 2,
avctx->width, avctx->height);
break;
case AV_PIX_FMT_YUV420P:
for (i = 0; i < 3; i++) {
ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height >> !!i,
plane_start[i], c->frame_pred == PRED_LEFT);
if (ret)
return ret;
if (c->frame_pred == PRED_MEDIAN) {
if (!c->interlaced) {
restore_median_planar(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height >> !!i,
c->slices, !i);
} else {
restore_median_planar_il(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i,
avctx->height >> !!i,
c->slices, !i);
}
} else if (c->frame_pred == PRED_GRADIENT) {
if (!c->interlaced) {
restore_gradient_planar(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height >> !!i,
c->slices, !i);
} else {
restore_gradient_planar_il(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i,
avctx->height >> !!i,
c->slices, !i);
}
}
}
break;
case AV_PIX_FMT_YUV422P:
for (i = 0; i < 3; i++) {
ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height,
plane_start[i], c->frame_pred == PRED_LEFT);
if (ret)
return ret;
if (c->frame_pred == PRED_MEDIAN) {
if (!c->interlaced) {
restore_median_planar(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height,
c->slices, 0);
} else {
restore_median_planar_il(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height,
c->slices, 0);
}
} else if (c->frame_pred == PRED_GRADIENT) {
if (!c->interlaced) {
restore_gradient_planar(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height,
c->slices, 0);
} else {
restore_gradient_planar_il(c, frame.f->data[i], frame.f->linesize[i],
avctx->width >> !!i, avctx->height,
c->slices, 0);
}
}
}
break;
case AV_PIX_FMT_YUV444P:
for (i = 0; i < 3; i++) {
ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i],
avctx->width, avctx->height,
plane_start[i], c->frame_pred == PRED_LEFT);
if (ret)
return ret;
if (c->frame_pred == PRED_MEDIAN) {
if (!c->interlaced) {
restore_median_planar(c, frame.f->data[i], frame.f->linesize[i],
avctx->width, avctx->height,
c->slices, 0);
} else {
restore_median_planar_il(c, frame.f->data[i], frame.f->linesize[i],
avctx->width, avctx->height,
c->slices, 0);
}
} else if (c->frame_pred == PRED_GRADIENT) {
if (!c->interlaced) {
restore_gradient_planar(c, frame.f->data[i], frame.f->linesize[i],
avctx->width, avctx->height,
c->slices, 0);
} else {
restore_gradient_planar_il(c, frame.f->data[i], frame.f->linesize[i],
avctx->width, avctx->height,
c->slices, 0);
}
}
}
break;
case AV_PIX_FMT_YUV420P10:
for (i = 0; i < 3; i++) {
ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], frame.f->linesize[i] / 2,
avctx->width >> !!i, avctx->height >> !!i,
plane_start[i], plane_start[i + 1] - 1024, c->frame_pred == PRED_LEFT);
if (ret)
return ret;
}
break;
case AV_PIX_FMT_YUV422P10:
for (i = 0; i < 3; i++) {
ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], frame.f->linesize[i] / 2,
avctx->width >> !!i, avctx->height,
plane_start[i], plane_start[i + 1] - 1024, c->frame_pred == PRED_LEFT);
if (ret)
return ret;
}
break;
}
frame.f->key_frame = 1;
frame.f->pict_type = AV_PICTURE_TYPE_I;
frame.f->interlaced_frame = !!c->interlaced;
*got_frame = 1;
/* always report that the buffer was completely consumed */
return buf_size;
}
static av_cold int decode_init(AVCodecContext *avctx)
{
UtvideoContext * const c = avctx->priv_data;
int h_shift, v_shift;
c->avctx = avctx;
ff_utvideodsp_init(&c->utdsp);
ff_bswapdsp_init(&c->bdsp);
ff_llviddsp_init(&c->llviddsp);
c->slice_bits_size = 0;
switch (avctx->codec_tag) {
case MKTAG('U', 'L', 'R', 'G'):
c->planes = 3;
avctx->pix_fmt = AV_PIX_FMT_GBRP;
break;
case MKTAG('U', 'L', 'R', 'A'):
c->planes = 4;
avctx->pix_fmt = AV_PIX_FMT_GBRAP;
break;
case MKTAG('U', 'L', 'Y', '0'):
c->planes = 3;
avctx->pix_fmt = AV_PIX_FMT_YUV420P;
avctx->colorspace = AVCOL_SPC_BT470BG;
break;
case MKTAG('U', 'L', 'Y', '2'):
c->planes = 3;
avctx->pix_fmt = AV_PIX_FMT_YUV422P;
avctx->colorspace = AVCOL_SPC_BT470BG;
break;
case MKTAG('U', 'L', 'Y', '4'):
c->planes = 3;
avctx->pix_fmt = AV_PIX_FMT_YUV444P;
avctx->colorspace = AVCOL_SPC_BT470BG;
