FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/svq1enc.c
Date: 2025-06-01 09:29:47
Exec Total Coverage
Lines: 355 374 94.9%
Functions: 8 8 100.0%
Branches: 159 186 85.5%
Line Branch Exec Source
1 /*
2 * SVQ1 Encoder
3 * Copyright (C) 2004 Mike Melanson <melanson@pcisys.net>
4 *
5 * This file is part of FFmpeg.
6 *
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.
11 *
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.
16 *
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
20 */
21
22 /**
23 * @file
24 * Sorenson Vector Quantizer #1 (SVQ1) video codec.
25 * For more information of the SVQ1 algorithm, visit:
26 * http://www.pcisys.net/~melanson/codecs/
27 */
28
29 #include "libavutil/emms.h"
30 #include "libavutil/mem.h"
31 #include "avcodec.h"
32 #include "codec_internal.h"
33 #include "encode.h"
34 #include "hpeldsp.h"
35 #include "me_cmp.h"
36 #include "mpegvideo.h"
37 #include "h263.h"
38 #include "h263enc.h"
39 #include "internal.h"
40 #include "mpegutils.h"
41 #include "packet_internal.h"
42 #include "put_bits.h"
43 #include "svq1.h"
44 #include "svq1encdsp.h"
45 #include "svq1enc_cb.h"
46 #include "version.h"
47
48 #include "libavutil/avassert.h"
49 #include "libavutil/frame.h"
50 #include "libavutil/mem_internal.h"
51
52 // Workaround for GCC bug 102513
53 #if AV_GCC_VERSION_AT_LEAST(10, 0) && AV_GCC_VERSION_AT_MOST(12, 0) \
54 && !defined(__clang__) && !defined(__INTEL_COMPILER)
55 #pragma GCC optimize ("no-ipa-cp-clone")
56 #endif
57
58 typedef struct SVQ1EncContext {
59 /* FIXME: Needed for motion estimation, should not be used for anything
60 * else, the idea is to make the motion estimation eventually independent
61 * of MPVEncContext, so this will be removed then. */
62 MPVEncContext m;
63 AVCodecContext *avctx;
64 MECmpContext mecc;
65 HpelDSPContext hdsp;
66 AVFrame *current_picture;
67 AVFrame *last_picture;
68
69 /* Some compression statistics */
70 enum AVPictureType pict_type;
71 int quality;
72
73 /* why ooh why this sick breadth first order,
74 * everything is slower and more complex */
75 PutBitContext reorder_pb[6];
76
77 int frame_width;
78 int frame_height;
79
80 /* Y plane block dimensions */
81 int y_block_width;
82 int y_block_height;
83
84 /* U & V plane (C planes) block dimensions */
85 int c_block_width;
86 int c_block_height;
87
88 DECLARE_ALIGNED(16, int16_t, encoded_block_levels)[6][7][256];
89
90 uint16_t *mb_type;
91 uint32_t *dummy;
92 int16_t (*motion_val8[3])[2];
93 int16_t (*motion_val16[3])[2];
94
95 int64_t rd_total;
96
97 uint8_t *scratchbuf;
98
99 SVQ1EncDSPContext svq1encdsp;
100 } SVQ1EncContext;
101
102 200 static void svq1_write_header(SVQ1EncContext *s, PutBitContext *pb, int frame_type)
103 {
104 int i;
105
106 /* frame code */
107 200 put_bits(pb, 22, 0x20);
108
109 /* temporal reference (sure hope this is a "don't care") */
110 200 put_bits(pb, 8, 0x00);
111
112 /* frame type */
113 200 put_bits(pb, 2, frame_type - 1);
114
115
2/2
✓ Branch 0 taken 20 times.
✓ Branch 1 taken 180 times.
 200 if (frame_type == AV_PICTURE_TYPE_I) {
116 /* no checksum since frame code is 0x20 */
117 /* no embedded string either */
118 /* output 5 unknown bits (2 + 2 + 1) */
119 20 put_bits(pb, 5, 2); /* 2 needed by quicktime decoder */
120
121 20 i = ff_match_2uint16(ff_svq1_frame_size_table,
122 FF_ARRAY_ELEMS(ff_svq1_frame_size_table),
123 s->frame_width, s->frame_height);
124 20 put_bits(pb, 3, i);
125
126
2/2
✓ Branch 0 taken 5 times.
✓ Branch 1 taken 15 times.
 20 if (i == 7) {
127 5 put_bits(pb, 12, s->frame_width);
128 5 put_bits(pb, 12, s->frame_height);
129 }
130 }
131
132 /* no checksum or extra data (next 2 bits get 0) */
133 200 put_bits(pb, 2, 0);
134 200 }
135
136 #define QUALITY_THRESHOLD 100
137 #define THRESHOLD_MULTIPLIER 0.6
138
139 4302522 static int encode_block(SVQ1EncContext *s, uint8_t *src, uint8_t *ref,
140 uint8_t *decoded, int stride, unsigned level,
141 int threshold, int lambda, int intra)
142 {
143 int count, y, x, i, j, split, best_mean, best_score, best_count;
144 int best_vector[6];
145 4302522 int block_sum[7] = { 0, 0, 0, 0, 0, 0 };
146 4302522 int w = 2 << (level + 2 >> 1);
147 4302522 int h = 2 << (level + 1 >> 1);
148 4302522 int size = w * h;
149 4302522 int16_t (*block)[256] = s->encoded_block_levels[level];
150 const int8_t *codebook_sum, *codebook;
151 const uint16_t(*mean_vlc)[2];
152 const uint8_t(*multistage_vlc)[2];
153
154 4302522 best_score = 0;
155 // FIXME: Optimize, this does not need to be done multiple times.
