• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2002-2014 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * see http://www.pcisys.net/~melanson/codecs/huffyuv.txt for a description of
5  * the algorithm used
6  *
7  * This file is part of FFmpeg.
8  *
9  * FFmpeg is free software; you can redistribute it and/or
10  * modify it under the terms of the GNU Lesser General Public
11  * License as published by the Free Software Foundation; either
12  * version 2.1 of the License, or (at your option) any later version.
13  *
14  * FFmpeg is distributed in the hope that it will be useful,
15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
17  * Lesser General Public License for more details.
18  *
19  * You should have received a copy of the GNU Lesser General Public
20  * License along with FFmpeg; if not, write to the Free Software
21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22  *
23  * yuva, gray, 4:4:4, 4:1:1, 4:1:0 and >8 bit per sample support sponsored by NOA
24  */
25 
26 /**
27  * @file
28  * huffyuv encoder
29  */
30 
31 #include "config_components.h"
32 
33 #include "avcodec.h"
34 #include "codec_internal.h"
35 #include "encode.h"
36 #include "huffyuv.h"
37 #include "huffman.h"
38 #include "huffyuvencdsp.h"
39 #include "lossless_videoencdsp.h"
40 #include "put_bits.h"
41 #include "libavutil/opt.h"
42 #include "libavutil/pixdesc.h"
43 
diff_bytes(HYuvContext * s,uint8_t * dst,const uint8_t * src0,const uint8_t * src1,int w)44 static inline void diff_bytes(HYuvContext *s, uint8_t *dst,
45                               const uint8_t *src0, const uint8_t *src1, int w)
46 {
47     if (s->bps <= 8) {
48         s->llvidencdsp.diff_bytes(dst, src0, src1, w);
49     } else {
50         s->hencdsp.diff_int16((uint16_t *)dst, (const uint16_t *)src0, (const uint16_t *)src1, s->n - 1, w);
51     }
52 }
53 
sub_left_prediction(HYuvContext * s,uint8_t * dst,const uint8_t * src,int w,int left)54 static inline int sub_left_prediction(HYuvContext *s, uint8_t *dst,
55                                       const uint8_t *src, int w, int left)
56 {
57     int i;
58     int min_width = FFMIN(w, 32);
59 
60     if (s->bps <= 8) {
61         for (i = 0; i < min_width; i++) { /* scalar loop before dsp call */
62             const int temp = src[i];
63             dst[i] = temp - left;
64             left   = temp;
65         }
66         if (w < 32)
67             return left;
68         s->llvidencdsp.diff_bytes(dst + 32, src + 32, src + 31, w - 32);
69         return src[w-1];
70     } else {
71         const uint16_t *src16 = (const uint16_t *)src;
72         uint16_t       *dst16 = (      uint16_t *)dst;
73         for (i = 0; i < min_width; i++) { /* scalar loop before dsp call */
74             const int temp = src16[i];
75             dst16[i] = temp - left;
76             left   = temp;
77         }
78         if (w < 32)
79             return left;
80         s->hencdsp.diff_int16(dst16 + 32, src16 + 32, src16 + 31, s->n - 1, w - 32);
81         return src16[w-1];
82     }
83 }
84 
sub_left_prediction_bgr32(HYuvContext * s,uint8_t * dst,const uint8_t * src,int w,int * red,int * green,int * blue,int * alpha)85 static inline void sub_left_prediction_bgr32(HYuvContext *s, uint8_t *dst,
86                                              const uint8_t *src, int w,
87                                              int *red, int *green, int *blue,
88                                              int *alpha)
89 {
90     int i;
91     int r, g, b, a;
92     int min_width = FFMIN(w, 8);
93     r = *red;
94     g = *green;
95     b = *blue;
96     a = *alpha;
97 
98     for (i = 0; i < min_width; i++) {
99         const int rt = src[i * 4 + R];
100         const int gt = src[i * 4 + G];
101         const int bt = src[i * 4 + B];
102         const int at = src[i * 4 + A];
103         dst[i * 4 + R] = rt - r;
104         dst[i * 4 + G] = gt - g;
105         dst[i * 4 + B] = bt - b;
106         dst[i * 4 + A] = at - a;
107         r = rt;
108         g = gt;
109         b = bt;
110         a = at;
111     }
112 
113     s->llvidencdsp.diff_bytes(dst + 32, src + 32, src + 32 - 4, w * 4 - 32);
114 
115     *red   = src[(w - 1) * 4 + R];
116     *green = src[(w - 1) * 4 + G];
117     *blue  = src[(w - 1) * 4 + B];
118     *alpha = src[(w - 1) * 4 + A];
119 }
120 
sub_left_prediction_rgb24(HYuvContext * s,uint8_t * dst,uint8_t * src,int w,int * red,int * green,int * blue)121 static inline void sub_left_prediction_rgb24(HYuvContext *s, uint8_t *dst,
122                                              uint8_t *src, int w,
123                                              int *red, int *green, int *blue)
124 {
125     int i;
126     int r, g, b;
127     r = *red;
128     g = *green;
129     b = *blue;
130     for (i = 0; i < FFMIN(w, 16); i++) {
131         const int rt = src[i * 3 + 0];
132         const int gt = src[i * 3 + 1];
133         const int bt = src[i * 3 + 2];
134         dst[i * 3 + 0] = rt - r;
135         dst[i * 3 + 1] = gt - g;
136         dst[i * 3 + 2] = bt - b;
137         r = rt;
138         g = gt;
139         b = bt;
140     }
141 
142     s->llvidencdsp.diff_bytes(dst + 48, src + 48, src + 48 - 3, w * 3 - 48);
143 
144     *red   = src[(w - 1) * 3 + 0];
145     *green = src[(w - 1) * 3 + 1];
146     *blue  = src[(w - 1) * 3 + 2];
