Invoke future RV30/40 loop filter for already decoded rows instead of
[libav.git] / libavcodec / rv40.c
CommitLineData
ae4ca617
KS
1/*
2 * RV40 decoder
3 * Copyright (c) 2007 Konstantin Shishkov
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 rv40.c
24 * RV40 decoder
25 */
26
27#include "avcodec.h"
28#include "dsputil.h"
29#include "mpegvideo.h"
4978618b 30#include "golomb.h"
ae4ca617
KS
31
32#include "rv34.h"
33#include "rv40vlc2.h"
34#include "rv40data.h"
35
36static VLC aic_top_vlc;
37static VLC aic_mode1_vlc[AIC_MODE1_NUM], aic_mode2_vlc[AIC_MODE2_NUM];
38static VLC ptype_vlc[NUM_PTYPE_VLCS], btype_vlc[NUM_BTYPE_VLCS];
39
40/**
41 * Initialize all tables.
42 */
98a6fff9 43static av_cold void rv40_init_tables()
ae4ca617
KS
44{
45 int i;
46
47 init_vlc(&aic_top_vlc, AIC_TOP_BITS, AIC_TOP_SIZE,
48 rv40_aic_top_vlc_bits, 1, 1,
49 rv40_aic_top_vlc_codes, 1, 1, INIT_VLC_USE_STATIC);
50 for(i = 0; i < AIC_MODE1_NUM; i++){
51 // Every tenth VLC table is empty
52 if((i % 10) == 9) continue;
53 init_vlc(&aic_mode1_vlc[i], AIC_MODE1_BITS, AIC_MODE1_SIZE,
54 aic_mode1_vlc_bits[i], 1, 1,
55 aic_mode1_vlc_codes[i], 1, 1, INIT_VLC_USE_STATIC);
56 }
57 for(i = 0; i < AIC_MODE2_NUM; i++){
58 init_vlc(&aic_mode2_vlc[i], AIC_MODE2_BITS, AIC_MODE2_SIZE,
59 aic_mode2_vlc_bits[i], 1, 1,
60 aic_mode2_vlc_codes[i], 2, 2, INIT_VLC_USE_STATIC);
61 }
62 for(i = 0; i < NUM_PTYPE_VLCS; i++)
63 init_vlc_sparse(&ptype_vlc[i], PTYPE_VLC_BITS, PTYPE_VLC_SIZE,
64 ptype_vlc_bits[i], 1, 1,
65 ptype_vlc_codes[i], 1, 1,
66 ptype_vlc_syms, 1, 1, INIT_VLC_USE_STATIC);
67 for(i = 0; i < NUM_BTYPE_VLCS; i++)
68 init_vlc_sparse(&btype_vlc[i], BTYPE_VLC_BITS, BTYPE_VLC_SIZE,
69 btype_vlc_bits[i], 1, 1,
70 btype_vlc_codes[i], 1, 1,
71 btype_vlc_syms, 1, 1, INIT_VLC_USE_STATIC);
72}
73
74/**
75 * Get stored dimension from bitstream.
76 *
77 * If the width/height is the standard one then it's coded as a 3-bit index.
78 * Otherwise it is coded as escaped 8-bit portions.
79 */
80static int get_dimension(GetBitContext *gb, const int *dim)
81{
82 int t = get_bits(gb, 3);
83 int val = dim[t];
84 if(val < 0)
85 val = dim[get_bits1(gb) - val];
86 if(!val){
87 do{
88 t = get_bits(gb, 8);
89 val += t << 2;
90 }while(t == 0xFF);
91 }
92 return val;
93}
94
95/**
96 * Get encoded picture size - usually this is called from rv40_parse_slice_header.