break;
case MKTAG('U', 'Q', 'Y', '0'):
c->planes = 3;
c->pro = 1;
avctx->pix_fmt = AV_PIX_FMT_YUV420P10;
break;
case MKTAG('U', 'Q', 'Y', '2'):
c->planes = 3;
c->pro = 1;
avctx->pix_fmt = AV_PIX_FMT_YUV422P10;
break;
case MKTAG('U', 'Q', 'R', 'G'):
c->planes = 3;
c->pro = 1;
avctx->pix_fmt = AV_PIX_FMT_GBRP10;
break;
case MKTAG('U', 'Q', 'R', 'A'):
c->planes = 4;
c->pro = 1;
avctx->pix_fmt = AV_PIX_FMT_GBRAP10;
break;
case MKTAG('U', 'L', 'H', '0'):
c->planes = 3;
avctx->pix_fmt = AV_PIX_FMT_YUV420P;
avctx->colorspace = AVCOL_SPC_BT709;
break;
case MKTAG('U', 'L', 'H', '2'):
c->planes = 3;
avctx->pix_fmt = AV_PIX_FMT_YUV422P;
avctx->colorspace = AVCOL_SPC_BT709;
break;
case MKTAG('U', 'L', 'H', '4'):
c->planes = 3;
avctx->pix_fmt = AV_PIX_FMT_YUV444P;
avctx->colorspace = AVCOL_SPC_BT709;
break;
case MKTAG('U', 'M', 'Y', '2'):
c->planes = 3;
c->pack = 1;
avctx->pix_fmt = AV_PIX_FMT_YUV422P;
avctx->colorspace = AVCOL_SPC_BT470BG;
break;
case MKTAG('U', 'M', 'H', '2'):
c->planes = 3;
c->pack = 1;
avctx->pix_fmt = AV_PIX_FMT_YUV422P;
avctx->colorspace = AVCOL_SPC_BT709;
break;
case MKTAG('U', 'M', 'Y', '4'):
c->planes = 3;
c->pack = 1;
avctx->pix_fmt = AV_PIX_FMT_YUV444P;
avctx->colorspace = AVCOL_SPC_BT470BG;
break;
case MKTAG('U', 'M', 'H', '4'):
c->planes = 3;
c->pack = 1;
avctx->pix_fmt = AV_PIX_FMT_YUV444P;
avctx->colorspace = AVCOL_SPC_BT709;
break;
case MKTAG('U', 'M', 'R', 'G'):
c->planes = 3;
c->pack = 1;
avctx->pix_fmt = AV_PIX_FMT_GBRP;
break;
case MKTAG('U', 'M', 'R', 'A'):
c->planes = 4;
c->pack = 1;
avctx->pix_fmt = AV_PIX_FMT_GBRAP;
break;
default:
av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
avctx->codec_tag);
return AVERROR_INVALIDDATA;
}
av_pix_fmt_get_chroma_sub_sample(avctx->pix_fmt, &h_shift, &v_shift);
if ((avctx->width & ((1<<h_shift)-1)) ||
(avctx->height & ((1<<v_shift)-1))) {
avpriv_request_sample(avctx, "Odd dimensions");
return AVERROR_PATCHWELCOME;
}
if (c->pack && avctx->extradata_size >= 16) {
av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
avctx->extradata[3], avctx->extradata[2],
avctx->extradata[1], avctx->extradata[0]);
av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
AV_RB32(avctx->extradata + 4));
c->compression = avctx->extradata[8];
if (c->compression != 2)
avpriv_request_sample(avctx, "Unknown compression type");
c->slices = avctx->extradata[9] + 1;
} else if (!c->pro && avctx->extradata_size >= 16) {
av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
avctx->extradata[3], avctx->extradata[2],
avctx->extradata[1], avctx->extradata[0]);
av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
AV_RB32(avctx->extradata + 4));
c->frame_info_size = AV_RL32(avctx->extradata + 8);
c->flags = AV_RL32(avctx->extradata + 12);
if (c->frame_info_size != 4)
avpriv_request_sample(avctx, "Frame info not 4 bytes");
av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
c->slices = (c->flags >> 24) + 1;
c->compression = c->flags & 1;
c->interlaced = c->flags & 0x800;
} else if (c->pro && avctx->extradata_size == 8) {
av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
avctx->extradata[3], avctx->extradata[2],
avctx->extradata[1], avctx->extradata[0]);
av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
AV_RB32(avctx->extradata + 4));
c->interlaced = 0;
c->frame_info_size = 4;
} else {
av_log(avctx, AV_LOG_ERROR,
"Insufficient extradata size %d, should be at least 16\n",
avctx->extradata_size);
return AVERROR_INVALIDDATA;
}
return 0;
}
static av_cold int decode_end(AVCodecContext *avctx)
{
UtvideoContext * const c = avctx->priv_data;
av_freep(&c->slice_bits);
return 0;
}
const AVCodec ff_utvideo_decoder = {
.name = "utvideo",
.long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
.type = AVMEDIA_TYPE_VIDEO,
.id = AV_CODEC_ID_UTVIDEO,
.priv_data_size = sizeof(UtvideoContext),
.init = decode_init,
.close = decode_end,
.decode = decode_frame,
.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
.caps_internal = FF_CODEC_CAP_INIT_THREADSAFE,
};