147 }
148 
sub_median_prediction(HYuvContext * s,uint8_t * dst,const uint8_t * src1,const uint8_t * src2,int w,int * left,int * left_top)149 static void sub_median_prediction(HYuvContext *s, uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top)
150 {
151     if (s->bps <= 8) {
152         s->llvidencdsp.sub_median_pred(dst, src1, src2, w , left, left_top);
153     } else {
154         s->hencdsp.sub_hfyu_median_pred_int16((uint16_t *)dst, (const uint16_t *)src1, (const uint16_t *)src2, s->n - 1, w , left, left_top);
155     }
156 }
157 
store_table(HYuvContext * s,const uint8_t * len,uint8_t * buf)158 static int store_table(HYuvContext *s, const uint8_t *len, uint8_t *buf)
159 {
160     int i;
161     int index = 0;
162     int n = s->vlc_n;
163 
164     for (i = 0; i < n;) {
165         int val = len[i];
166         int repeat = 0;
167 
168         for (; i < n && len[i] == val && repeat < 255; i++)
169             repeat++;
170 
171         av_assert0(val < 32 && val >0 && repeat < 256 && repeat>0);
172         if (repeat > 7) {
173             buf[index++] = val;
174             buf[index++] = repeat;
175         } else {
176             buf[index++] = val | (repeat << 5);
177         }
178     }
179 
180     return index;
181 }
182 
store_huffman_tables(HYuvContext * s,uint8_t * buf)183 static int store_huffman_tables(HYuvContext *s, uint8_t *buf)
184 {
185     int i, ret;
186     int size = 0;
187     int count = 3;
188 
189     if (s->version > 2)
190         count = 1 + s->alpha + 2*s->chroma;
191 
192     for (i = 0; i < count; i++) {
193         if ((ret = ff_huff_gen_len_table(s->len[i], s->stats[i], s->vlc_n, 0)) < 0)
194             return ret;
195 
196         if (ff_huffyuv_generate_bits_table(s->bits[i], s->len[i], s->vlc_n) < 0) {
197             return -1;
198         }
199 
200         size += store_table(s, s->len[i], buf + size);
201     }
202     return size;
203 }
204 
encode_init(AVCodecContext * avctx)205 static av_cold int encode_init(AVCodecContext *avctx)
206 {
207     HYuvContext *s = avctx->priv_data;
208     int i, j;
209     int ret;
210     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(avctx->pix_fmt);
211 
212     ff_huffyuv_common_init(avctx);
213     ff_huffyuvencdsp_init(&s->hencdsp, avctx);
214     ff_llvidencdsp_init(&s->llvidencdsp);
215 
216     avctx->extradata = av_mallocz(3*MAX_N + 4);
217     if (s->flags&AV_CODEC_FLAG_PASS1) {
218 #define STATS_OUT_SIZE 21*MAX_N*3 + 4
219         avctx->stats_out = av_mallocz(STATS_OUT_SIZE); // 21*256*3(%llu ) + 3(\n) + 1(0) = 16132
220         if (!avctx->stats_out)
221             return AVERROR(ENOMEM);
222     }
223     s->version = 2;
224 
225     if (!avctx->extradata)
226         return AVERROR(ENOMEM);
227 
228     s->bps = desc->comp[0].depth;
229     s->yuv = !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components >= 2;
230     s->chroma = desc->nb_components > 2;
231     s->alpha = !!(desc->flags & AV_PIX_FMT_FLAG_ALPHA);
232     av_pix_fmt_get_chroma_sub_sample(avctx->pix_fmt,
233                                      &s->chroma_h_shift,
234                                      &s->chroma_v_shift);
235 
236     switch (avctx->pix_fmt) {
237     case AV_PIX_FMT_YUV420P:
238     case AV_PIX_FMT_YUV422P:
239         if (s->width & 1) {
240             av_log(avctx, AV_LOG_ERROR, "Width must be even for this colorspace.\n");
241             return AVERROR(EINVAL);
242         }
243         s->bitstream_bpp = avctx->pix_fmt == AV_PIX_FMT_YUV420P ? 12 : 16;
244         break;
245     case AV_PIX_FMT_YUV444P:
246     case AV_PIX_FMT_YUV410P:
247     case AV_PIX_FMT_YUV411P:
248     case AV_PIX_FMT_YUV440P:
249     case AV_PIX_FMT_GBRP:
250     case AV_PIX_FMT_GBRP9:
251     case AV_PIX_FMT_GBRP10:
252     case AV_PIX_FMT_GBRP12:
253     case AV_PIX_FMT_GBRP14:
254     case AV_PIX_FMT_GBRP16:
255     case AV_PIX_FMT_GRAY8:
256     case AV_PIX_FMT_GRAY16:
257     case AV_PIX_FMT_YUVA444P:
258     case AV_PIX_FMT_YUVA420P:
259     case AV_PIX_FMT_YUVA422P:
260     case AV_PIX_FMT_GBRAP:
261     case AV_PIX_FMT_YUV420P9:
262     case AV_PIX_FMT_YUV420P10:
263     case AV_PIX_FMT_YUV420P12:
264     case AV_PIX_FMT_YUV420P14:
265     case AV_PIX_FMT_YUV420P16:
266     case AV_PIX_FMT_YUV422P9:
267     case AV_PIX_FMT_YUV422P10:
268     case AV_PIX_FMT_YUV422P12:
269     case AV_PIX_FMT_YUV422P14:
270     case AV_PIX_FMT_YUV422P16:
271     case AV_PIX_FMT_YUV444P9:
272     case AV_PIX_FMT_YUV444P10:
273     case AV_PIX_FMT_YUV444P12:
274     case AV_PIX_FMT_YUV444P14:
275     case AV_PIX_FMT_YUV444P16:
276     case AV_PIX_FMT_YUVA420P9:
277     case AV_PIX_FMT_YUVA420P10:
278     case AV_PIX_FMT_YUVA420P16:
279     case AV_PIX_FMT_YUVA422P9:
280     case AV_PIX_FMT_YUVA422P10:
281     case AV_PIX_FMT_YUVA422P16:
282     case AV_PIX_FMT_YUVA444P9:
283     case AV_PIX_FMT_YUVA444P10:
284     case AV_PIX_FMT_YUVA444P16:
285         s->version = 3;
286         break;
287     case AV_PIX_FMT_RGB32:
288         s->bitstream_bpp = 32;
289         break;
290     case AV_PIX_FMT_RGB24:
291         s->bitstream_bpp = 24;
292         break;
293     default:
294         av_log(avctx, AV_LOG_ERROR, "format not supported\n");
295         return AVERROR(EINVAL);
296     }
297     s->n = 1<<s->bps;
298     s->vlc_n = FFMIN(s->n, MAX_VLC_N);