97 */
98static void rv40_parse_picture_size(GetBitContext *gb, int *w, int *h)
99{
100 *w = get_dimension(gb, rv40_standard_widths);
101 *h = get_dimension(gb, rv40_standard_heights);
102}
103
104static int rv40_parse_slice_header(RV34DecContext *r, GetBitContext *gb, SliceInfo *si)
105{
39902a8c 106 int mb_bits;
ae4ca617
KS
107 int w = r->s.width, h = r->s.height;
108 int mb_size;
109
110 memset(si, 0, sizeof(SliceInfo));
111 if(get_bits1(gb))
112 return -1;
113 si->type = get_bits(gb, 2);
114 if(si->type == 1) si->type = 0;
115 si->quant = get_bits(gb, 5);
116 if(get_bits(gb, 2))
117 return -1;
118 si->vlc_set = get_bits(gb, 2);
119 skip_bits1(gb);
39902a8c 120 si->pts = get_bits(gb, 13);
ae4ca617
KS
121 if(!si->type || !get_bits1(gb))
122 rv40_parse_picture_size(gb, &w, &h);
b192e564
KS
123 if(avcodec_check_dimensions(r->s.avctx, w, h) < 0)
124 return -1;
ae4ca617
KS
125 si->width = w;
126 si->height = h;
127 mb_size = ((w + 15) >> 4) * ((h + 15) >> 4);
128 mb_bits = ff_rv34_get_start_offset(gb, mb_size);
129 si->start = get_bits(gb, mb_bits);
130
131 return 0;
132}
133
134/**
135 * Decode 4x4 intra types array.
136 */
ea61e33a 137static int rv40_decode_intra_types(RV34DecContext *r, GetBitContext *gb, int8_t *dst)
ae4ca617
KS
138{
139 MpegEncContext *s = &r->s;
140 int i, j, k, v;
141 int A, B, C;
142 int pattern;
ea61e33a 143 int8_t *ptr;
ae4ca617
KS
144
145 for(i = 0; i < 4; i++, dst += s->b4_stride){
146 if(!i && s->first_slice_line){
147 pattern = get_vlc2(gb, aic_top_vlc.table, AIC_TOP_BITS, 1);
148 dst[0] = (pattern >> 2) & 2;
149 dst[1] = (pattern >> 1) & 2;
150 dst[2] = pattern & 2;
151 dst[3] = (pattern << 1) & 2;
152 continue;
153 }
154 ptr = dst;
155 for(j = 0; j < 4; j++){
156 /* Coefficients are read using VLC chosen by the prediction pattern
157 * The first one (used for retrieving a pair of coefficients) is
158 * constructed from the top, top right and left coefficients
159 * The second one (used for retrieving only one coefficient) is
160 * top + 10 * left.
161 */
162 A = ptr[-s->b4_stride + 1]; // it won't be used for the last coefficient in a row
163 B = ptr[-s->b4_stride];
164 C = ptr[-1];
165 pattern = A + (B << 4) + (C << 8);
166 for(k = 0; k < MODE2_PATTERNS_NUM; k++)
167 if(pattern == rv40_aic_table_index[k])
168 break;
169 if(j < 3 && k < MODE2_PATTERNS_NUM){ //pattern is found, decoding 2 coefficients
170 v = get_vlc2(gb, aic_mode2_vlc[k].table, AIC_MODE2_BITS, 2);
171 *ptr++ = v/9;
172 *ptr++ = v%9;
173 j++;
174 }else{
175 if(B != -1 && C != -1)
176 v = get_vlc2(gb, aic_mode1_vlc[B + C*10].table, AIC_MODE1_BITS, 1);
177 else{ // tricky decoding
178 v = 0;
179 switch(C){
180 case -1: // code 0 -> 1, 1 -> 0
181 if(B < 2)
182 v = get_bits1(gb) ^ 1;
183 break;
184 case 0:
185 case 2: // code 0 -> 2, 1 -> 0
186 v = (get_bits1(gb) ^ 1) << 1;
187 break;
188 }
189 }
190 *ptr++ = v;
191 }
192 }
193 }
194 return 0;
195}
196
197/**
198 * Decode macroblock information.