156
2/2
✓ Branch 0 taken 643715 times.
✓ Branch 1 taken 3658807 times.
 4302522 if (intra) {
157 // level is 5 when encode_block is called from svq1_encode_plane
158 // and always < 4 when called recursively from this function.
159
2/2
✓ Branch 0 taken 612686 times.
✓ Branch 1 taken 31029 times.
 643715 codebook_sum = level < 4 ? svq1_intra_codebook_sum[level] : NULL;
160 643715 codebook = ff_svq1_intra_codebooks[level];
161 643715 mean_vlc = ff_svq1_intra_mean_vlc;
162 643715 multistage_vlc = ff_svq1_intra_multistage_vlc[level];
163
2/2
✓ Branch 0 taken 2296676 times.
✓ Branch 1 taken 643715 times.
 2940391 for (y = 0; y < h; y++) {
164
2/2
✓ Branch 0 taken 15798160 times.
✓ Branch 1 taken 2296676 times.
 18094836 for (x = 0; x < w; x++) {
165 15798160 int v = src[x + y * stride];
166 15798160 block[0][x + w * y] = v;
167 15798160 best_score += v * v;
168 15798160 block_sum[0] += v;
169 }
170 }
171 } else {
172 // level is 5 or < 4, see above for details.
173
2/2
✓ Branch 0 taken 3483018 times.
✓ Branch 1 taken 175789 times.
 3658807 codebook_sum = level < 4 ? svq1_inter_codebook_sum[level] : NULL;
174 3658807 codebook = ff_svq1_inter_codebooks[level];
175 3658807 mean_vlc = ff_svq1_inter_mean_vlc + 256;
176 3658807 multistage_vlc = ff_svq1_inter_multistage_vlc[level];
177
2/2
✓ Branch 0 taken 13044696 times.
✓ Branch 1 taken 3658807 times.
 16703503 for (y = 0; y < h; y++) {
178
2/2
✓ Branch 0 taken 89655008 times.
✓ Branch 1 taken 13044696 times.
 102699704 for (x = 0; x < w; x++) {
179 89655008 int v = src[x + y * stride] - ref[x + y * stride];
180 89655008 block[0][x + w * y] = v;
181 89655008 best_score += v * v;
182 89655008 block_sum[0] += v;
183 }
184 }
185 }
186
187 4302522 best_count = 0;
188 4302522 best_score -= (int)((unsigned)block_sum[0] * block_sum[0] >> (level + 3));
189 4302522 best_mean = block_sum[0] + (size >> 1) >> (level + 3);
190
191
2/2
✓ Branch 0 taken 4095704 times.
✓ Branch 1 taken 206818 times.
 4302522 if (level < 4) {
192
2/2
✓ Branch 0 taken 24574224 times.
✓ Branch 1 taken 4095704 times.
 28669928 for (count = 1; count < 7; count++) {
193 24574224 int best_vector_score = INT_MAX;
194 24574224 int best_vector_sum = -999, best_vector_mean = -999;
195 24574224 const int stage = count - 1;
196 const int8_t *vector;
197
198
2/2
✓ Branch 0 taken 393187584 times.
✓ Branch 1 taken 24574224 times.
 417761808 for (i = 0; i < 16; i++) {
199 393187584 int sum = codebook_sum[stage * 16 + i];
200 int sqr, diff, score;
201
202 393187584 vector = codebook + stage * size * 16 + i * size;
203 393187584 sqr = s->svq1encdsp.ssd_int8_vs_int16(vector, block[stage], size);
204 393187584 diff = block_sum[stage] - sum;
205 393187584 score = sqr - (diff * (int64_t)diff >> (level + 3)); // FIXME: 64 bits slooow
206
2/2
✓ Branch 0 taken 65426753 times.
✓ Branch 1 taken 327760831 times.
 393187584 if (score < best_vector_score) {
207 65426753 int mean = diff + (size >> 1) >> (level + 3);
208 av_assert2(mean > -300 && mean < 300);
209
2/2
✓ Branch 0 taken 9912177 times.
✓ Branch 1 taken 55514576 times.
 65426753 mean = av_clip(mean, intra ? 0 : -256, 255);
210 65426753 best_vector_score = score;
211 65426753 best_vector[stage] = i;
212 65426753 best_vector_sum = sum;
213 65426753 best_vector_mean = mean;
214 }
215 }
216
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 24574224 times.