299 
300     avctx->bits_per_coded_sample = s->bitstream_bpp;
301     s->decorrelate = s->bitstream_bpp >= 24 && !s->yuv && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR);
302     s->interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_ME ? 1 : 0;
303     if (s->context) {
304         if (s->flags & (AV_CODEC_FLAG_PASS1 | AV_CODEC_FLAG_PASS2)) {
305             av_log(avctx, AV_LOG_ERROR,
306                    "context=1 is not compatible with "
307                    "2 pass huffyuv encoding\n");
308             return AVERROR(EINVAL);
309         }
310     }
311 
312     if (avctx->codec->id == AV_CODEC_ID_HUFFYUV) {
313         if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
314             av_log(avctx, AV_LOG_ERROR,
315                    "Error: YV12 is not supported by huffyuv; use "
316                    "vcodec=ffvhuff or format=422p\n");
317             return AVERROR(EINVAL);
318         }
319         if (s->interlaced != ( s->height > 288 ))
320             av_log(avctx, AV_LOG_INFO,
321                    "using huffyuv 2.2.0 or newer interlacing flag\n");
322     }
323 
324     if (s->version > 3 && avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
325         av_log(avctx, AV_LOG_ERROR, "Ver > 3 is under development, files encoded with it may not be decodable with future versions!!!\n"
326                "Use vstrict=-2 / -strict -2 to use it anyway.\n");
327         return AVERROR(EINVAL);
328     }
329 
330     if (s->bitstream_bpp >= 24 && s->predictor == MEDIAN && s->version <= 2) {
331         av_log(avctx, AV_LOG_ERROR,
332                "Error: RGB is incompatible with median predictor\n");
333         return AVERROR(EINVAL);
334     }
335 
336     ((uint8_t*)avctx->extradata)[0] = s->predictor | (s->decorrelate << 6);
337     ((uint8_t*)avctx->extradata)[2] = s->interlaced ? 0x10 : 0x20;
338     if (s->context)
339         ((uint8_t*)avctx->extradata)[2] |= 0x40;
340     if (s->version < 3) {
341         ((uint8_t*)avctx->extradata)[1] = s->bitstream_bpp;
342         ((uint8_t*)avctx->extradata)[3] = 0;
343     } else {
344         ((uint8_t*)avctx->extradata)[1] = ((s->bps-1)<<4) | s->chroma_h_shift | (s->chroma_v_shift<<2);
345         if (s->chroma)
346             ((uint8_t*)avctx->extradata)[2] |= s->yuv ? 1 : 2;
347         if (s->alpha)
348             ((uint8_t*)avctx->extradata)[2] |= 4;
349         ((uint8_t*)avctx->extradata)[3] = 1;
350     }
351     s->avctx->extradata_size = 4;
352 
353     if (avctx->stats_in) {
354         char *p = avctx->stats_in;
355 
356         for (i = 0; i < 4; i++)
357             for (j = 0; j < s->vlc_n; j++)
358                 s->stats[i][j] = 1;
359 
360         for (;;) {
361             for (i = 0; i < 4; i++) {
362                 char *next;
363 
364                 for (j = 0; j < s->vlc_n; j++) {
365                     s->stats[i][j] += strtol(p, &next, 0);
366                     if (next == p) return -1;
367                     p = next;
368                 }
369             }
370             if (p[0] == 0 || p[1] == 0 || p[2] == 0) break;
371         }
372     } else {
373         for (i = 0; i < 4; i++)
374             for (j = 0; j < s->vlc_n; j++) {
375                 int d = FFMIN(j, s->vlc_n - j);
376 
377                 s->stats[i][j] = 100000000 / (d*d + 1);
378             }
379     }
380 
381     ret = store_huffman_tables(s, s->avctx->extradata + s->avctx->extradata_size);
382     if (ret < 0)
383         return ret;
384     s->avctx->extradata_size += ret;
385 
386     if (s->context) {
387         for (i = 0; i < 4; i++) {
388             int pels = s->width * s->height / (i ? 40 : 10);
389             for (j = 0; j < s->vlc_n; j++) {
390                 int d = FFMIN(j, s->vlc_n - j);
391                 s->stats[i][j] = pels/(d*d + 1);
392             }
393         }
394     } else {
395         for (i = 0; i < 4; i++)
396             for (j = 0; j < s->vlc_n; j++)
397                 s->stats[i][j]= 0;
398     }
399 
400     if (ff_huffyuv_alloc_temp(s)) {
401         ff_huffyuv_common_end(s);
402         return AVERROR(ENOMEM);
403     }
404 
405     s->picture_number=0;
406 
407     return 0;
408 }
encode_422_bitstream(HYuvContext * s,int offset,int count)409 static int encode_422_bitstream(HYuvContext *s, int offset, int count)
410 {
411     int i;
412     const uint8_t *y = s->temp[0] + offset;
413     const uint8_t *u = s->temp[1] + offset / 2;
414     const uint8_t *v = s->temp[2] + offset / 2;
415 
416     if (put_bytes_left(&s->pb, 0) < 2 * 4 * count) {
417         av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
418         return -1;
419     }
420 
421 #define LOAD4\
422             int y0 = y[2 * i];\
423             int y1 = y[2 * i + 1];\
424             int u0 = u[i];\
425             int v0 = v[i];
426 
427     count /= 2;
428 
429     if (s->flags & AV_CODEC_FLAG_PASS1) {
430         for(i = 0; i < count; i++) {
431             LOAD4;
432             s->stats[0][y0]++;
433             s->stats[1][u0]++;
434             s->stats[0][y1]++;
435             s->stats[2][v0]++;
436         }
437     }
438     if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
439         return 0;
440     if (s->context) {
441         for (i = 0; i < count; i++) {
442             LOAD4;
