2 * Copyright (c) 2010-2011 Maxim Poliakovski
3 * Copyright (c) 2010-2011 Elvis Presley
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Known FOURCCs: 'apch' (HQ), 'apcn' (SD), 'apcs' (LT), 'acpo' (Proxy), 'ap4h' (4444)
29 #define LONG_BITSTREAM_READER
33 #include "simple_idct.h"
34 #include "proresdec.h"
36 static void permute(uint8_t *dst, const uint8_t *src, const uint8_t permutation[64])
39 for (i = 0; i < 64; i++)
40 dst[i] = permutation[src[i]];
43 static const uint8_t progressive_scan[64] = {
44 0, 1, 8, 9, 2, 3, 10, 11,
45 16, 17, 24, 25, 18, 19, 26, 27,
46 4, 5, 12, 20, 13, 6, 7, 14,
47 21, 28, 29, 22, 15, 23, 30, 31,
48 32, 33, 40, 48, 41, 34, 35, 42,
49 49, 56, 57, 50, 43, 36, 37, 44,
50 51, 58, 59, 52, 45, 38, 39, 46,
51 53, 60, 61, 54, 47, 55, 62, 63
54 static const uint8_t interlaced_scan[64] = {
55 0, 8, 1, 9, 16, 24, 17, 25,
56 2, 10, 3, 11, 18, 26, 19, 27,
57 32, 40, 33, 34, 41, 48, 56, 49,
58 42, 35, 43, 50, 57, 58, 51, 59,
59 4, 12, 5, 6, 13, 20, 28, 21,
60 14, 7, 15, 22, 29, 36, 44, 37,
61 30, 23, 31, 38, 45, 52, 60, 53,
62 46, 39, 47, 54, 61, 62, 55, 63,
65 static av_cold int decode_init(AVCodecContext *avctx)
67 ProresContext *ctx = avctx->priv_data;
68 uint8_t idct_permutation[64];
70 avctx->bits_per_raw_sample = 10;
72 dsputil_init(&ctx->dsp, avctx);
73 ff_proresdsp_init(&ctx->prodsp, avctx);
75 avctx->coded_frame = &ctx->frame;
76 ctx->frame.type = FF_I_TYPE;
77 ctx->frame.key_frame = 1;
79 ff_init_scantable_permutation(idct_permutation,
80 ctx->prodsp.idct_permutation_type);
82 permute(ctx->progressive_scan, progressive_scan, idct_permutation);
83 permute(ctx->interlaced_scan, interlaced_scan, idct_permutation);
88 static int decode_frame_header(ProresContext *ctx, const uint8_t *buf,
89 const int data_size, AVCodecContext *avctx)
91 int hdr_size, width, height, flags;
95 hdr_size = AV_RB16(buf);
96 av_dlog(avctx, "header size %d\n", hdr_size);
97 if (hdr_size > data_size) {
98 av_log(avctx, AV_LOG_ERROR, "error, wrong header size\n");
102 version = AV_RB16(buf + 2);
103 av_dlog(avctx, "%.4s version %d\n", buf+4, version);
105 av_log(avctx, AV_LOG_ERROR, "unsupported version: %d\n", version);
109 width = AV_RB16(buf + 8);
110 height = AV_RB16(buf + 10);
111 if (width != avctx->width || height != avctx->height) {
112 av_log(avctx, AV_LOG_ERROR, "picture resolution change: %dx%d -> %dx%d\n",
113 avctx->width, avctx->height, width, height);
117 ctx->frame_type = (buf[12] >> 2) & 3;
119 av_dlog(avctx, "frame type %d\n", ctx->frame_type);
121 if (ctx->frame_type == 0) {
122 ctx->scan = ctx->progressive_scan; // permuted
124 ctx->scan = ctx->interlaced_scan; // permuted
125 ctx->frame.interlaced_frame = 1;
126 ctx->frame.top_field_first = ctx->frame_type == 1;
129 avctx->pix_fmt = (buf[12] & 0xC0) == 0xC0 ? PIX_FMT_YUV444P10 : PIX_FMT_YUV422P10;
133 av_dlog(avctx, "flags %x\n", flags);
136 permute(ctx->qmat_luma, ctx->prodsp.idct_permutation, ptr);
139 memset(ctx->qmat_luma, 4, 64);
143 permute(ctx->qmat_chroma, ctx->prodsp.idct_permutation, ptr);
145 memset(ctx->qmat_chroma, 4, 64);
151 static int decode_picture_header(AVCodecContext *avctx, const uint8_t *buf, const int buf_size)
153 ProresContext *ctx = avctx->priv_data;
154 int i, hdr_size, slice_count;
155 unsigned pic_data_size;
156 int log2_slice_mb_width, log2_slice_mb_height;
157 int slice_mb_count, mb_x, mb_y;
158 const uint8_t *data_ptr, *index_ptr;
160 hdr_size = buf[0] >> 3;
161 if (hdr_size < 8 || hdr_size > buf_size) {
162 av_log(avctx, AV_LOG_ERROR, "error, wrong picture header size\n");
166 pic_data_size = AV_RB32(buf + 1);