 24574224 av_assert0(best_vector_mean != -999);
217 24574224 vector = codebook + stage * size * 16 + best_vector[stage] * size;
218
2/2
✓ Branch 0 taken 420929184 times.
✓ Branch 1 taken 24574224 times.
 445503408 for (j = 0; j < size; j++)
219 420929184 block[stage + 1][j] = block[stage][j] - vector[j];
220 24574224 block_sum[stage + 1] = block_sum[stage] - best_vector_sum;
221 24574224 best_vector_score += lambda *
222 24574224 (+1 + 4 * count +
223 24574224 multistage_vlc[1 + count][1]
224 24574224 + mean_vlc[best_vector_mean][1]);
225
226
2/2
✓ Branch 0 taken 5416629 times.
✓ Branch 1 taken 19157595 times.
 24574224 if (best_vector_score < best_score) {
227 5416629 best_score = best_vector_score;
228 5416629 best_count = count;
229 5416629 best_mean = best_vector_mean;
230 }
231 }
232 }
233
234
2/2
✓ Branch 0 taken 12 times.
✓ Branch 1 taken 4302510 times.
 4302522 if (best_mean == -128)
235 12 best_mean = -127;
236
2/2
✓ Branch 0 taken 6210 times.
✓ Branch 1 taken 4296300 times.
 4302510 else if (best_mean == 128)
237 6210 best_mean = 127;
238
239 4302522 split = 0;
240
4/4
✓ Branch 0 taken 4243230 times.
✓ Branch 1 taken 59292 times.
✓ Branch 2 taken 2116786 times.
✓ Branch 3 taken 2126444 times.
 4302522 if (best_score > threshold && level) {
241 2116786 int score = 0;
242
2/2
✓ Branch 0 taken 1431267 times.
✓ Branch 1 taken 685519 times.
 2116786 int offset = level & 1 ? stride * h / 2 : w / 2;
243 PutBitContext backup[6];
244
245
2/2
✓ Branch 0 taken 3903357 times.
✓ Branch 1 taken 2116786 times.
 6020143 for (i = level - 1; i >= 0; i--)
246 3903357 backup[i] = s->reorder_pb[i];
247 2116786 score += encode_block(s, src, ref, decoded, stride, level - 1,
248 threshold >> 1, lambda, intra);
249 2116786 score += encode_block(s, src + offset, ref + offset, decoded + offset,
250 stride, level - 1, threshold >> 1, lambda, intra);
251 2116786 score += lambda;
252
253
2/2
✓ Branch 0 taken 1359243 times.
✓ Branch 1 taken 757543 times.
 2116786 if (score < best_score) {
254 1359243 best_score = score;
255 1359243 split = 1;
256 } else {
257
2/2
✓ Branch 0 taken 1046147 times.
✓ Branch 1 taken 757543 times.
 1803690 for (i = level - 1; i >= 0; i--)
258 1046147 s->reorder_pb[i] = backup[i];
259 }
260 }
261
2/2
✓ Branch 0 taken 2127616 times.
✓ Branch 1 taken 2174906 times.
 4302522 if (level > 0)
262 2127616 put_bits(&s->reorder_pb[level], 1, split);
263
264
2/2
✓ Branch 0 taken 2943279 times.
✓ Branch 1 taken 1359243 times.
 4302522 if (!split) {
265 av_assert1(best_mean >= 0 && best_mean < 256 || !intra);
266 av_assert1(best_mean >= -256 && best_mean < 256);
267 av_assert1(best_count >= 0 && best_count < 7);
268 av_assert1(level < 4 || best_count == 0);
269
270 /* output the encoding */
271 2943279 put_bits(&s->reorder_pb[level],
272 2943279 multistage_vlc[1 + best_count][1],
273 2943279 multistage_vlc[1 + best_count][0]);
274 2943279 put_bits(&s->reorder_pb[level], mean_vlc[best_mean][1],
275 2943279 mean_vlc[best_mean][0]);
276
277
2/2
✓ Branch 0 taken 3649044 times.
✓ Branch 1 taken 2943279 times.
 6592323 for (i = 0; i < best_count; i++) {
278 av_assert2(best_vector[i] >= 0 && best_vector[i] < 16);
279 3649044 put_bits(&s->reorder_pb[level], 4, best_vector[i]);
280 }
281
282
2/2
✓ Branch 0 taken 7680728 times.
✓ Branch 1 taken 2943279 times.
 10624007 for (y = 0; y < h; y++)
283
2/2
✓ Branch 0 taken 36196848 times.
✓ Branch 1 taken 7680728 times.
 43877576 for (x = 0; x < w; x++)
284 36196848 decoded[x + y * stride] = src[x + y * stride] -
285 36196848 block[best_count][x + w * y] +
286 best_mean;
287 }
288
289 4302522 return best_score;
290 }
291
292 131005 static void init_block_index(MpegEncContext *const s)
293 {
294 131005 s->block_index[0]= s->b8_stride*(s->mb_y*2 ) + s->mb_x*2;
295 131005 s->block_index[1]= s->b8_stride*(s->mb_y*2 ) + 1 + s->mb_x*2;
296 131005 s->block_index[2]= s->b8_stride*(s->mb_y*2 + 1) + s->mb_x*2;
297 131005 s->block_index[3]= s->b8_stride*(s->mb_y*2 + 1) + 1 + s->mb_x*2;
298 131005 s->block_index[4]= s->mb_stride*(s->mb_y + 1) + s->b8_stride*s->mb_height*2 + s->mb_x;
299 131005 s->block_index[5]= s->mb_stride*(s->mb_y + s->mb_height + 2) + s->b8_stride*s->mb_height*2 + s->mb_x;
300 131005 }
301
302 600 static int svq1_encode_plane(SVQ1EncContext *s, int plane,
303 PutBitContext *pb,
304 const unsigned char *src_plane,
305 unsigned char *ref_plane,
306 unsigned char *decoded_plane,
307 int width, int height, int src_stride, int stride)
308 {
309 600 MpegEncContext *const s2 = &s->m.c;
310 int x, y;
311 int i;
312 int block_width, block_height;
313 int level;
314 int threshold[6];
315 600 uint8_t *src = s->scratchbuf + stride * 32;
316 600 const int lambda = (s->quality * s->quality) >>
317 (2 * FF_LAMBDA_SHIFT);
318
319 /* figure out the acceptable level thresholds in advance */
320 600 threshold[5] = QUALITY_THRESHOLD;
321
2/2
✓ Branch 0 taken 3000 times.
✓ Branch 1 taken 600 times.
 3600 for (level = 4; level >= 0; level--)
322 3000 threshold[level] = threshold[level + 1] * THRESHOLD_MULTIPLIER;
323
324 600 block_width = (width + 15) / 16;
325 600 block_height = (height + 15) / 16;
326
327
2/2
✓ Branch 0 taken 540 times.
✓ Branch 1 taken 60 times.
 600 if (s->pict_type == AV_PICTURE_TYPE_P) {
328 540 s2->avctx = s->avctx;
329 540 s2->last_pic.data[0] = ref_plane;
330 540 s2->linesize =
331 540 s2->last_pic.linesize[0] =
332 540 s->m.new_pic->linesize[0] =
333 540 s2->cur_pic.linesize[0] = stride;
334 540 s2->width = width;
335 540 s2->height = height;
336 540 s2->mb_width = block_width;
337 540 s2->mb_height = block_height;
338 540 s2->mb_stride = s2->mb_width + 1;
339 540 s2->b8_stride = 2 * s2->mb_width + 1;
340 540 s->m.f_code = 1;
341 540 s2->pict_type = s->pict_type;
342 540 s->m.me.scene_change_score = 0;
343 // s2->out_format = FMT_H263;
344 // s2->unrestricted_mv = 1;
345 540 s->m.lambda = s->quality;
346 540 s2->qscale = s->m.lambda * 139 +
347 540 FF_LAMBDA_SCALE * 64 >>
348 FF_LAMBDA_SHIFT + 7;
349 540 s->m.lambda2 = s->m.lambda * s->m.lambda +
350 540 FF_LAMBDA_SCALE / 2 >>
351 FF_LAMBDA_SHIFT;
352
353
2/2
✓ Branch 0 taken 12 times.
✓ Branch 1 taken 528 times.
 540 if (!s->motion_val8[plane]) {
354 24 s->motion_val8[plane] = av_mallocz((s2->b8_stride *
355 12 block_height * 2 + 2) *
356 2 * sizeof(int16_t));
357 24 s->motion_val16[plane] = av_mallocz((s2->mb_stride *
358 12 (block_height + 2) + 1) *
359 2 * sizeof(int16_t));
360
2/4
✓ Branch 0 taken 12 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 12 times.
 12 if (!s->motion_val8[plane] || !s->motion_val16[plane])
361 ✗ return AVERROR(ENOMEM);
362 }
363
364 540 s->m.mb_type = s->mb_type;
365
366 // dummies, to avoid segfaults
367 540 s->m.mb_mean = (uint8_t *)s->dummy;
368 540 s->m.mb_var = (uint16_t *)s->dummy;
369 540 s->m.mc_mb_var = (uint16_t *)s->dummy;
370 540 s2->cur_pic.mb_type = s->dummy;
371
372 540 s2->cur_pic.motion_val[0] = s->motion_val8[plane] + 2;
373 540 s->m.p_mv_table = s->motion_val16[plane] +
374 540 s2->mb_stride + 1;
375 540 ff_me_init_pic(&s->m);
376
377 540 s->m.me.dia_size = s->avctx->dia_size;
378 540 s2->first_slice_line = 1;
379
2/2
✓ Branch 0 taken 4005 times.
✓ Branch 1 taken 540 times.
 4545 for (y = 0; y < block_height; y++) {
380 4005 s->m.new_pic->data[0] = src - y * 16 * stride; // ugly
381 4005 s2->mb_y = y;
382
383
4/4
✓ Branch 0 taken 60975 times.
✓ Branch 1 taken 3600 times.
✓ Branch 2 taken 60570 times.
✓ Branch 3 taken 405 times.
 64575 for (i = 0; i < 16 && i + 16 * y < height; i++) {
384 60570 memcpy(&src[i * stride], &src_plane[(i + 16 * y) * src_stride],
385 width);
386
2/2
✓ Branch 0 taken 182700 times.
✓ Branch 1 taken 60570 times.
 243270 for (x = width; x < 16 * block_width; x++)
387 182700 src[i * stride + x] = src[i * stride + x - 1];
388 }
389
3/4
✓ Branch 0 taken 3510 times.
✓ Branch 1 taken 4005 times.
✓ Branch 2 taken 3510 times.
✗ Branch 3 not taken.
 7515 for (; i < 16 && i + 16 * y < 16 * block_height; i++)
390 3510 memcpy(&src[i * stride], &src[(i - 1) * stride],
391 3510 16 * block_width);
392
393
2/2
✓ Branch 0 taken 62055 times.
✓ Branch 1 taken 4005 times.
 66060 for (x = 0; x < block_width; x++) {
394 62055 s2->mb_x = x;
395 62055 init_block_index(s2);
396
397 62055 ff_estimate_p_frame_motion(&s->m, x, y);
398 }
399 4005 s2->first_slice_line = 0;
400 }
401
402 540 ff_fix_long_p_mvs(&s->m, CANDIDATE_MB_TYPE_INTRA);
403 540 ff_fix_long_mvs(&s->m, NULL, 0, s->m.p_mv_table, s->m.f_code,
404 CANDIDATE_MB_TYPE_INTER, 0);
405 }
406
407 600 s2->first_slice_line = 1;
408
2/2
✓ Branch 0 taken 4450 times.
✓ Branch 1 taken 600 times.
 5050 for (y = 0; y < block_height; y++) {
409
4/4
✓ Branch 0 taken 67750 times.
✓ Branch 1 taken 4000 times.
✓ Branch 2 taken 67300 times.
✓ Branch 3 taken 450 times.
 71750 for (i = 0; i < 16 && i + 16 * y < height; i++) {
410 67300 memcpy(&src[i * stride], &src_plane[(i + 16 * y) * src_stride],
411 width);
412
2/2
✓ Branch 0 taken 203000 times.
✓ Branch 1 taken 67300 times.
 270300 for (x = width; x < 16 * block_width; x++)
413 203000 src[i * stride + x] = src[i * stride + x - 1];
414 }
415
3/4
✓ Branch 0 taken 3900 times.
✓ Branch 1 taken 4450 times.
✓ Branch 2 taken 3900 times.
✗ Branch 3 not taken.
 8350 for (; i < 16 && i + 16 * y < 16 * block_height; i++)
416 3900 memcpy(&src[i * stride], &src[(i - 1) * stride], 16 * block_width);
417
418 4450 s2->mb_y = y;
419
2/2
✓ Branch 0 taken 68950 times.
✓ Branch 1 taken 4450 times.
 73400 for (x = 0; x < block_width; x++) {
420 uint8_t reorder_buffer[2][6][7 * 32];
421 int count[2][6];
422 68950 int offset = y * 16 * stride + x * 16;
423 68950 uint8_t *decoded = decoded_plane + offset;
424 68950 const uint8_t *ref = ref_plane + offset;
425 68950 int score[4] = { 0, 0, 0, 0 }, best;
426 68950 uint8_t *temp = s->scratchbuf;
427
428
1/2
✗ Branch 1 not taken.
✓ Branch 2 taken 68950 times.
 68950 if (put_bytes_left(pb, 0) < 3000) { // FIXME: check size
429 ✗ av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
430 ✗ return -1;
431 }
432
433 68950 s2->mb_x = x;
434 68950 init_block_index(s2);
435
436
2/2
✓ Branch 0 taken 62055 times.
✓ Branch 1 taken 6895 times.
 68950 if (s->pict_type == AV_PICTURE_TYPE_I ||
437
2/2
✓ Branch 0 taken 3450 times.
✓ Branch 1 taken 58605 times.
 62055 (s->m.mb_type[x + y * s2->mb_stride] &
438 CANDIDATE_MB_TYPE_INTRA)) {
439
2/2
✓ Branch 0 taken 62070 times.
✓ Branch 1 taken 10345 times.
 72415 for (i = 0; i < 6; i++)
440 62070 init_put_bits(&s->reorder_pb[i], reorder_buffer[0][i],
441 7 * 32);
442
2/2
✓ Branch 0 taken 3450 times.
✓ Branch 1 taken 6895 times.
 10345 if (s->pict_type == AV_PICTURE_TYPE_P) {
443 3450 put_bits(&s->reorder_pb[5], SVQ1_BLOCK_INTRA_LEN, SVQ1_BLOCK_INTRA_CODE);
444 3450 score[0] = SVQ1_BLOCK_INTRA_LEN * lambda;
445 }
446 10345 score[0] += encode_block(s, src + 16 * x, src + 16 * x /* unused */,
447 temp, stride, 5, 64, lambda, 1);
448
2/2
✓ Branch 0 taken 62070 times.
✓ Branch 1 taken 10345 times.
 72415 for (i = 0; i < 6; i++) {
449 62070 count[0][i] = put_bits_count(&s->reorder_pb[i]);
450 62070 flush_put_bits(&s->reorder_pb[i]);
451 }
452 } else
453 58605 score[0] = INT_MAX;
454
455 68950 best = 0;
456
457
2/2
✓ Branch 0 taken 62055 times.
✓ Branch 1 taken 6895 times.
 68950 if (s->pict_type == AV_PICTURE_TYPE_P) {
458 int mx, my, pred_x, pred_y, dxy;
459 int16_t *motion_ptr;
460
461 62055 motion_ptr = ff_h263_pred_motion(s2, 0, 0, &pred_x, &pred_y);
462
2/2
✓ Branch 0 taken 58605 times.
✓ Branch 1 taken 3450 times.
 62055 if (s->m.mb_type[x + y * s2->mb_stride] &
463 CANDIDATE_MB_TYPE_INTER) {
464
2/2
✓ Branch 0 taken 351630 times.
✓ Branch 1 taken 58605 times.
 410235 for (i = 0; i < 6; i++)
465 351630 init_put_bits(&s->reorder_pb[i], reorder_buffer[1][i],
466 7 * 32);
467
468 58605 put_bits(&s->reorder_pb[5], SVQ1_BLOCK_INTER_LEN, SVQ1_BLOCK_INTER_CODE);
469
470 58605 mx = motion_ptr[0];
471 58605 my = motion_ptr[1];
472 av_assert1(mx >= -32 && mx <= 31);
473 av_assert1(my >= -32 && my <= 31);
474 av_assert1(pred_x >= -32 && pred_x <= 31);
475 av_assert1(pred_y >= -32 && pred_y <= 31);
476 58605 ff_h263_encode_motion(&s->reorder_pb[5], mx - pred_x, 1);
477 58605 ff_h263_encode_motion(&s->reorder_pb[5], my - pred_y, 1);
478 58605 score[1] += lambda * put_bits_count(&s->reorder_pb[5]);
479
480 58605 dxy = (mx & 1) + 2 * (my & 1);
481
482 58605 s->hdsp.put_pixels_tab[0][dxy](temp + 16*stride,
483 58605 ref + (mx >> 1) +
484 58605 stride * (my >> 1),
485 stride, 16);
486
487 58605 score[1] += encode_block(s, src + 16 * x, temp + 16*stride,
488 decoded, stride, 5, 64, lambda, 0);
489 58605 best = score[1] <= score[0];
490
491 58605 score[2] = s->mecc.sse[0](NULL, src + 16 * x, ref,
492 stride, 16);
493 58605 score[2] += SVQ1_BLOCK_SKIP_LEN * lambda;
494
6/6
✓ Branch 0 taken 119 times.
✓ Branch 1 taken 58486 times.
✓ Branch 2 taken 88 times.
✓ Branch 3 taken 31 times.
✓ Branch 4 taken 83 times.
✓ Branch 5 taken 5 times.
 58605 if (score[2] < score[best] && mx == 0 && my == 0) {
495 83 best = 2;
496 83 s->hdsp.put_pixels_tab[0][0](decoded, ref, stride, 16);
497 83 put_bits(pb, SVQ1_BLOCK_SKIP_LEN, SVQ1_BLOCK_SKIP_CODE);
498 }
499 }
500
501
2/2
✓ Branch 0 taken 58522 times.
✓ Branch 1 taken 3533 times.
 62055 if (best == 1) {
502
2/2
✓ Branch 0 taken 351132 times.
✓ Branch 1 taken 58522 times.
 409654 for (i = 0; i < 6; i++) {
503 351132 count[1][i] = put_bits_count(&s->reorder_pb[i]);
504 351132 flush_put_bits(&s->reorder_pb[i]);
505 }
506 } else {
507 3533 motion_ptr[0] =
508 3533 motion_ptr[1] =
509 3533 motion_ptr[2] =
510 3533 motion_ptr[3] =
511 3533 motion_ptr[0 + 2 * s2->b8_stride] =
512 3533 motion_ptr[1 + 2 * s2->b8_stride] =
513 3533 motion_ptr[2 + 2 * s2->b8_stride] =
514 3533 motion_ptr[3 + 2 * s2->b8_stride] = 0;
515 }
516 }
517
518 68950 s->rd_total += score[best];
519
520
2/2
✓ Branch 0 taken 68867 times.
✓ Branch 1 taken 83 times.
 68950 if (best != 2)
521
2/2
✓ Branch 0 taken 413202 times.
✓ Branch 1 taken 68867 times.
 482069 for (i = 5; i >= 0; i--)
522 413202 ff_copy_bits(pb, reorder_buffer[best][i],
523 count[best][i]);
524
2/2
✓ Branch 0 taken 10345 times.
✓ Branch 1 taken 58605 times.
 68950 if (best == 0)
525 10345 s->hdsp.put_pixels_tab[0][0](decoded, temp, stride, 16);
526 }
527 4450 s2->first_slice_line = 0;
528 }
529 600 return 0;
530 }
531
532 4 static av_cold int svq1_encode_end(AVCodecContext *avctx)
533 {
534 4 SVQ1EncContext *const s = avctx->priv_data;
535 int i;
536
537
1/2
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
 4 if (avctx->frame_num)
538 4 av_log(avctx, AV_LOG_DEBUG, "RD: %f\n",
539 4 s->rd_total / (double)(avctx->width * avctx->height *
540 4 avctx->frame_num));
541
542 4 av_freep(&s->m.me.scratchpad);
543 4 av_freep(&s->mb_type);
544 4 av_freep(&s->dummy);
545 4 av_freep(&s->scratchbuf);
546
547 4 s->m.mb_type = NULL;
548 4 ff_mpv_common_end(&s->m.c);
549
550
2/2
✓ Branch 0 taken 12 times.
✓ Branch 1 taken 4 times.
 16 for (i = 0; i < 3; i++) {
551 12 av_freep(&s->motion_val8[i]);
552 12 av_freep(&s->motion_val16[i]);
553 }
554
555 4 av_frame_free(&s->current_picture);
556 4 av_frame_free(&s->last_picture);
557 4 av_frame_free(&s->m.new_pic);
558
559 4 return 0;
560 }
561
562 4 static av_cold int write_ident(AVCodecContext *avctx, const char *ident)
563 {
564 4 int size = strlen(ident);
565 4 avctx->extradata = av_malloc(size + 8);
566
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (!avctx->extradata)
567 ✗ return AVERROR(ENOMEM);
568 4 AV_WB32(avctx->extradata, size + 8);
569 4 AV_WL32(avctx->extradata + 4, MKTAG('S', 'V', 'Q', '1'));
570 4 memcpy(avctx->extradata + 8, ident, size);
571 4 avctx->extradata_size = size + 8;
572 4 return 0;
573 }
574
575 4 static av_cold int svq1_encode_init(AVCodecContext *avctx)
576 {
577 4 SVQ1EncContext *const s = avctx->priv_data;
578 int ret;
579
580
2/4
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 4 times.
 4 if (avctx->width >= 4096 || avctx->height >= 4096) {
581 ✗ av_log(avctx, AV_LOG_ERROR, "Dimensions too large, maximum is 4095x4095\n");
582 ✗ return AVERROR(EINVAL);
583 }
584
585 4 ff_hpeldsp_init(&s->hdsp, avctx->flags);
586 4 ff_me_cmp_init(&s->mecc, avctx);
587 4 ret = ff_me_init(&s->m.me, avctx, &s->mecc, 0);
588
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (ret < 0)
589 ✗ return ret;
590 4 ff_mpegvideoencdsp_init(&s->m.mpvencdsp, avctx);
591
592 4 s->current_picture = av_frame_alloc();
593 4 s->last_picture = av_frame_alloc();
594
2/4
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 4 times.
 4 if (!s->current_picture || !s->last_picture) {
595 ✗ return AVERROR(ENOMEM);
596 }
597 4 ret = ff_encode_alloc_frame(avctx, s->current_picture);
598
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (ret < 0)
599 ✗ return ret;
600 4 ret = ff_encode_alloc_frame(avctx, s->last_picture);
601
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (ret < 0)
602 ✗ return ret;
603 4 s->scratchbuf = av_malloc_array(s->current_picture->linesize[0], 16 * 3);
604
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (!s->scratchbuf)
605 ✗ return AVERROR(ENOMEM);
606
607 4 s->frame_width = avctx->width;
608 4 s->frame_height = avctx->height;
609
610 4 s->y_block_width = (s->frame_width + 15) / 16;
611 4 s->y_block_height = (s->frame_height + 15) / 16;
612
613 4 s->c_block_width = (s->frame_width / 4 + 15) / 16;
614 4 s->c_block_height = (s->frame_height / 4 + 15) / 16;
615
616 4 s->avctx = avctx;
617 4 s->m.c.avctx = avctx;
618
619 4 ret = ff_mpv_common_init(&s->m.c);
620
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (ret < 0)
621 ✗ return ret;
622
623 4 s->m.c.picture_structure = PICT_FRAME;
624 4 s->m.me.temp =
625 8 s->m.me.scratchpad = av_mallocz((avctx->width + 64) *
626 4 2 * 16 * 2 * sizeof(uint8_t));
627 8 s->mb_type = av_mallocz((s->y_block_width + 1) *
628 4 s->y_block_height * sizeof(int16_t));
629 8 s->dummy = av_mallocz((s->y_block_width + 1) *
630 4 s->y_block_height * sizeof(int32_t));
631 4 s->m.new_pic = av_frame_alloc();
632
633
1/2
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
 4 if (!s->m.me.scratchpad ||
634
3/6
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 4 times.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
✓ Branch 5 taken 4 times.
 4 !s->mb_type || !s->dummy || !s->m.new_pic)
635 ✗ return AVERROR(ENOMEM);
636
637 4 ff_svq1enc_init(&s->svq1encdsp);
638
639 4 s->m.me.mv_penalty = ff_h263_get_mv_penalty();
640
641
1/2
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
 4 return write_ident(avctx, s->avctx->flags & AV_CODEC_FLAG_BITEXACT ? "Lavc" : LIBAVCODEC_IDENT);
642 }
643
644 200 static int svq1_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
645 const AVFrame *pict, int *got_packet)
646 {
647 200 SVQ1EncContext *const s = avctx->priv_data;
648 PutBitContext pb;
649 int i, ret;
650
651 200 ret = ff_alloc_packet(avctx, pkt, s->y_block_width * s->y_block_height *
652 200 MAX_MB_BYTES * 3 + FF_INPUT_BUFFER_MIN_SIZE);
653
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 200 times.
 200 if (ret < 0)
654 ✗ return ret;
655
656 200 FFSWAP(AVFrame*, s->current_picture, s->last_picture);
657
658
3/4
✓ Branch 0 taken 200 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 180 times.
✓ Branch 3 taken 20 times.
 200 if (avctx->gop_size && (avctx->frame_num % avctx->gop_size))
659 180 s->pict_type = AV_PICTURE_TYPE_P;
660 else
661 20 s->pict_type = AV_PICTURE_TYPE_I;
662 200 s->quality = pict->quality;
663
664 200 ff_side_data_set_encoder_stats(pkt, pict->quality, NULL, 0, s->pict_type);
665
666 200 init_put_bits(&pb, pkt->data, pkt->size);
667 200 svq1_write_header(s, &pb, s->pict_type);
668
2/2
✓ Branch 0 taken 600 times.
✓ Branch 1 taken 200 times.
 800 for (i = 0; i < 3; i++) {
669 1800 int ret = svq1_encode_plane(s, i, &pb,
670 600 pict->data[i],
671 600 s->last_picture->data[i],
672 600 s->current_picture->data[i],
673
2/2
✓ Branch 0 taken 400 times.
✓ Branch 1 taken 200 times.
 600 s->frame_width / (i ? 4 : 1),
674 600 s->frame_height / (i ? 4 : 1),
675 600 pict->linesize[i],
676
2/2
✓ Branch 0 taken 400 times.
✓ Branch 1 taken 200 times.
 600 s->current_picture->linesize[i]);
677 600 emms_c();
678
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 600 times.
 600 if (ret < 0) {
679 int j;
680 ✗ for (j = 0; j < i; j++) {
681 ✗ av_freep(&s->motion_val8[j]);
682 ✗ av_freep(&s->motion_val16[j]);
683 }
684 ✗ av_freep(&s->scratchbuf);
685 ✗ return -1;
686 }
687 }
688
689 // align_put_bits(&pb);
690
2/2
✓ Branch 1 taken 3154 times.
✓ Branch 2 taken 200 times.
 3354 while (put_bits_count(&pb) & 31)
691 3154 put_bits(&pb, 1, 0);
692
693 200 flush_put_bits(&pb);
694
695 200 pkt->size = put_bytes_output(&pb);
696
2/2
✓ Branch 0 taken 20 times.
✓ Branch 1 taken 180 times.
 200 if (s->pict_type == AV_PICTURE_TYPE_I)
697 20 pkt->flags |= AV_PKT_FLAG_KEY;
698 200 *got_packet = 1;
699
700 200 return 0;
701 }
702
703 #define OFFSET(x) offsetof(struct SVQ1EncContext, x)
704 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
705 static const AVOption options[] = {
706 { "motion-est", "Motion estimation algorithm", OFFSET(m.me.motion_est), AV_OPT_TYPE_INT, { .i64 = FF_ME_EPZS }, FF_ME_ZERO, FF_ME_XONE, VE, .unit = "motion-est"},
707 { "zero", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ZERO }, 0, 0, FF_MPV_OPT_FLAGS, .unit = "motion-est" },
708 { "epzs", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_EPZS }, 0, 0, FF_MPV_OPT_FLAGS, .unit = "motion-est" },
709 { "xone", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_XONE }, 0, 0, FF_MPV_OPT_FLAGS, .unit = "motion-est" },
710
711 { NULL },
712 };
713
714 static const AVClass svq1enc_class = {
715 .class_name = "svq1enc",
716 .item_name = av_default_item_name,
717 .option = options,
718 .version = LIBAVUTIL_VERSION_INT,
719 };
720
721 const FFCodec ff_svq1_encoder = {
722 .p.name = "svq1",
723 CODEC_LONG_NAME("Sorenson Vector Quantizer 1 / Sorenson Video 1 / SVQ1"),
724 .p.type = AVMEDIA_TYPE_VIDEO,
725 .p.id = AV_CODEC_ID_SVQ1,
726 .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_ENCODER_REORDERED_OPAQUE,
727 .priv_data_size = sizeof(SVQ1EncContext),
728 .p.priv_class = &svq1enc_class,
729 .init = svq1_encode_init,
730 FF_CODEC_ENCODE_CB(svq1_encode_frame),
731 .close = svq1_encode_end,
732 CODEC_PIXFMTS(AV_PIX_FMT_YUV410P),
733 .color_ranges = AVCOL_RANGE_MPEG,
734 .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
735 };
736