443             s->stats[0][y0]++;
444             put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);
445             s->stats[1][u0]++;
446             put_bits(&s->pb, s->len[1][u0], s->bits[1][u0]);
447             s->stats[0][y1]++;
448             put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
449             s->stats[2][v0]++;
450             put_bits(&s->pb, s->len[2][v0], s->bits[2][v0]);
451         }
452     } else {
453         for(i = 0; i < count; i++) {
454             LOAD4;
455             put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);
456             put_bits(&s->pb, s->len[1][u0], s->bits[1][u0]);
457             put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
458             put_bits(&s->pb, s->len[2][v0], s->bits[2][v0]);
459         }
460     }
461     return 0;
462 }
463 
encode_plane_bitstream(HYuvContext * s,int width,int plane)464 static int encode_plane_bitstream(HYuvContext *s, int width, int plane)
465 {
466     int i, count = width/2;
467 
468     if (put_bytes_left(&s->pb, 0) < count * s->bps / 2) {
469         av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
470         return -1;
471     }
472 
473 #define LOADEND\
474             int y0 = s->temp[0][width-1];
475 #define LOADEND_14\
476             int y0 = s->temp16[0][width-1] & mask;
477 #define LOADEND_16\
478             int y0 = s->temp16[0][width-1];
479 #define STATEND\
480             s->stats[plane][y0]++;
481 #define STATEND_16\
482             s->stats[plane][y0>>2]++;
483 #define WRITEEND\
484             put_bits(&s->pb, s->len[plane][y0], s->bits[plane][y0]);
485 #define WRITEEND_16\
486             put_bits(&s->pb, s->len[plane][y0>>2], s->bits[plane][y0>>2]);\
487             put_bits(&s->pb, 2, y0&3);
488 
489 #define LOAD2\
490             int y0 = s->temp[0][2 * i];\
491             int y1 = s->temp[0][2 * i + 1];
492 #define LOAD2_14\
493             int y0 = s->temp16[0][2 * i] & mask;\
494             int y1 = s->temp16[0][2 * i + 1] & mask;
495 #define LOAD2_16\
496             int y0 = s->temp16[0][2 * i];\
497             int y1 = s->temp16[0][2 * i + 1];
498 #define STAT2\
499             s->stats[plane][y0]++;\
500             s->stats[plane][y1]++;
501 #define STAT2_16\
502             s->stats[plane][y0>>2]++;\
503             s->stats[plane][y1>>2]++;
504 #define WRITE2\
505             put_bits(&s->pb, s->len[plane][y0], s->bits[plane][y0]);\
506             put_bits(&s->pb, s->len[plane][y1], s->bits[plane][y1]);
507 #define WRITE2_16\
508             put_bits(&s->pb, s->len[plane][y0>>2], s->bits[plane][y0>>2]);\
509             put_bits(&s->pb, 2, y0&3);\
510             put_bits(&s->pb, s->len[plane][y1>>2], s->bits[plane][y1>>2]);\
511             put_bits(&s->pb, 2, y1&3);
512 
513     if (s->bps <= 8) {
514     if (s->flags & AV_CODEC_FLAG_PASS1) {
515         for (i = 0; i < count; i++) {
516             LOAD2;
517             STAT2;
518         }
519         if (width&1) {
520             LOADEND;
521             STATEND;
522         }
523     }
524     if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
525         return 0;
526 
527     if (s->context) {
528         for (i = 0; i < count; i++) {
529             LOAD2;
530             STAT2;
531             WRITE2;
532         }
533         if (width&1) {
534             LOADEND;
535             STATEND;
536             WRITEEND;
537         }
538     } else {
539         for (i = 0; i < count; i++) {
540             LOAD2;
541             WRITE2;
542         }
543         if (width&1) {
544             LOADEND;
545             WRITEEND;
546         }
547     }
548     } else if (s->bps <= 14) {
549         int mask = s->n - 1;
550         if (s->flags & AV_CODEC_FLAG_PASS1) {
551             for (i = 0; i < count; i++) {
552                 LOAD2_14;
553                 STAT2;
554             }
555             if (width&1) {
556                 LOADEND_14;
557                 STATEND;
558             }
559         }
560         if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
561             return 0;
562 
563         if (s->context) {
564             for (i = 0; i < count; i++) {
565                 LOAD2_14;
566                 STAT2;
567                 WRITE2;
568             }
569             if (width&1) {
570                 LOADEND_14;
571                 STATEND;
572                 WRITEEND;
573             }
574         } else {
575             for (i = 0; i < count; i++) {
576                 LOAD2_14;
577                 WRITE2;
578             }
579             if (width&1) {
580                 LOADEND_14;
581                 WRITEEND;
582             }
583         }
584     } else {
585         if (s->flags & AV_CODEC_FLAG_PASS1) {
586             for (i = 0; i < count; i++) {
587                 LOAD2_16;
588                 STAT2_16;
589             }
590             if (width&1) {
591                 LOADEND_16;
592                 STATEND_16;
593             }
594         }
595         if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
596             return 0;
597 
598         if (s->context) {
599             for (i = 0; i < count; i++) {
600                 LOAD2_16;
601                 STAT2_16;
602                 WRITE2_16;
603             }
604             if (width&1) {