167 if (pic_data_size > buf_size) {
168 av_log(avctx, AV_LOG_ERROR, "error, wrong picture data size\n");
172 log2_slice_mb_width = buf[7] >> 4;
173 log2_slice_mb_height = buf[7] & 0xF;
174 if (log2_slice_mb_width > 3 || log2_slice_mb_height) {
175 av_log(avctx, AV_LOG_ERROR, "unsupported slice resolution: %dx%d\n",
176 1 << log2_slice_mb_width, 1 << log2_slice_mb_height);
180 ctx->mb_width = (avctx->width + 15) >> 4;
182 ctx->mb_height = (avctx->height + 31) >> 5;
184 ctx->mb_height = (avctx->height + 15) >> 4;
186 slice_count = AV_RB16(buf + 5);
188 if (ctx->slice_count != slice_count || !ctx->slices) {
189 av_freep(&ctx->slices);
190 ctx->slices = av_mallocz(slice_count * sizeof(*ctx->slices));
192 return AVERROR(ENOMEM);
193 ctx->slice_count = slice_count;
197 return AVERROR(EINVAL);
199 if (hdr_size + slice_count*2 > buf_size) {
200 av_log(avctx, AV_LOG_ERROR, "error, wrong slice count\n");
204 // parse slice information
205 index_ptr = buf + hdr_size;
206 data_ptr = index_ptr + slice_count*2;
208 slice_mb_count = 1 << log2_slice_mb_width;
212 for (i = 0; i < slice_count; i++) {
213 SliceContext *slice = &ctx->slices[i];
215 slice->data = data_ptr;
216 data_ptr += AV_RB16(index_ptr + i*2);
218 while (ctx->mb_width - mb_x < slice_mb_count)
219 slice_mb_count >>= 1;
223 slice->mb_count = slice_mb_count;
224 slice->data_size = data_ptr - slice->data;
226 if (slice->data_size < 6) {
227 av_log(avctx, AV_LOG_ERROR, "error, wrong slice data size\n");
231 mb_x += slice_mb_count;
232 if (mb_x == ctx->mb_width) {
233 slice_mb_count = 1 << log2_slice_mb_width;
237 if (data_ptr > buf + buf_size) {
238 av_log(avctx, AV_LOG_ERROR, "error, slice out of bounds\n");
243 if (mb_x || mb_y != ctx->mb_height) {
244 av_log(avctx, AV_LOG_ERROR, "error wrong mb count y %d h %d\n",
245 mb_y, ctx->mb_height);
249 return pic_data_size;
252 #define DECODE_CODEWORD(val, codebook) \
254 unsigned int rice_order, exp_order, switch_bits; \
255 unsigned int q, buf, bits; \
257 UPDATE_CACHE(re, gb); \
258 buf = GET_CACHE(re, gb); \
260 /* number of bits to switch between rice and exp golomb */ \
261 switch_bits = codebook & 3; \
262 rice_order = codebook >> 5; \
263 exp_order = (codebook >> 2) & 7; \
265 q = 31 - av_log2(buf); \
267 if (q > switch_bits) { /* exp golomb */ \
268 bits = exp_order - switch_bits + (q<<1); \
269 val = SHOW_UBITS(re, gb, bits) - (1 << exp_order) + \
270 ((switch_bits + 1) << rice_order); \
271 SKIP_BITS(re, gb, bits); \
272 } else if (rice_order) { \
273 SKIP_BITS(re, gb, q+1); \
274 val = (q << rice_order) + SHOW_UBITS(re, gb, rice_order); \
275 SKIP_BITS(re, gb, rice_order); \
278 SKIP_BITS(re, gb, q+1); \
282 #define TOSIGNED(x) (((x) >> 1) ^ (-((x) & 1)))
284 #define FIRST_DC_CB 0xB8
286 static const uint8_t dc_codebook[7] = { 0x04, 0x28, 0x28, 0x4D, 0x4D, 0x70, 0x70};
288 static av_always_inline void decode_dc_coeffs(GetBitContext *gb, DCTELEM *out,
289 int blocks_per_slice)
296 DECODE_CODEWORD(code, FIRST_DC_CB);
297 prev_dc = TOSIGNED(code);
300 out += 64; // dc coeff for the next block
304 for (i = 1; i < blocks_per_slice; i++, out += 64) {
305 DECODE_CODEWORD(code, dc_codebook[FFMIN(code, 6)]);
306 if(code) sign ^= -(code & 1);
308 prev_dc += (((code + 1) >> 1) ^ sign) - sign;
311 CLOSE_READER(re, gb);
314 // adaptive codebook switching lut according to previous run/level values
315 static const uint8_t run_to_cb[16] = { 0x06, 0x06, 0x05, 0x05, 0x04, 0x29, 0x29, 0x29, 0x29, 0x28, 0x28, 0x28, 0x28, 0x28, 0x28, 0x4C };
316 static const uint8_t lev_to_cb[10] = { 0x04, 0x0A, 0x05, 0x06, 0x04, 0x28, 0x28, 0x28, 0x28, 0x4C };
