Imported Debian version 2.5.3~trusty1
[deb_ffmpeg.git] / ffmpeg / libavcodec / h264_mb_template.c
1 /*
2 * H.26L/H.264/AVC/JVT/14496-10/... decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
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 #include "svq3.h"
23
24 #undef FUNC
25 #undef PIXEL_SHIFT
26
27 #if SIMPLE
28 # define FUNC(n) AV_JOIN(n ## _simple_, BITS)
29 # define PIXEL_SHIFT (BITS >> 4)
30 #else
31 # define FUNC(n) n ## _complex
32 # define PIXEL_SHIFT h->pixel_shift
33 #endif
34
35 #undef CHROMA_IDC
36 #define CHROMA_IDC 1
37 #include "h264_mc_template.c"
38
39 #undef CHROMA_IDC
40 #define CHROMA_IDC 2
41 #include "h264_mc_template.c"
42
43 static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
44 {
45 const int mb_x = h->mb_x;
46 const int mb_y = h->mb_y;
47 const int mb_xy = h->mb_xy;
48 const int mb_type = h->cur_pic.mb_type[mb_xy];
49 uint8_t *dest_y, *dest_cb, *dest_cr;
50 int linesize, uvlinesize /*dct_offset*/;
51 int i, j;
52 int *block_offset = &h->block_offset[0];
53 const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass);
54 /* is_h264 should always be true if SVQ3 is disabled. */
55 const int is_h264 = !CONFIG_SVQ3_DECODER || SIMPLE || h->avctx->codec_id == AV_CODEC_ID_H264;
56 void (*idct_add)(uint8_t *dst, int16_t *block, int stride);
57 const int block_h = 16 >> h->chroma_y_shift;
58 const int chroma422 = CHROMA422(h);
59
60 dest_y = h->cur_pic.f.data[0] + ((mb_x << PIXEL_SHIFT) + mb_y * h->linesize) * 16;
61 dest_cb = h->cur_pic.f.data[1] + (mb_x << PIXEL_SHIFT) * 8 + mb_y * h->uvlinesize * block_h;
62 dest_cr = h->cur_pic.f.data[2] + (mb_x << PIXEL_SHIFT) * 8 + mb_y * h->uvlinesize * block_h;
63
64 h->vdsp.prefetch(dest_y + (h->mb_x & 3) * 4 * h->linesize + (64 << PIXEL_SHIFT), h->linesize, 4);
65 h->vdsp.prefetch(dest_cb + (h->mb_x & 7) * h->uvlinesize + (64 << PIXEL_SHIFT), dest_cr - dest_cb, 2);
66
67 h->list_counts[mb_xy] = h->list_count;
68
69 if (!SIMPLE && MB_FIELD(h)) {
70 linesize = h->mb_linesize = h->linesize * 2;
71 uvlinesize = h->mb_uvlinesize = h->uvlinesize * 2;
72 block_offset = &h->block_offset[48];
73 if (mb_y & 1) { // FIXME move out of this function?
74 dest_y -= h->linesize * 15;
75 dest_cb -= h->uvlinesize * (block_h - 1);
76 dest_cr -= h->uvlinesize * (block_h - 1);
77 }
78 if (FRAME_MBAFF(h)) {
79 int list;
80 for (list = 0; list < h->list_count; list++) {
81 if (!USES_LIST(mb_type, list))
82 continue;
83 if (IS_16X16(mb_type)) {
84 int8_t *ref = &h->ref_cache[list][scan8[0]];
85 fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (h->mb_y & 1), 1);
86 } else {
87 for (i = 0; i < 16; i += 4) {
88 int ref = h->ref_cache[list][scan8[i]];
89 if (ref >= 0)
90 fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
91 8, (16 + ref) ^ (h->mb_y & 1), 1);
92 }
93 }
94 }
95 }
96 } else {
97 linesize = h->mb_linesize = h->linesize;
98 uvlinesize = h->mb_uvlinesize = h->uvlinesize;
99 // dct_offset = s->linesize * 16;
100 }
101
102 if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
103 const int bit_depth = h->sps.bit_depth_luma;
104 if (PIXEL_SHIFT) {
105 int j;
106 GetBitContext gb;
107 init_get_bits(&gb, h->intra_pcm_ptr,
108 ff_h264_mb_sizes[h->sps.chroma_format_idc] * bit_depth);
109
110 for (i = 0; i < 16; i++) {
111 uint16_t *tmp_y = (uint16_t *)(dest_y + i * linesize);
112 for (j = 0; j < 16; j++)
113 tmp_y[j] = get_bits(&gb, bit_depth);
114 }
115 if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
116 if (!h->sps.chroma_format_idc) {
117 for (i = 0; i < block_h; i++) {
118 uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
119 uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
120 for (j = 0; j < 8; j++) {
121 tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1);
122 }
123 }
124 } else {
125 for (i = 0; i < block_h; i++) {
126 uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
127 for (j = 0; j < 8; j++)
128 tmp_cb[j] = get_bits(&gb, bit_depth);
129 }
130 for (i = 0; i < block_h; i++) {
131 uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
132 for (j = 0; j < 8; j++)
133 tmp_cr[j] = get_bits(&gb, bit_depth);
134 }
135 }
136 }
137 } else {
138 for (i = 0; i < 16; i++)