605                 LOADEND_16;
606                 STATEND_16;
607                 WRITEEND_16;
608             }
609         } else {
610             for (i = 0; i < count; i++) {
611                 LOAD2_16;
612                 WRITE2_16;
613             }
614             if (width&1) {
615                 LOADEND_16;
616                 WRITEEND_16;
617             }
618         }
619     }
620 #undef LOAD2
621 #undef STAT2
622 #undef WRITE2
623     return 0;
624 }
625 
encode_gray_bitstream(HYuvContext * s,int count)626 static int encode_gray_bitstream(HYuvContext *s, int count)
627 {
628     int i;
629 
630     if (put_bytes_left(&s->pb, 0) < 4 * count) {
631         av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
632         return -1;
633     }
634 
635 #define LOAD2\
636             int y0 = s->temp[0][2 * i];\
637             int y1 = s->temp[0][2 * i + 1];
638 #define STAT2\
639             s->stats[0][y0]++;\
640             s->stats[0][y1]++;
641 #define WRITE2\
642             put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);\
643             put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
644 
645     count /= 2;
646 
647     if (s->flags & AV_CODEC_FLAG_PASS1) {
648         for (i = 0; i < count; i++) {
649             LOAD2;
650             STAT2;
651         }
652     }
653     if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
654         return 0;
655 
656     if (s->context) {
657         for (i = 0; i < count; i++) {
658             LOAD2;
659             STAT2;
660             WRITE2;
661         }
662     } else {
663         for (i = 0; i < count; i++) {
664             LOAD2;
665             WRITE2;
666         }
667     }
668     return 0;
669 }
670 
encode_bgra_bitstream(HYuvContext * s,int count,int planes)671 static inline int encode_bgra_bitstream(HYuvContext *s, int count, int planes)
672 {
673     int i;
674 
675     if (put_bytes_left(&s->pb, 0) < 4 * planes * count) {
676         av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
677         return -1;
678     }
679 
680 #define LOAD_GBRA                                                       \
681     int g = s->temp[0][planes == 3 ? 3 * i + 1 : 4 * i + G];            \
682     int b =(s->temp[0][planes == 3 ? 3 * i + 2 : 4 * i + B] - g) & 0xFF;\
683     int r =(s->temp[0][planes == 3 ? 3 * i + 0 : 4 * i + R] - g) & 0xFF;\
684     int a = s->temp[0][planes * i + A];
685 
686 #define STAT_BGRA                                                       \
687     s->stats[0][b]++;                                                   \
688     s->stats[1][g]++;                                                   \
689     s->stats[2][r]++;                                                   \
690     if (planes == 4)                                                    \
691         s->stats[2][a]++;
692 
693 #define WRITE_GBRA                                                      \
694     put_bits(&s->pb, s->len[1][g], s->bits[1][g]);                      \
695     put_bits(&s->pb, s->len[0][b], s->bits[0][b]);                      \
696     put_bits(&s->pb, s->len[2][r], s->bits[2][r]);                      \
697     if (planes == 4)                                                    \
698         put_bits(&s->pb, s->len[2][a], s->bits[2][a]);
699 
700     if ((s->flags & AV_CODEC_FLAG_PASS1) &&
701         (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)) {
702         for (i = 0; i < count; i++) {
703             LOAD_GBRA;
704             STAT_BGRA;
705         }
706     } else if (s->context || (s->flags & AV_CODEC_FLAG_PASS1)) {
707         for (i = 0; i < count; i++) {
708             LOAD_GBRA;
709             STAT_BGRA;
710             WRITE_GBRA;
711         }
712     } else {
713         for (i = 0; i < count; i++) {
714             LOAD_GBRA;
715             WRITE_GBRA;
716         }
717     }
718     return 0;
719 }
720 
encode_frame(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * pict,int * got_packet)721 static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
722                         const AVFrame *pict, int *got_packet)
723 {
724     HYuvContext *s = avctx->priv_data;
725     const int width = s->width;
726     const int width2 = s->width>>1;
727     const int height = s->height;
728     const int fake_ystride = s->interlaced ? pict->linesize[0]*2  : pict->linesize[0];
729     const int fake_ustride = s->interlaced ? pict->linesize[1]*2  : pict->linesize[1];
730     const int fake_vstride = s->interlaced ? pict->linesize[2]*2  : pict->linesize[2];
731     const AVFrame * const p = pict;
732     int i, j, size = 0, ret;
733 
734     if ((ret = ff_alloc_packet(avctx, pkt, width * height * 3 * 4 + AV_INPUT_BUFFER_MIN_SIZE)) < 0)
735         return ret;
736 
737     if (s->context) {
738         size = store_huffman_tables(s, pkt->data);
739         if (size < 0)
740             return size;
741 
742         for (i = 0; i < 4; i++)
743             for (j = 0; j < s->vlc_n; j++)
744                 s->stats[i][j] >>= 1;
745     }
746 
747     init_put_bits(&s->pb, pkt->data + size, pkt->size - size);
748 
749     if (avctx->pix_fmt == AV_PIX_FMT_YUV422P ||
750         avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
751         int lefty, leftu, leftv, y, cy;
752 
753         put_bits(&s->pb, 8, leftv = p->data[2][0]);
754         put_bits(&s->pb, 8, lefty = p->data[0][1]);
755         put_bits(&s->pb, 8, leftu = p->data[1][0]);
756         put_bits(&s->pb, 8,         p->data[0][0]);
757 
758         lefty = sub_left_prediction(s, s->temp[0], p->data[0], width , 0);
759         leftu = sub_left_prediction(s, s->temp[1], p->data[1], width2, 0);
760         leftv = sub_left_prediction(s, s->temp[2], p->data[2], width2, 0);
761 
762         encode_422_bitstream(s, 2, width-2);
763 
764         if (s->predictor==MEDIAN) {
765             int lefttopy, lefttopu, lefttopv;
766             cy = y = 1;
767             if (s->interlaced) {
768                 lefty = sub_left_prediction(s, s->temp[0], p->data[0] + p->linesize[0], width , lefty);
769                 leftu = sub_left_prediction(s, s->temp[1], p->data[1] + p->linesize[1], width2, leftu);
770                 leftv = sub_left_prediction(s, s->temp[2], p->data[2] + p->linesize[2], width2, leftv);
771 
772                 encode_422_bitstream(s, 0, width);
773                 y++; cy++;
774             }
775 
776             lefty = sub_left_prediction(s, s->temp[0], p->data[0] + fake_ystride, 4, lefty);
777             leftu = sub_left_prediction(s, s->temp[1], p->data[1] + fake_ustride, 2, leftu);
778             leftv = sub_left_prediction(s, s->temp[2], p->data[2] + fake_vstride, 2, leftv);
779 
780             encode_422_bitstream(s, 0, 4);
781 
782             lefttopy = p->data[0][3];
783             lefttopu = p->data[1][1];
784             lefttopv = p->data[2][1];
785             s->llvidencdsp.sub_median_pred(s->temp[0], p->data[0] + 4, p->data[0] + fake_ystride + 4, width  - 4, &lefty, &lefttopy);
786             s->llvidencdsp.sub_median_pred(s->temp[1], p->data[1] + 2, p->data[1] + fake_ustride + 2, width2 - 2, &leftu, &lefttopu);
787             s->llvidencdsp.sub_median_pred(s->temp[2], p->data[2] + 2, p->data[2] + fake_vstride + 2, width2 - 2, &leftv, &lefttopv);
788             encode_422_bitstream(s, 0, width - 4);
789             y++; cy++;
790 
791             for (; y < height; y++,cy++) {
792                 uint8_t *ydst, *udst, *vdst;
793 
794                 if (s->bitstream_bpp == 12) {
795                     while (2 * cy > y) {
796                         ydst = p->data[0] + p->linesize[0] * y;
797                         s->llvidencdsp.sub_median_pred(s->temp[0], ydst - fake_ystride, ydst, width, &lefty, &lefttopy);
798                         encode_gray_bitstream(s, width);
799                         y++;
800                     }
801                     if (y >= height) break;
802                 }
803                 ydst = p->data[0] + p->linesize[0] * y;
804                 udst = p->data[1] + p->linesize[1] * cy;
805                 vdst = p->data[2] + p->linesize[2] * cy;
806 
807                 s->llvidencdsp.sub_median_pred(s->temp[0], ydst - fake_ystride, ydst, width,  &lefty, &lefttopy);
808                 s->llvidencdsp.sub_median_pred(s->temp[1], udst - fake_ustride, udst, width2, &leftu, &lefttopu);
809                 s->llvidencdsp.sub_median_pred(s->temp[2], vdst - fake_vstride, vdst, width2, &leftv, &lefttopv);
810 
811                 encode_422_bitstream(s, 0, width);
812             }
813         } else {
814             for (cy = y = 1; y < height; y++, cy++) {
815                 uint8_t *ydst, *udst, *vdst;
816 
817                 /* encode a luma only line & y++ */
818                 if (s->bitstream_bpp == 12) {
819                     ydst = p->data[0] + p->linesize[0] * y;
820 
821                     if (s->predictor == PLANE && s->interlaced < y) {
822                         s->llvidencdsp.diff_bytes(s->temp[1], ydst, ydst - fake_ystride, width);
823 
824                         lefty = sub_left_prediction(s, s->temp[0], s->temp[1], width , lefty);
825                     } else {
826                         lefty = sub_left_prediction(s, s->temp[0], ydst, width , lefty);
827                     }
828                     encode_gray_bitstream(s, width);
829                     y++;
830                     if (y >= height) break;
831                 }
832 
833                 ydst = p->data[0] + p->linesize[0] * y;
834                 udst = p->data[1] + p->linesize[1] * cy;
835                 vdst = p->data[2] + p->linesize[2] * cy;
836 
837                 if (s->predictor == PLANE && s->interlaced < cy) {
838                     s->llvidencdsp.diff_bytes(s->temp[1],          ydst, ydst - fake_ystride, width);
839                     s->llvidencdsp.diff_bytes(s->temp[2],          udst, udst - fake_ustride, width2);
840                     s->llvidencdsp.diff_bytes(s->temp[2] + width2, vdst, vdst - fake_vstride, width2);
841 
842                     lefty = sub_left_prediction(s, s->temp[0], s->temp[1], width , lefty);
843                     leftu = sub_left_prediction(s, s->temp[1], s->temp[2], width2, leftu);
844                     leftv = sub_left_prediction(s, s->temp[2], s->temp[2] + width2, width2, leftv);
845                 } else {
846                     lefty = sub_left_prediction(s, s->temp[0], ydst, width , lefty);
847                     leftu = sub_left_prediction(s, s->temp[1], udst, width2, leftu);
848                     leftv = sub_left_prediction(s, s->temp[2], vdst, width2, leftv);
849                 }
850 
851                 encode_422_bitstream(s, 0, width);
852             }
853         }
854     } else if(avctx->pix_fmt == AV_PIX_FMT_RGB32) {
855         uint8_t *data = p->data[0] + (height - 1) * p->linesize[0];
856         const int stride = -p->linesize[0];
857         const int fake_stride = -fake_ystride;
858         int y;
859         int leftr, leftg, leftb, lefta;
860 
861         put_bits(&s->pb, 8, lefta = data[A]);
862         put_bits(&s->pb, 8, leftr = data[R]);
863         put_bits(&s->pb, 8, leftg = data[G]);
864         put_bits(&s->pb, 8, leftb = data[B]);
865 
866         sub_left_prediction_bgr32(s, s->temp[0], data + 4, width - 1,
867                                   &leftr, &leftg, &leftb, &lefta);
868         encode_bgra_bitstream(s, width - 1, 4);
869 
870         for (y = 1; y < s->height; y++) {
871             uint8_t *dst = data + y*stride;
872             if (s->predictor == PLANE && s->interlaced < y) {
873                 s->llvidencdsp.diff_bytes(s->temp[1], dst, dst - fake_stride, width * 4);
874                 sub_left_prediction_bgr32(s, s->temp[0], s->temp[1], width,
875                                           &leftr, &leftg, &leftb, &lefta);
876             } else {
877                 sub_left_prediction_bgr32(s, s->temp[0], dst, width,
878                                           &leftr, &leftg, &leftb, &lefta);
879             }
880             encode_bgra_bitstream(s, width, 4);
881         }
882     } else if (avctx->pix_fmt == AV_PIX_FMT_RGB24) {
883         uint8_t *data = p->data[0] + (height - 1) * p->linesize[0];
884         const int stride = -p->linesize[0];
885         const int fake_stride = -fake_ystride;
886         int y;
887         int leftr, leftg, leftb;
888 
889         put_bits(&s->pb, 8, leftr = data[0]);
890         put_bits(&s->pb, 8, leftg = data[1]);
891         put_bits(&s->pb, 8, leftb = data[2]);
892         put_bits(&s->pb, 8, 0);
893 
894         sub_left_prediction_rgb24(s, s->temp[0], data + 3, width - 1,
895                                   &leftr, &leftg, &leftb);
896         encode_bgra_bitstream(s, width-1, 3);
897 
898         for (y = 1; y < s->height; y++) {
899             uint8_t *dst = data + y * stride;
900             if (s->predictor == PLANE && s->interlaced < y) {
901                 s->llvidencdsp.diff_bytes(s->temp[1], dst, dst - fake_stride,
902                                       width * 3);
903                 sub_left_prediction_rgb24(s, s->temp[0], s->temp[1], width,
904                                           &leftr, &leftg, &leftb);
905             } else {
906                 sub_left_prediction_rgb24(s, s->temp[0], dst, width,
907                                           &leftr, &leftg, &leftb);
908             }
909             encode_bgra_bitstream(s, width, 3);
910         }
911     } else if (s->version > 2) {
912         int plane;
913         for (plane = 0; plane < 1 + 2*s->chroma + s->alpha; plane++) {
914             int left, y;
915             int w = width;
916             int h = height;
917             int fake_stride = fake_ystride;
918 
919             if (s->chroma && (plane == 1 || plane == 2)) {
920                 w >>= s->chroma_h_shift;
921                 h >>= s->chroma_v_shift;
922                 fake_stride = plane == 1 ? fake_ustride : fake_vstride;
923             }
924 
925             left = sub_left_prediction(s, s->temp[0], p->data[plane], w , 0);
926 
927             encode_plane_bitstream(s, w, plane);
928 
929             if (s->predictor==MEDIAN) {
930                 int lefttop;
931                 y = 1;
932                 if (s->interlaced) {
933                     left = sub_left_prediction(s, s->temp[0], p->data[plane] + p->linesize[plane], w , left);
934 
935                     encode_plane_bitstream(s, w, plane);
936                     y++;
937                 }
938 
939                 lefttop = p->data[plane][0];
940 
941                 for (; y < h; y++) {
942                     uint8_t *dst = p->data[plane] + p->linesize[plane] * y;
943 
944                     sub_median_prediction(s, s->temp[0], dst - fake_stride, dst, w , &left, &lefttop);
945 
946                     encode_plane_bitstream(s, w, plane);
947                 }
948             } else {
949                 for (y = 1; y < h; y++) {
950                     uint8_t *dst = p->data[plane] + p->linesize[plane] * y;
951 
952                     if (s->predictor == PLANE && s->interlaced < y) {
953                         diff_bytes(s, s->temp[1], dst, dst - fake_stride, w);
954 
955                         left = sub_left_prediction(s, s->temp[0], s->temp[1], w , left);
956                     } else {
957                         left = sub_left_prediction(s, s->temp[0], dst, w , left);
958                     }
959 
960                     encode_plane_bitstream(s, w, plane);
961                 }
962             }
963         }
964     } else {
965         av_log(avctx, AV_LOG_ERROR, "Format not supported!\n");
966     }
967     emms_c();
968 
969     size += (put_bits_count(&s->pb) + 31) / 8;
970     put_bits(&s->pb, 16, 0);
971     put_bits(&s->pb, 15, 0);
972     size /= 4;
973 
974     if ((s->flags & AV_CODEC_FLAG_PASS1) && (s->picture_number & 31) == 0) {
975         int j;