318 static av_always_inline void decode_ac_coeffs(AVCodecContext *avctx, GetBitContext *gb,
319 DCTELEM *out, int blocks_per_slice)
321 ProresContext *ctx = avctx->priv_data;
322 int block_mask, sign;
323 unsigned pos, run, level;
324 int max_coeffs, i, bits_left;
325 int log2_block_count = av_log2(blocks_per_slice);
328 UPDATE_CACHE(re, gb); \
332 max_coeffs = 64 << log2_block_count;
333 block_mask = blocks_per_slice - 1;
335 for (pos = block_mask;;) {
336 bits_left = gb->size_in_bits - re_index;
337 if (!bits_left || (bits_left < 32 && !SHOW_UBITS(re, gb, bits_left)))
340 DECODE_CODEWORD(run, run_to_cb[FFMIN(run, 15)]);
342 if (pos >= max_coeffs) {
343 av_log(avctx, AV_LOG_ERROR, "ac tex damaged %d, %d\n", pos, max_coeffs);
347 DECODE_CODEWORD(level, lev_to_cb[FFMIN(level, 9)]);
350 i = pos >> log2_block_count;
352 sign = SHOW_SBITS(re, gb, 1);
353 SKIP_BITS(re, gb, 1);
354 out[((pos & block_mask) << 6) + ctx->scan[i]] = ((level ^ sign) - sign);
357 CLOSE_READER(re, gb);
360 static void decode_slice_luma(AVCodecContext *avctx, SliceContext *slice,
361 uint8_t *dst, int dst_stride,
362 const uint8_t *buf, unsigned buf_size,
365 ProresContext *ctx = avctx->priv_data;
366 LOCAL_ALIGNED_16(DCTELEM, blocks, [8*4*64]);
369 int i, blocks_per_slice = slice->mb_count<<2;
371 for (i = 0; i < blocks_per_slice; i++)
372 ctx->dsp.clear_block(blocks+(i<<6));
374 init_get_bits(&gb, buf, buf_size << 3);
376 decode_dc_coeffs(&gb, blocks, blocks_per_slice);
377 decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice);
380 for (i = 0; i < slice->mb_count; i++) {
381 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
382 ctx->prodsp.idct_put(dst+16, dst_stride, block+(1<<6), qmat);
383 ctx->prodsp.idct_put(dst+8*dst_stride, dst_stride, block+(2<<6), qmat);
384 ctx->prodsp.idct_put(dst+8*dst_stride+16, dst_stride, block+(3<<6), qmat);
390 static void decode_slice_chroma(AVCodecContext *avctx, SliceContext *slice,
391 uint8_t *dst, int dst_stride,
392 const uint8_t *buf, unsigned buf_size,
393 const int16_t *qmat, int log2_blocks_per_mb)
395 ProresContext *ctx = avctx->priv_data;
396 LOCAL_ALIGNED_16(DCTELEM, blocks, [8*4*64]);
399 int i, j, blocks_per_slice = slice->mb_count << log2_blocks_per_mb;
401 for (i = 0; i < blocks_per_slice; i++)
402 ctx->dsp.clear_block(blocks+(i<<6));
404 init_get_bits(&gb, buf, buf_size << 3);
406 decode_dc_coeffs(&gb, blocks, blocks_per_slice);
407 decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice);
410 for (i = 0; i < slice->mb_count; i++) {
411 for (j = 0; j < log2_blocks_per_mb; j++) {
412 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
413 ctx->prodsp.idct_put(dst+8*dst_stride, dst_stride, block+(1<<6), qmat);
420 static int decode_slice_thread(AVCodecContext *avctx, void *arg, int jobnr, int threadnr)
422 ProresContext *ctx = avctx->priv_data;
423 SliceContext *slice = &ctx->slices[jobnr];
424 const uint8_t *buf = slice->data;
425 AVFrame *pic = avctx->coded_frame;
426 int i, hdr_size, qscale, log2_chroma_blocks_per_mb;
427 int luma_stride, chroma_stride;
428 int y_data_size, u_data_size, v_data_size;
429 uint8_t *dest_y, *dest_u, *dest_v;
430 int16_t qmat_luma_scaled[64];
431 int16_t qmat_chroma_scaled[64];
434 //av_log(avctx, AV_LOG_INFO, "slice %d mb width %d mb x %d y %d\n",
435 // jobnr, slice->mb_count, slice->mb_x, slice->mb_y);
438 hdr_size = buf[0] >> 3;
439 qscale = av_clip(buf[1], 1, 224);
440 qscale = qscale > 128 ? qscale - 96 << 2: qscale;
441 y_data_size = AV_RB16(buf + 2);
442 u_data_size = AV_RB16(buf + 4);