199 */
200static int rv40_decode_mb_info(RV34DecContext *r)
201{
202 MpegEncContext *s = &r->s;
203 GetBitContext *gb = &s->gb;
204 int q, i;
205 int prev_type = 0;
206 int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
207 int blocks[RV34_MB_TYPES] = {0};
208 int count = 0;
209
210 if(!r->s.mb_skip_run)
4978618b 211 r->s.mb_skip_run = svq3_get_ue_golomb(gb) + 1;
ae4ca617
KS
212
213 if(--r->s.mb_skip_run)
214 return RV34_MB_SKIP;
215
6cd1500d 216 if(r->avail_cache[5-1])
ae4ca617 217 blocks[r->mb_type[mb_pos - 1]]++;
6cd1500d 218 if(r->avail_cache[5-4]){
ae4ca617 219 blocks[r->mb_type[mb_pos - s->mb_stride]]++;
6cd1500d 220 if(r->avail_cache[5-2])
ae4ca617 221 blocks[r->mb_type[mb_pos - s->mb_stride + 1]]++;
6cd1500d 222 if(r->avail_cache[5-5])
ae4ca617
KS
223 blocks[r->mb_type[mb_pos - s->mb_stride - 1]]++;
224 }
225
226 for(i = 0; i < RV34_MB_TYPES; i++){
227 if(blocks[i] > count){
228 count = blocks[i];
229 prev_type = i;
230 }
231 }
9701840b 232 if(s->pict_type == FF_P_TYPE){
ae4ca617
KS
233 prev_type = block_num_to_ptype_vlc_num[prev_type];
234 q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);
235 if(q < PBTYPE_ESCAPE)
236 return q;
237 q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);
238 av_log(s->avctx, AV_LOG_ERROR, "Dquant for P-frame\n");
239 }else{
240 prev_type = block_num_to_btype_vlc_num[prev_type];
241 q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1);
242 if(q < PBTYPE_ESCAPE)
243 return q;
244 q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1);
245 av_log(s->avctx, AV_LOG_ERROR, "Dquant for B-frame\n");
246 }
247 return 0;
248}
249
70f42303
KS
250#define CLIP_SYMM(a, b) av_clip(a, -(b), b)
251/**
252 * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1
253 */
254static inline void rv40_weak_loop_filter(uint8_t *src, const int step,
255 const int filter_p1, const int filter_q1,
256 const int alpha, const int beta,
257 const int lim_p0q0,
258 const int lim_q1, const int lim_p1,
259 const int diff_p1p0, const int diff_q1q0,
260 const int diff_p1p2, const int diff_q1q2)
261{
262 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
263 int t, u, diff;
264
265 t = src[0*step] - src[-1*step];
266 if(!t)
267 return;
268 u = (alpha * FFABS(t)) >> 7;
269 if(u > 3 - (filter_p1 && filter_q1))
270 return;
271
272 t <<= 2;
273 if(filter_p1 && filter_q1)
274 t += src[-2*step] - src[1*step];
275 diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);
276 src[-1*step] = cm[src[-1*step] + diff];
277 src[ 0*step] = cm[src[ 0*step] - diff];
278 if(FFABS(diff_p1p2) <= beta && filter_p1){
279 t = (diff_p1p0 + diff_p1p2 - diff) >> 1;
280 src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];
281 }
282 if(FFABS(diff_q1q2) <= beta && filter_q1){
283 t = (diff_q1q0 + diff_q1q2 + diff) >> 1;
284 src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];
285 }
286}
287
e9d0fdef
KS
288static inline void rv40_adaptive_loop_filter(uint8_t *src, const int step,
289 const int stride, const int dmode,
290 const int lim_q1, const int lim_p1,
291 const int alpha,
292 const int beta, const int beta2,
293 const int chroma, const int edge)
294{
295 int diff_p1p0[4], diff_q1q0[4], diff_p1p2[4], diff_q1q2[4];
296 int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;
297 uint8_t *ptr;
298 int flag_strong0 = 1, flag_strong1 = 1;
299 int filter_p1, filter_q1;
300 int i;
301 int lims;
302
303 for(i = 0, ptr = src; i < 4; i++, ptr += stride){
304 diff_p1p0[i] = ptr[-2*step] - ptr[-1*step];
305 diff_q1q0[i] = ptr[ 1*step] - ptr[ 0*step];
306 sum_p1p0 += diff_p1p0[i];
307 sum_q1q0 += diff_q1q0[i];
308 }
309 filter_p1 = FFABS(sum_p1p0) < (beta<<2);
310 filter_q1 = FFABS(sum_q1q0) < (beta<<2);
311 if(!filter_p1 && !filter_q1)
312 return;
313
314 for(i = 0, ptr = src; i < 4; i++, ptr += stride){
315 diff_p1p2[i] = ptr[-2*step] - ptr[-3*step];
316 diff_q1q2[i] = ptr[ 1*step] - ptr[ 2*step];
317 sum_p1p2 += diff_p1p2[i];
318 sum_q1q2 += diff_q1q2[i];
319 }
320
321 if(edge){
322 flag_strong0 = filter_p1 && (FFABS(sum_p1p2) < beta2);
323 flag_strong1 = filter_q1 && (FFABS(sum_q1q2) < beta2);
324 }else{
325 flag_strong0 = flag_strong1 = 0;
326 }
327
328 lims = filter_p1 + filter_q1 + ((lim_q1 + lim_p1) >> 1) + 1;
329 if(flag_strong0 && flag_strong1){ /* strong filtering */
330 for(i = 0; i < 4; i++, src += stride){
331 int sflag, p0, q0, p1, q1;
332 int t = src[0*step] - src[-1*step];
333
334 if(!t) continue;
335 sflag = (alpha * FFABS(t)) >> 7;
336 if(sflag > 1) continue;
337
338 p0 = (25*src[-3*step] + 26*src[-2*step]
339 + 26*src[-1*step]
340 + 26*src[ 0*step] + 25*src[ 1*step] + rv40_dither_l[dmode + i]) >> 7;
341 q0 = (25*src[-2*step] + 26*src[-1*step]
342 + 26*src[ 0*step]
343 + 26*src[ 1*step] + 25*src[ 2*step] + rv40_dither_r[dmode + i]) >> 7;
344 if(sflag){
345 p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);
346 q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);
347 }
348 p1 = (25*src[-4*step] + 26*src[-3*step]
349 + 26*src[-2*step]
350 + 26*p0 + 25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;
351 q1 = (25*src[-1*step] + 26*q0
352 + 26*src[ 1*step]
353 + 26*src[ 2*step] + 25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;
354 if(sflag){
355 p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);
356 q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);
357 }
358 src[-2*step] = p1;
359 src[-1*step] = p0;
360 src[ 0*step] = q0;
361 src[ 1*step] = q1;
362 if(!chroma){
363 src[-3*step] = (25*src[-1*step] + 26*src[-2*step] + 51*src[-3*step] + 26*src[-4*step] + 64) >> 7;
364 src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] + 51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;
365 }
366 }
367 }else if(filter_p1 && filter_q1){
368 for(i = 0; i < 4; i++, src += stride)
369 rv40_weak_loop_filter(src, step, 1, 1, alpha, beta, lims, lim_q1, lim_p1,
370 diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);
371 }else{
372 for(i = 0; i < 4; i++, src += stride)
373 rv40_weak_loop_filter(src, step, filter_p1, filter_q1,
374 alpha, beta, lims>>1, lim_q1>>1, lim_p1>>1,
375 diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);
376 }
377}
378
379static void rv40_v_loop_filter(uint8_t *src, int stride, int dmode,
380 int lim_q1, int lim_p1,
381 int alpha, int beta, int beta2, int chroma, int edge){
382 rv40_adaptive_loop_filter(src, 1, stride, dmode, lim_q1, lim_p1,
383 alpha, beta, beta2, chroma, edge);
384}
385static void rv40_h_loop_filter(uint8_t *src, int stride, int dmode,
386 int lim_q1, int lim_p1,
387 int alpha, int beta, int beta2, int chroma, int edge){
388 rv40_adaptive_loop_filter(src, stride, 1, dmode, lim_q1, lim_p1,
389 alpha, beta, beta2, chroma, edge);
390}
391
ae4ca617
KS
392/**
393 * Initialize decoder.
394 */
98a6fff9 395static av_cold int rv40_decode_init(AVCodecContext *avctx)
ae4ca617
KS
396{
397 RV34DecContext *r = avctx->priv_data;
398
399 r->rv30 = 0;
400 ff_rv34_decode_init(avctx);
401 if(!aic_top_vlc.bits)
402 rv40_init_tables();
403 r->parse_slice_header = rv40_parse_slice_header;
404 r->decode_intra_types = rv40_decode_intra_types;
405 r->decode_mb_info = rv40_decode_mb_info;
406 r->luma_dc_quant_i = rv40_luma_dc_quant[0];
407 r->luma_dc_quant_p = rv40_luma_dc_quant[1];
408 return 0;
409}
410
411AVCodec rv40_decoder = {
412 "rv40",
413 CODEC_TYPE_VIDEO,
414 CODEC_ID_RV40,
415 sizeof(RV34DecContext),
416 rv40_decode_init,
417 NULL,
418 ff_rv34_decode_end,
419 ff_rv34_decode_frame,
52b91753 420 CODEC_CAP_DR1 | CODEC_CAP_DELAY,
fe4bf374 421 .long_name = NULL_IF_CONFIG_SMALL("RealVideo 4.0"),
ae4ca617 422};