976         char *p = avctx->stats_out;
977         char *end = p + STATS_OUT_SIZE;
978         for (i = 0; i < 4; i++) {
979             for (j = 0; j < s->vlc_n; j++) {
980                 snprintf(p, end-p, "%"PRIu64" ", s->stats[i][j]);
981                 p += strlen(p);
982                 s->stats[i][j]= 0;
983             }
984             snprintf(p, end-p, "\n");
985             p++;
986             if (end <= p)
987                 return AVERROR(ENOMEM);
988         }
989     } else if (avctx->stats_out)
990         avctx->stats_out[0] = '\0';
991     if (!(s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)) {
992         flush_put_bits(&s->pb);
993         s->bdsp.bswap_buf((uint32_t *) pkt->data, (uint32_t *) pkt->data, size);
994     }
995 
996     s->picture_number++;
997 
998     pkt->size   = size * 4;
999     *got_packet = 1;
1000 
1001     return 0;
1002 }
1003 
encode_end(AVCodecContext * avctx)1004 static av_cold int encode_end(AVCodecContext *avctx)
1005 {
1006     HYuvContext *s = avctx->priv_data;
1007 
1008     ff_huffyuv_common_end(s);
1009 
1010     av_freep(&avctx->stats_out);
1011 
1012     return 0;
1013 }
1014 
1015 #define OFFSET(x) offsetof(HYuvContext, x)
1016 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
1017 
1018 #define COMMON_OPTIONS \
1019     { "non_deterministic", "Allow multithreading for e.g. context=1 at the expense of determinism", \
1020       OFFSET(non_determ), AV_OPT_TYPE_BOOL, { .i64 = 0 }, \
1021       0, 1, VE }, \
1022     { "pred", "Prediction method", OFFSET(predictor), AV_OPT_TYPE_INT, { .i64 = LEFT }, LEFT, MEDIAN, VE, "pred" }, \
1023         { "left",   NULL, 0, AV_OPT_TYPE_CONST, { .i64 = LEFT },   INT_MIN, INT_MAX, VE, "pred" }, \
1024         { "plane",  NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PLANE },  INT_MIN, INT_MAX, VE, "pred" }, \
1025         { "median", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MEDIAN }, INT_MIN, INT_MAX, VE, "pred" }, \
1026 
1027 static const AVOption normal_options[] = {
1028     COMMON_OPTIONS
1029     { NULL },
1030 };
1031 
1032 static const AVOption ff_options[] = {
1033     COMMON_OPTIONS
1034     { "context", "Set per-frame huffman tables", OFFSET(context), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
1035     { NULL },
1036 };
1037 
1038 static const AVClass normal_class = {
1039     .class_name = "huffyuv",
1040     .item_name  = av_default_item_name,
1041     .option     = normal_options,
1042     .version    = LIBAVUTIL_VERSION_INT,
1043 };
1044 
1045 static const AVClass ff_class = {
1046     .class_name = "ffvhuff",
1047     .item_name  = av_default_item_name,
1048     .option     = ff_options,
1049     .version    = LIBAVUTIL_VERSION_INT,
1050 };
1051 
1052 const FFCodec ff_huffyuv_encoder = {
1053     .p.name         = "huffyuv",
1054     .p.long_name    = NULL_IF_CONFIG_SMALL("Huffyuv / HuffYUV"),
1055     .p.type         = AVMEDIA_TYPE_VIDEO,
1056     .p.id           = AV_CODEC_ID_HUFFYUV,
1057     .priv_data_size = sizeof(HYuvContext),
1058     .init           = encode_init,
1059     FF_CODEC_ENCODE_CB(encode_frame),
1060     .close          = encode_end,
1061     .p.capabilities = AV_CODEC_CAP_FRAME_THREADS,
1062     .p.priv_class   = &normal_class,
1063     .p.pix_fmts     = (const enum AVPixelFormat[]){
1064         AV_PIX_FMT_YUV422P, AV_PIX_FMT_RGB24,
1065         AV_PIX_FMT_RGB32, AV_PIX_FMT_NONE
1066     },
1067     .caps_internal  = FF_CODEC_CAP_INIT_THREADSAFE |
1068                       FF_CODEC_CAP_INIT_CLEANUP,
1069 };
1070 
1071 #if CONFIG_FFVHUFF_ENCODER
1072 const FFCodec ff_ffvhuff_encoder = {
1073     .p.name         = "ffvhuff",
1074     .p.long_name    = NULL_IF_CONFIG_SMALL("Huffyuv FFmpeg variant"),
1075     .p.type         = AVMEDIA_TYPE_VIDEO,
1076     .p.id           = AV_CODEC_ID_FFVHUFF,
1077     .priv_data_size = sizeof(HYuvContext),
1078     .init           = encode_init,
1079     FF_CODEC_ENCODE_CB(encode_frame),
1080     .close          = encode_end,
1081     .p.capabilities = AV_CODEC_CAP_FRAME_THREADS,
1082     .p.priv_class   = &ff_class,
1083     .p.pix_fmts     = (const enum AVPixelFormat[]){
1084         AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV411P,
1085         AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV440P,
1086         AV_PIX_FMT_GBRP,
1087         AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16,
1088         AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY16,
1089         AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
1090         AV_PIX_FMT_GBRAP,
1091         AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV420P16,
1092         AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV422P16,
1093         AV_PIX_FMT_YUV444P9, AV_PIX_FMT_YUV444P10, AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV444P14, AV_PIX_FMT_YUV444P16,
1094         AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA420P16,
1095         AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA422P16,
1096         AV_PIX_FMT_YUVA444P9, AV_PIX_FMT_YUVA444P10, AV_PIX_FMT_YUVA444P16,
1097         AV_PIX_FMT_RGB24,
1098         AV_PIX_FMT_RGB32, AV_PIX_FMT_NONE
1099     },
1100     .caps_internal  = FF_CODEC_CAP_INIT_THREADSAFE |
1101                       FF_CODEC_CAP_INIT_CLEANUP,
1102 };
1103 #endif
1104