443 v_data_size = slice->data_size - y_data_size - u_data_size - hdr_size;
444 if (hdr_size > 7) v_data_size = AV_RB16(buf + 6);
446 if (y_data_size < 0 || u_data_size < 0 || v_data_size < 0
447 || hdr_size+y_data_size+u_data_size+v_data_size > slice->data_size){
448 av_log(avctx, AV_LOG_ERROR, "invalid plane data size\n");
454 for (i = 0; i < 64; i++) {
455 qmat_luma_scaled [i] = ctx->qmat_luma [i] * qscale;
456 qmat_chroma_scaled[i] = ctx->qmat_chroma[i] * qscale;
459 if (ctx->frame_type == 0) {
460 luma_stride = pic->linesize[0];
461 chroma_stride = pic->linesize[1];
463 luma_stride = pic->linesize[0] << 1;
464 chroma_stride = pic->linesize[1] << 1;
467 if (avctx->pix_fmt == PIX_FMT_YUV444P10) {
469 log2_chroma_blocks_per_mb = 2;
472 log2_chroma_blocks_per_mb = 1;
475 dest_y = pic->data[0] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
476 dest_u = pic->data[1] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
477 dest_v = pic->data[2] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
479 if (ctx->frame_type && ctx->first_field ^ ctx->frame.top_field_first) {
480 dest_y += pic->linesize[0];
481 dest_u += pic->linesize[1];
482 dest_v += pic->linesize[2];
485 decode_slice_luma(avctx, slice, dest_y, luma_stride,
486 buf, y_data_size, qmat_luma_scaled);
488 if (!(avctx->flags & CODEC_FLAG_GRAY)) {
489 decode_slice_chroma(avctx, slice, dest_u, chroma_stride,
490 buf + y_data_size, u_data_size,
491 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
492 decode_slice_chroma(avctx, slice, dest_v, chroma_stride,
493 buf + y_data_size + u_data_size, v_data_size,
494 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
500 static int decode_picture(AVCodecContext *avctx)
502 ProresContext *ctx = avctx->priv_data;
503 int i, threads_ret[ctx->slice_count];
505 avctx->execute2(avctx, decode_slice_thread, NULL, threads_ret, ctx->slice_count);
507 for (i = 0; i < ctx->slice_count; i++)
508 if (threads_ret[i] < 0)
509 return threads_ret[i];
514 static int decode_frame(AVCodecContext *avctx, void *data, int *data_size,
517 ProresContext *ctx = avctx->priv_data;
518 AVFrame *frame = avctx->coded_frame;
519 const uint8_t *buf = avpkt->data;
520 int buf_size = avpkt->size;
521 int frame_hdr_size, pic_size;
523 if (buf_size < 28 || AV_RL32(buf + 4) != AV_RL32("icpf")) {
524 av_log(avctx, AV_LOG_ERROR, "invalid frame header\n");
528 ctx->first_field = 1;
533 frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
534 if (frame_hdr_size < 0)
537 buf += frame_hdr_size;
538 buf_size -= frame_hdr_size;
541 avctx->release_buffer(avctx, frame);
543 if (avctx->get_buffer(avctx, frame) < 0)
547 pic_size = decode_picture_header(avctx, buf, buf_size);
549 av_log(avctx, AV_LOG_ERROR, "error decoding picture header\n");
553 if (decode_picture(avctx)) {
554 av_log(avctx, AV_LOG_ERROR, "error decoding picture\n");
559 buf_size -= pic_size;
561 if (ctx->frame_type && buf_size > 0 && ctx->first_field) {
562 ctx->first_field = 0;
566 *data_size = sizeof(AVFrame);
567 *(AVFrame*)data = *frame;
572 static av_cold int decode_close(AVCodecContext *avctx)
574 ProresContext *ctx = avctx->priv_data;
576 AVFrame *frame = avctx->coded_frame;
578 avctx->release_buffer(avctx, frame);
579 av_freep(&ctx->slices);
584 AVCodec ff_prores_decoder = {
586 .type = AVMEDIA_TYPE_VIDEO,
587 .id = CODEC_ID_PRORES,
588 .priv_data_size = sizeof(ProresContext),
590 .close = decode_close,
591 .decode = decode_frame,
592 .long_name = NULL_IF_CONFIG_SMALL("ProRes"),
593 .capabilities = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS,