Commit | Line | Data |
---|---|---|
2ba45a60 DM |
1 | /* |
2 | * Apple ProRes encoder | |
3 | * | |
4 | * Copyright (c) 2011 Anatoliy Wasserman | |
5 | * | |
6 | * This file is part of FFmpeg. | |
7 | * | |
8 | * FFmpeg is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU Lesser General Public | |
10 | * License as published by the Free Software Foundation; either | |
11 | * version 2.1 of the License, or (at your option) any later version. | |
12 | * | |
13 | * FFmpeg is distributed in the hope that it will be useful, | |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 | * Lesser General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU Lesser General Public | |
19 | * License along with FFmpeg; if not, write to the Free Software | |
20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
21 | */ | |
22 | ||
23 | /** | |
24 | * @file | |
25 | * Apple ProRes encoder (Anatoliy Wasserman version) | |
26 | * Known FOURCCs: 'apch' (HQ), 'apcn' (SD), 'apcs' (LT), 'acpo' (Proxy) | |
27 | */ | |
28 | ||
29 | #include "avcodec.h" | |
30 | #include "dct.h" | |
31 | #include "internal.h" | |
32 | #include "put_bits.h" | |
33 | #include "bytestream.h" | |
34 | #include "fdctdsp.h" | |
35 | ||
36 | #define DEFAULT_SLICE_MB_WIDTH 8 | |
37 | ||
38 | #define FF_PROFILE_PRORES_PROXY 0 | |
39 | #define FF_PROFILE_PRORES_LT 1 | |
40 | #define FF_PROFILE_PRORES_STANDARD 2 | |
41 | #define FF_PROFILE_PRORES_HQ 3 | |
42 | ||
43 | static const AVProfile profiles[] = { | |
44 | { FF_PROFILE_PRORES_PROXY, "apco"}, | |
45 | { FF_PROFILE_PRORES_LT, "apcs"}, | |
46 | { FF_PROFILE_PRORES_STANDARD, "apcn"}, | |
47 | { FF_PROFILE_PRORES_HQ, "apch"}, | |
48 | { FF_PROFILE_UNKNOWN } | |
49 | }; | |
50 | ||
51 | static const int qp_start_table[4] = { 4, 1, 1, 1 }; | |
52 | static const int qp_end_table[4] = { 8, 9, 6, 6 }; | |
53 | static const int bitrate_table[5] = { 1000, 2100, 3500, 5400 }; | |
54 | ||
55 | static const uint8_t progressive_scan[64] = { | |
56 | 0, 1, 8, 9, 2, 3, 10, 11, | |
57 | 16, 17, 24, 25, 18, 19, 26, 27, | |
58 | 4, 5, 12, 20, 13, 6, 7, 14, | |
59 | 21, 28, 29, 22, 15, 23, 30, 31, | |
60 | 32, 33, 40, 48, 41, 34, 35, 42, | |
61 | 49, 56, 57, 50, 43, 36, 37, 44, | |
62 | 51, 58, 59, 52, 45, 38, 39, 46, | |
63 | 53, 60, 61, 54, 47, 55, 62, 63 | |
64 | }; | |
65 | ||
66 | static const uint8_t QMAT_LUMA[4][64] = { | |
67 | { | |
68 | 4, 7, 9, 11, 13, 14, 15, 63, | |
69 | 7, 7, 11, 12, 14, 15, 63, 63, | |
70 | 9, 11, 13, 14, 15, 63, 63, 63, | |
71 | 11, 11, 13, 14, 63, 63, 63, 63, | |
72 | 11, 13, 14, 63, 63, 63, 63, 63, | |
73 | 13, 14, 63, 63, 63, 63, 63, 63, | |
74 | 13, 63, 63, 63, 63, 63, 63, 63, | |
75 | 63, 63, 63, 63, 63, 63, 63, 63 | |
76 | }, { | |
77 | 4, 5, 6, 7, 9, 11, 13, 15, | |
78 | 5, 5, 7, 8, 11, 13, 15, 17, | |
79 | 6, 7, 9, 11, 13, 15, 15, 17, | |
80 | 7, 7, 9, 11, 13, 15, 17, 19, | |
81 | 7, 9, 11, 13, 14, 16, 19, 23, | |
82 | 9, 11, 13, 14, 16, 19, 23, 29, | |
83 | 9, 11, 13, 15, 17, 21, 28, 35, | |
84 | 11, 13, 16, 17, 21, 28, 35, 41 | |
85 | }, { | |
86 | 4, 4, 5, 5, 6, 7, 7, 9, | |
87 | 4, 4, 5, 6, 7, 7, 9, 9, | |
88 | 5, 5, 6, 7, 7, 9, 9, 10, | |
89 | 5, 5, 6, 7, 7, 9, 9, 10, | |
90 | 5, 6, 7, 7, 8, 9, 10, 12, | |
91 | 6, 7, 7, 8, 9, 10, 12, 15, | |
92 | 6, 7, 7, 9, 10, 11, 14, 17, | |
93 | 7, 7, 9, 10, 11, 14, 17, 21 | |
94 | }, { | |
95 | 4, 4, 4, 4, 4, 4, 4, 4, | |
96 | 4, 4, 4, 4, 4, 4, 4, 4, | |
97 | 4, 4, 4, 4, 4, 4, 4, 4, | |
98 | 4, 4, 4, 4, 4, 4, 4, 5, | |
99 | 4, 4, 4, 4, 4, 4, 5, 5, | |
100 | 4, 4, 4, 4, 4, 5, 5, 6, | |
101 | 4, 4, 4, 4, 5, 5, 6, 7, | |
102 | 4, 4, 4, 4, 5, 6, 7, 7 | |
103 | } | |
104 | }; | |
105 | ||
106 | static const uint8_t QMAT_CHROMA[4][64] = { | |
107 | { | |
108 | 4, 7, 9, 11, 13, 14, 63, 63, | |
109 | 7, 7, 11, 12, 14, 63, 63, 63, | |
110 | 9, 11, 13, 14, 63, 63, 63, 63, | |
111 | 11, 11, 13, 14, 63, 63, 63, 63, | |
112 | 11, 13, 14, 63, 63, 63, 63, 63, | |
113 | 13, 14, 63, 63, 63, 63, 63, 63, | |
114 | 13, 63, 63, 63, 63, 63, 63, 63, | |
115 | 63, 63, 63, 63, 63, 63, 63, 63 | |
116 | }, { | |
117 | 4, 5, 6, 7, 9, 11, 13, 15, | |
118 | 5, 5, 7, 8, 11, 13, 15, 17, | |
119 | 6, 7, 9, 11, 13, 15, 15, 17, | |
120 | 7, 7, 9, 11, 13, 15, 17, 19, | |
121 | 7, 9, 11, 13, 14, 16, 19, 23, | |
122 | 9, 11, 13, 14, 16, 19, 23, 29, | |
123 | 9, 11, 13, 15, 17, 21, 28, 35, | |
124 | 11, 13, 16, 17, 21, 28, 35, 41 | |
125 | }, { | |
126 | 4, 4, 5, 5, 6, 7, 7, 9, | |
127 | 4, 4, 5, 6, 7, 7, 9, 9, | |
128 | 5, 5, 6, 7, 7, 9, 9, 10, | |
129 | 5, 5, 6, 7, 7, 9, 9, 10, | |
130 | 5, 6, 7, 7, 8, 9, 10, 12, | |
131 | 6, 7, 7, 8, 9, 10, 12, 15, | |
132 | 6, 7, 7, 9, 10, 11, 14, 17, | |
133 | 7, 7, 9, 10, 11, 14, 17, 21 | |
134 | }, { | |
135 | 4, 4, 4, 4, 4, 4, 4, 4, | |
136 | 4, 4, 4, 4, 4, 4, 4, 4, | |
137 | 4, 4, 4, 4, 4, 4, 4, 4, | |
138 | 4, 4, 4, 4, 4, 4, 4, 5, | |
139 | 4, 4, 4, 4, 4, 4, 5, 5, | |
140 | 4, 4, 4, 4, 4, 5, 5, 6, | |
141 | 4, 4, 4, 4, 5, 5, 6, 7, | |
142 | 4, 4, 4, 4, 5, 6, 7, 7 | |
143 | } | |
144 | }; | |
145 | ||
146 | ||
147 | typedef struct { | |
148 | FDCTDSPContext fdsp; | |
149 | uint8_t* fill_y; | |
150 | uint8_t* fill_u; | |
151 | uint8_t* fill_v; | |
152 | ||
153 | int qmat_luma[16][64]; | |
154 | int qmat_chroma[16][64]; | |
155 | } ProresContext; | |
156 | ||
157 | static void encode_codeword(PutBitContext *pb, int val, int codebook) | |
158 | { | |
159 | unsigned int rice_order, exp_order, switch_bits, first_exp, exp, zeros, | |
160 | mask; | |
161 | ||
162 | /* number of bits to switch between rice and exp golomb */ | |
163 | switch_bits = codebook & 3; | |
164 | rice_order = codebook >> 5; | |
165 | exp_order = (codebook >> 2) & 7; | |
166 | ||
167 | first_exp = ((switch_bits + 1) << rice_order); | |
168 | ||
169 | if (val >= first_exp) { /* exp golomb */ | |
170 | val -= first_exp; | |
171 | val += (1 << exp_order); | |
172 | exp = av_log2(val); | |
173 | zeros = exp - exp_order + switch_bits + 1; | |
174 | put_bits(pb, zeros, 0); | |
175 | put_bits(pb, exp + 1, val); | |
176 | } else if (rice_order) { | |
177 | mask = (1 << rice_order) - 1; | |
178 | put_bits(pb, (val >> rice_order), 0); | |
179 | put_bits(pb, 1, 1); | |
180 | put_bits(pb, rice_order, val & mask); | |
181 | } else { | |
182 | put_bits(pb, val, 0); | |
183 | put_bits(pb, 1, 1); | |
184 | } | |
185 | } | |
186 | ||
187 | #define QSCALE(qmat,ind,val) ((val) / ((qmat)[ind])) | |
188 | #define TO_GOLOMB(val) (((val) << 1) ^ ((val) >> 31)) | |
189 | #define DIFF_SIGN(val, sign) (((val) >> 31) ^ (sign)) | |
190 | #define IS_NEGATIVE(val) ((((val) >> 31) ^ -1) + 1) | |
191 | #define TO_GOLOMB2(val,sign) ((val)==0 ? 0 : ((val) << 1) + (sign)) | |
192 | ||
193 | static av_always_inline int get_level(int val) | |
194 | { | |
195 | int sign = (val >> 31); | |
196 | return (val ^ sign) - sign; | |
197 | } | |
198 | ||
199 | #define FIRST_DC_CB 0xB8 | |
200 | ||
201 | static const uint8_t dc_codebook[7] = { 0x04, 0x28, 0x28, 0x4D, 0x4D, 0x70, 0x70}; | |
202 | ||
203 | static void encode_dc_coeffs(PutBitContext *pb, int16_t *in, | |
204 | int blocks_per_slice, int *qmat) | |
205 | { | |
206 | int prev_dc, code; | |
207 | int i, sign, idx; | |
208 | int new_dc, delta, diff_sign, new_code; | |
209 | ||
210 | prev_dc = QSCALE(qmat, 0, in[0] - 16384); | |
211 | code = TO_GOLOMB(prev_dc); | |
212 | encode_codeword(pb, code, FIRST_DC_CB); | |
213 | ||
214 | code = 5; sign = 0; idx = 64; | |
215 | for (i = 1; i < blocks_per_slice; i++, idx += 64) { | |
216 | new_dc = QSCALE(qmat, 0, in[idx] - 16384); | |
217 | delta = new_dc - prev_dc; | |
218 | diff_sign = DIFF_SIGN(delta, sign); | |
219 | new_code = TO_GOLOMB2(get_level(delta), diff_sign); | |
220 | ||
221 | encode_codeword(pb, new_code, dc_codebook[FFMIN(code, 6)]); | |
222 | ||
223 | code = new_code; | |
224 | sign = delta >> 31; | |
225 | prev_dc = new_dc; | |
226 | } | |
227 | } | |
228 | ||
229 | static const uint8_t run_to_cb[16] = { 0x06, 0x06, 0x05, 0x05, 0x04, 0x29, | |
230 | 0x29, 0x29, 0x29, 0x28, 0x28, 0x28, 0x28, 0x28, 0x28, 0x4C }; | |
231 | static const uint8_t lev_to_cb[10] = { 0x04, 0x0A, 0x05, 0x06, 0x04, 0x28, | |
232 | 0x28, 0x28, 0x28, 0x4C }; | |
233 | ||
234 | static void encode_ac_coeffs(AVCodecContext *avctx, PutBitContext *pb, | |
235 | int16_t *in, int blocks_per_slice, int *qmat) | |
236 | { | |
237 | int prev_run = 4; | |
238 | int prev_level = 2; | |
239 | ||
240 | int run = 0, level, code, i, j; | |
241 | for (i = 1; i < 64; i++) { | |
242 | int indp = progressive_scan[i]; | |
243 | for (j = 0; j < blocks_per_slice; j++) { | |
244 | int val = QSCALE(qmat, indp, in[(j << 6) + indp]); | |
245 | if (val) { | |
246 | encode_codeword(pb, run, run_to_cb[FFMIN(prev_run, 15)]); | |
247 | ||
248 | prev_run = run; | |
249 | run = 0; | |
250 | level = get_level(val); | |
251 | code = level - 1; | |
252 | ||
253 | encode_codeword(pb, code, lev_to_cb[FFMIN(prev_level, 9)]); | |
254 | ||
255 | prev_level = level; | |
256 | ||
257 | put_bits(pb, 1, IS_NEGATIVE(val)); | |
258 | } else { | |
259 | ++run; | |
260 | } | |
261 | } | |
262 | } | |
263 | } | |
264 | ||
265 | static void get(uint8_t *pixels, int stride, int16_t* block) | |
266 | { | |
267 | int i; | |
268 | ||
269 | for (i = 0; i < 8; i++) { | |
270 | AV_WN64(block, AV_RN64(pixels)); | |
271 | AV_WN64(block+4, AV_RN64(pixels+8)); | |
272 | pixels += stride; | |
273 | block += 8; | |
274 | } | |
275 | } | |
276 | ||
277 | static void fdct_get(FDCTDSPContext *fdsp, uint8_t *pixels, int stride, int16_t* block) | |
278 | { | |
279 | get(pixels, stride, block); | |
280 | fdsp->fdct(block); | |
281 | } | |
282 | ||
283 | static int encode_slice_plane(AVCodecContext *avctx, int mb_count, | |
284 | uint8_t *src, int src_stride, uint8_t *buf, unsigned buf_size, | |
285 | int *qmat, int chroma) | |
286 | { | |
287 | ProresContext* ctx = avctx->priv_data; | |
288 | FDCTDSPContext *fdsp = &ctx->fdsp; | |
289 | DECLARE_ALIGNED(16, int16_t, blocks)[DEFAULT_SLICE_MB_WIDTH << 8], *block; | |
290 | int i, blocks_per_slice; | |
291 | PutBitContext pb; | |
292 | ||
293 | block = blocks; | |
294 | for (i = 0; i < mb_count; i++) { | |
295 | fdct_get(fdsp, src, src_stride, block + (0 << 6)); | |
296 | fdct_get(fdsp, src + 8 * src_stride, src_stride, block + ((2 - chroma) << 6)); | |
297 | if (!chroma) { | |
298 | fdct_get(fdsp, src + 16, src_stride, block + (1 << 6)); | |
299 | fdct_get(fdsp, src + 16 + 8 * src_stride, src_stride, block + (3 << 6)); | |
300 | } | |
301 | ||
302 | block += (256 >> chroma); | |
303 | src += (32 >> chroma); | |
304 | } | |
305 | ||
306 | blocks_per_slice = mb_count << (2 - chroma); | |
307 | init_put_bits(&pb, buf, buf_size << 3); | |
308 | ||
309 | encode_dc_coeffs(&pb, blocks, blocks_per_slice, qmat); | |
310 | encode_ac_coeffs(avctx, &pb, blocks, blocks_per_slice, qmat); | |
311 | ||
312 | flush_put_bits(&pb); | |
313 | return put_bits_ptr(&pb) - pb.buf; | |
314 | } | |
315 | ||
316 | static av_always_inline unsigned encode_slice_data(AVCodecContext *avctx, | |
317 | uint8_t *dest_y, uint8_t *dest_u, uint8_t *dest_v, int luma_stride, | |
318 | int chroma_stride, unsigned mb_count, uint8_t *buf, unsigned data_size, | |
319 | unsigned* y_data_size, unsigned* u_data_size, unsigned* v_data_size, | |
320 | int qp) | |
321 | { | |
322 | ProresContext* ctx = avctx->priv_data; | |
323 | ||
324 | *y_data_size = encode_slice_plane(avctx, mb_count, dest_y, luma_stride, | |
325 | buf, data_size, ctx->qmat_luma[qp - 1], 0); | |
326 | ||
327 | if (!(avctx->flags & CODEC_FLAG_GRAY)) { | |
328 | *u_data_size = encode_slice_plane(avctx, mb_count, dest_u, | |
329 | chroma_stride, buf + *y_data_size, data_size - *y_data_size, | |
330 | ctx->qmat_chroma[qp - 1], 1); | |
331 | ||
332 | *v_data_size = encode_slice_plane(avctx, mb_count, dest_v, | |
333 | chroma_stride, buf + *y_data_size + *u_data_size, | |
334 | data_size - *y_data_size - *u_data_size, | |
335 | ctx->qmat_chroma[qp - 1], 1); | |
336 | } | |
337 | ||
338 | return *y_data_size + *u_data_size + *v_data_size; | |
339 | } | |
340 | ||
341 | static void subimage_with_fill(uint16_t *src, unsigned x, unsigned y, | |
342 | unsigned stride, unsigned width, unsigned height, uint16_t *dst, | |
343 | unsigned dst_width, unsigned dst_height) | |
344 | { | |
345 | ||
346 | int box_width = FFMIN(width - x, dst_width); | |
347 | int box_height = FFMIN(height - y, dst_height); | |
348 | int i, j, src_stride = stride >> 1; | |
349 | uint16_t last_pix, *last_line; | |
350 | ||
351 | src += y * src_stride + x; | |
352 | for (i = 0; i < box_height; ++i) { | |
353 | for (j = 0; j < box_width; ++j) { | |
354 | dst[j] = src[j]; | |
355 | } | |
356 | last_pix = dst[j - 1]; | |
357 | for (; j < dst_width; j++) | |
358 | dst[j] = last_pix; | |
359 | src += src_stride; | |
360 | dst += dst_width; | |
361 | } | |
362 | last_line = dst - dst_width; | |
363 | for (; i < dst_height; i++) { | |
364 | for (j = 0; j < dst_width; ++j) { | |
365 | dst[j] = last_line[j]; | |
366 | } | |
367 | dst += dst_width; | |
368 | } | |
369 | } | |
370 | ||
371 | static int encode_slice(AVCodecContext *avctx, const AVFrame *pic, int mb_x, | |
372 | int mb_y, unsigned mb_count, uint8_t *buf, unsigned data_size, | |
373 | int unsafe, int *qp) | |
374 | { | |
375 | int luma_stride, chroma_stride; | |
376 | int hdr_size = 6, slice_size; | |
377 | uint8_t *dest_y, *dest_u, *dest_v; | |
378 | unsigned y_data_size = 0, u_data_size = 0, v_data_size = 0; | |
379 | ProresContext* ctx = avctx->priv_data; | |
380 | int tgt_bits = (mb_count * bitrate_table[avctx->profile]) >> 2; | |
381 | int low_bytes = (tgt_bits - (tgt_bits >> 3)) >> 3; // 12% bitrate fluctuation | |
382 | int high_bytes = (tgt_bits + (tgt_bits >> 3)) >> 3; | |
383 | ||
384 | luma_stride = pic->linesize[0]; | |
385 | chroma_stride = pic->linesize[1]; | |
386 | ||
387 | dest_y = pic->data[0] + (mb_y << 4) * luma_stride + (mb_x << 5); | |
388 | dest_u = pic->data[1] + (mb_y << 4) * chroma_stride + (mb_x << 4); | |
389 | dest_v = pic->data[2] + (mb_y << 4) * chroma_stride + (mb_x << 4); | |
390 | ||
391 | if (unsafe) { | |
392 | ||
393 | subimage_with_fill((uint16_t *) pic->data[0], mb_x << 4, mb_y << 4, | |
394 | luma_stride, avctx->width, avctx->height, | |
395 | (uint16_t *) ctx->fill_y, mb_count << 4, 16); | |
396 | subimage_with_fill((uint16_t *) pic->data[1], mb_x << 3, mb_y << 4, | |
397 | chroma_stride, avctx->width >> 1, avctx->height, | |
398 | (uint16_t *) ctx->fill_u, mb_count << 3, 16); | |
399 | subimage_with_fill((uint16_t *) pic->data[2], mb_x << 3, mb_y << 4, | |
400 | chroma_stride, avctx->width >> 1, avctx->height, | |
401 | (uint16_t *) ctx->fill_v, mb_count << 3, 16); | |
402 | ||
403 | encode_slice_data(avctx, ctx->fill_y, ctx->fill_u, ctx->fill_v, | |
404 | mb_count << 5, mb_count << 4, mb_count, buf + hdr_size, | |
405 | data_size - hdr_size, &y_data_size, &u_data_size, &v_data_size, | |
406 | *qp); | |
407 | } else { | |
408 | slice_size = encode_slice_data(avctx, dest_y, dest_u, dest_v, | |
409 | luma_stride, chroma_stride, mb_count, buf + hdr_size, | |
410 | data_size - hdr_size, &y_data_size, &u_data_size, &v_data_size, | |
411 | *qp); | |
412 | ||
413 | if (slice_size > high_bytes && *qp < qp_end_table[avctx->profile]) { | |
414 | do { | |
415 | *qp += 1; | |
416 | slice_size = encode_slice_data(avctx, dest_y, dest_u, dest_v, | |
417 | luma_stride, chroma_stride, mb_count, buf + hdr_size, | |
418 | data_size - hdr_size, &y_data_size, &u_data_size, | |
419 | &v_data_size, *qp); | |
420 | } while (slice_size > high_bytes && *qp < qp_end_table[avctx->profile]); | |
421 | } else if (slice_size < low_bytes && *qp | |
422 | > qp_start_table[avctx->profile]) { | |
423 | do { | |
424 | *qp -= 1; | |
425 | slice_size = encode_slice_data(avctx, dest_y, dest_u, dest_v, | |
426 | luma_stride, chroma_stride, mb_count, buf + hdr_size, | |
427 | data_size - hdr_size, &y_data_size, &u_data_size, | |
428 | &v_data_size, *qp); | |
429 | } while (slice_size < low_bytes && *qp > qp_start_table[avctx->profile]); | |
430 | } | |
431 | } | |
432 | ||
433 | buf[0] = hdr_size << 3; | |
434 | buf[1] = *qp; | |
435 | AV_WB16(buf + 2, y_data_size); | |
436 | AV_WB16(buf + 4, u_data_size); | |
437 | ||
438 | return hdr_size + y_data_size + u_data_size + v_data_size; | |
439 | } | |
440 | ||
441 | static int prores_encode_picture(AVCodecContext *avctx, const AVFrame *pic, | |
442 | uint8_t *buf, const int buf_size) | |
443 | { | |
444 | int mb_width = (avctx->width + 15) >> 4; | |
445 | int mb_height = (avctx->height + 15) >> 4; | |
446 | int hdr_size, sl_size, i; | |
447 | int mb_y, sl_data_size, qp; | |
448 | int unsafe_bot, unsafe_right; | |
449 | uint8_t *sl_data, *sl_data_sizes; | |
450 | int slice_per_line = 0, rem = mb_width; | |
451 | ||
452 | for (i = av_log2(DEFAULT_SLICE_MB_WIDTH); i >= 0; --i) { | |
453 | slice_per_line += rem >> i; | |
454 | rem &= (1 << i) - 1; | |
455 | } | |
456 | ||
457 | qp = qp_start_table[avctx->profile]; | |
458 | hdr_size = 8; sl_data_size = buf_size - hdr_size; | |
459 | sl_data_sizes = buf + hdr_size; | |
460 | sl_data = sl_data_sizes + (slice_per_line * mb_height * 2); | |
461 | for (mb_y = 0; mb_y < mb_height; mb_y++) { | |
462 | int mb_x = 0; | |
463 | int slice_mb_count = DEFAULT_SLICE_MB_WIDTH; | |
464 | while (mb_x < mb_width) { | |
465 | while (mb_width - mb_x < slice_mb_count) | |
466 | slice_mb_count >>= 1; | |
467 | ||
468 | unsafe_bot = (avctx->height & 0xf) && (mb_y == mb_height - 1); | |
469 | unsafe_right = (avctx->width & 0xf) && (mb_x + slice_mb_count == mb_width); | |
470 | ||
471 | sl_size = encode_slice(avctx, pic, mb_x, mb_y, slice_mb_count, | |
472 | sl_data, sl_data_size, unsafe_bot || unsafe_right, &qp); | |
473 | ||
474 | bytestream_put_be16(&sl_data_sizes, sl_size); | |
475 | sl_data += sl_size; | |
476 | sl_data_size -= sl_size; | |
477 | mb_x += slice_mb_count; | |
478 | } | |
479 | } | |
480 | ||
481 | buf[0] = hdr_size << 3; | |
482 | AV_WB32(buf + 1, sl_data - buf); | |
483 | AV_WB16(buf + 5, slice_per_line * mb_height); | |
484 | buf[7] = av_log2(DEFAULT_SLICE_MB_WIDTH) << 4; | |
485 | ||
486 | return sl_data - buf; | |
487 | } | |
488 | ||
489 | static int prores_encode_frame(AVCodecContext *avctx, AVPacket *pkt, | |
490 | const AVFrame *pict, int *got_packet) | |
491 | { | |
492 | int header_size = 148; | |
493 | uint8_t *buf; | |
494 | int pic_size, ret; | |
495 | int frame_size = FFALIGN(avctx->width, 16) * FFALIGN(avctx->height, 16)*16 + 500 + FF_MIN_BUFFER_SIZE; //FIXME choose tighter limit | |
496 | ||
497 | ||
498 | if ((ret = ff_alloc_packet2(avctx, pkt, frame_size + FF_MIN_BUFFER_SIZE)) < 0) | |
499 | return ret; | |
500 | ||
501 | buf = pkt->data; | |
502 | pic_size = prores_encode_picture(avctx, pict, buf + header_size + 8, | |
503 | pkt->size - header_size - 8); | |
504 | ||
505 | bytestream_put_be32(&buf, pic_size + 8 + header_size); | |
506 | bytestream_put_buffer(&buf, "icpf", 4); | |
507 | ||
508 | bytestream_put_be16(&buf, header_size); | |
509 | bytestream_put_be16(&buf, 0); | |
510 | bytestream_put_buffer(&buf, "fmpg", 4); | |
511 | bytestream_put_be16(&buf, avctx->width); | |
512 | bytestream_put_be16(&buf, avctx->height); | |
513 | *buf++ = 0x83; // {10}(422){00}{00}(frame){11} | |
514 | *buf++ = 0; | |
515 | *buf++ = 2; | |
516 | *buf++ = 2; | |
517 | *buf++ = 6; | |
518 | *buf++ = 32; | |
519 | *buf++ = 0; | |
520 | *buf++ = 3; | |
521 | ||
522 | bytestream_put_buffer(&buf, QMAT_LUMA[avctx->profile], 64); | |
523 | bytestream_put_buffer(&buf, QMAT_CHROMA[avctx->profile], 64); | |
524 | ||
525 | pkt->flags |= AV_PKT_FLAG_KEY; | |
526 | pkt->size = pic_size + 8 + header_size; | |
527 | *got_packet = 1; | |
528 | ||
529 | return 0; | |
530 | } | |
531 | ||
532 | static void scale_mat(const uint8_t* src, int* dst, int scale) | |
533 | { | |
534 | int i; | |
535 | for (i = 0; i < 64; i++) | |
536 | dst[i] = src[i] * scale; | |
537 | } | |
538 | ||
539 | static av_cold int prores_encode_init(AVCodecContext *avctx) | |
540 | { | |
541 | int i; | |
542 | ProresContext* ctx = avctx->priv_data; | |
543 | ||
544 | if (avctx->pix_fmt != AV_PIX_FMT_YUV422P10) { | |
545 | av_log(avctx, AV_LOG_ERROR, "need YUV422P10\n"); | |
546 | return -1; | |
547 | } | |
548 | avctx->bits_per_raw_sample = 10; | |
549 | ||
550 | if (avctx->width & 0x1) { | |
551 | av_log(avctx, AV_LOG_ERROR, | |
552 | "frame width needs to be multiple of 2\n"); | |
553 | return -1; | |
554 | } | |
555 | ||
556 | if (avctx->width > 65534 || avctx->height > 65535) { | |
557 | av_log(avctx, AV_LOG_ERROR, | |
558 | "The maximum dimensions are 65534x65535\n"); | |
559 | return AVERROR(EINVAL); | |
560 | } | |
561 | ||
562 | if ((avctx->height & 0xf) || (avctx->width & 0xf)) { | |
563 | ctx->fill_y = av_malloc(4 * (DEFAULT_SLICE_MB_WIDTH << 8)); | |
564 | if (!ctx->fill_y) | |
565 | return AVERROR(ENOMEM); | |
566 | ctx->fill_u = ctx->fill_y + (DEFAULT_SLICE_MB_WIDTH << 9); | |
567 | ctx->fill_v = ctx->fill_u + (DEFAULT_SLICE_MB_WIDTH << 8); | |
568 | } | |
569 | ||
570 | if (avctx->profile == FF_PROFILE_UNKNOWN) { | |
571 | avctx->profile = FF_PROFILE_PRORES_STANDARD; | |
572 | av_log(avctx, AV_LOG_INFO, | |
573 | "encoding with ProRes standard (apcn) profile\n"); | |
574 | ||
575 | } else if (avctx->profile < FF_PROFILE_PRORES_PROXY | |
576 | || avctx->profile > FF_PROFILE_PRORES_HQ) { | |
577 | av_log( | |
578 | avctx, | |
579 | AV_LOG_ERROR, | |
580 | "unknown profile %d, use [0 - apco, 1 - apcs, 2 - apcn (default), 3 - apch]\n", | |
581 | avctx->profile); | |
582 | return -1; | |
583 | } | |
584 | ||
585 | ff_fdctdsp_init(&ctx->fdsp, avctx); | |
586 | ||
587 | avctx->codec_tag = AV_RL32((const uint8_t*)profiles[avctx->profile].name); | |
588 | ||
589 | for (i = 1; i <= 16; i++) { | |
590 | scale_mat(QMAT_LUMA[avctx->profile] , ctx->qmat_luma[i - 1] , i); | |
591 | scale_mat(QMAT_CHROMA[avctx->profile], ctx->qmat_chroma[i - 1], i); | |
592 | } | |
593 | ||
594 | avctx->coded_frame = av_frame_alloc(); | |
595 | avctx->coded_frame->key_frame = 1; | |
596 | avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I; | |
597 | ||
598 | return 0; | |
599 | } | |
600 | ||
601 | static av_cold int prores_encode_close(AVCodecContext *avctx) | |
602 | { | |
603 | ProresContext* ctx = avctx->priv_data; | |
604 | av_freep(&avctx->coded_frame); | |
605 | av_freep(&ctx->fill_y); | |
606 | ||
607 | return 0; | |
608 | } | |
609 | ||
610 | AVCodec ff_prores_aw_encoder = { | |
611 | .name = "prores_aw", | |
612 | .long_name = NULL_IF_CONFIG_SMALL("Apple ProRes"), | |
613 | .type = AVMEDIA_TYPE_VIDEO, | |
614 | .id = AV_CODEC_ID_PRORES, | |
615 | .priv_data_size = sizeof(ProresContext), | |
616 | .init = prores_encode_init, | |
617 | .close = prores_encode_close, | |
618 | .encode2 = prores_encode_frame, | |
619 | .pix_fmts = (const enum AVPixelFormat[]){AV_PIX_FMT_YUV422P10, AV_PIX_FMT_NONE}, | |
620 | .capabilities = CODEC_CAP_FRAME_THREADS | CODEC_CAP_INTRA_ONLY, | |
621 | .profiles = profiles | |
622 | }; | |
623 | ||
624 | AVCodec ff_prores_encoder = { | |
625 | .name = "prores", | |
626 | .long_name = NULL_IF_CONFIG_SMALL("Apple ProRes"), | |
627 | .type = AVMEDIA_TYPE_VIDEO, | |
628 | .id = AV_CODEC_ID_PRORES, | |
629 | .priv_data_size = sizeof(ProresContext), | |
630 | .init = prores_encode_init, | |
631 | .close = prores_encode_close, | |
632 | .encode2 = prores_encode_frame, | |
633 | .pix_fmts = (const enum AVPixelFormat[]){AV_PIX_FMT_YUV422P10, AV_PIX_FMT_NONE}, | |
634 | .capabilities = CODEC_CAP_FRAME_THREADS | CODEC_CAP_INTRA_ONLY, | |
635 | .profiles = profiles | |
636 | }; |