139 memcpy(dest_y + i * linesize, h->intra_pcm_ptr + i * 16, 16);
140 if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
141 if (!h->sps.chroma_format_idc) {
142 for (i = 0; i < 8; i++) {
143 memset(dest_cb + i * uvlinesize, 1 << (bit_depth - 1), 8);
144 memset(dest_cr + i * uvlinesize, 1 << (bit_depth - 1), 8);
145 }
146 } else {
147 const uint8_t *src_cb = h->intra_pcm_ptr + 256;
148 const uint8_t *src_cr = h->intra_pcm_ptr + 256 + block_h * 8;
149 for (i = 0; i < block_h; i++) {
150 memcpy(dest_cb + i * uvlinesize, src_cb + i * 8, 8);
151 memcpy(dest_cr + i * uvlinesize, src_cr + i * 8, 8);
152 }
153 }
154 }
155 }
156 } else {
157 if (IS_INTRA(mb_type)) {
158 if (h->deblocking_filter)
159 xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize,
160 uvlinesize, 1, 0, SIMPLE, PIXEL_SHIFT);
161
162 if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
163 h->hpc.pred8x8[h->chroma_pred_mode](dest_cb, uvlinesize);
164 h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize);
165 }
166
167 hl_decode_mb_predict_luma(h, mb_type, is_h264, SIMPLE,
168 transform_bypass, PIXEL_SHIFT,
169 block_offset, linesize, dest_y, 0);
170
171 if (h->deblocking_filter)
172 xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize,
173 uvlinesize, 0, 0, SIMPLE, PIXEL_SHIFT);
174 } else if (is_h264) {
175 if (chroma422) {
176 FUNC(hl_motion_422)(h, dest_y, dest_cb, dest_cr,
177 h->qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
178 h->qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
179 h->h264dsp.weight_h264_pixels_tab,
180 h->h264dsp.biweight_h264_pixels_tab);
181 } else {
182 FUNC(hl_motion_420)(h, dest_y, dest_cb, dest_cr,
183 h->qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
184 h->qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
185 h->h264dsp.weight_h264_pixels_tab,
186 h->h264dsp.biweight_h264_pixels_tab);
187 }
188 }
189
190 hl_decode_mb_idct_luma(h, mb_type, is_h264, SIMPLE, transform_bypass,
191 PIXEL_SHIFT, block_offset, linesize, dest_y, 0);
192
193 if ((SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) &&
194 (h->cbp & 0x30)) {
195 uint8_t *dest[2] = { dest_cb, dest_cr };
196 if (transform_bypass) {
197 if (IS_INTRA(mb_type) && h->sps.profile_idc == 244 &&
198 (h->chroma_pred_mode == VERT_PRED8x8 ||
199 h->chroma_pred_mode == HOR_PRED8x8)) {
200 h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0],
201 block_offset + 16,
202 h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
203 uvlinesize);
204 h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1],
205 block_offset + 32,
206 h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
207 uvlinesize);
208 } else {
209 idct_add = h->h264dsp.h264_add_pixels4_clear;
210 for (j = 1; j < 3; j++) {
211 for (i = j * 16; i < j * 16 + 4; i++)
212 if (h->non_zero_count_cache[scan8[i]] ||
213 dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
214 idct_add(dest[j - 1] + block_offset[i],
215 h->mb + (i * 16 << PIXEL_SHIFT),
216 uvlinesize);
217 if (chroma422) {
218 for (i = j * 16 + 4; i < j * 16 + 8; i++)
219 if (h->non_zero_count_cache[scan8[i + 4]] ||
220 dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
221 idct_add(dest[j - 1] + block_offset[i + 4],
222 h->mb + (i * 16 << PIXEL_SHIFT),
223 uvlinesize);
224 }
225 }
226 }
227 } else {
228 if (is_h264) {
229 int qp[2];
230 if (chroma422) {
231 qp[0] = h->chroma_qp[0] + 3;
232 qp[1] = h->chroma_qp[1] + 3;
233 } else {
234 qp[0] = h->chroma_qp[0];
235 qp[1] = h->chroma_qp[1];
236 }
237 if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]])
238 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
239 h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]);
240 if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]])
241 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
242 h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]);
243 h->h264dsp.h264_idct_add8(dest, block_offset,
244 h->mb, uvlinesize,
245 h->non_zero_count_cache);
246 } else if (CONFIG_SVQ3_DECODER) {
247 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 1,
248 h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][h->chroma_qp[0]][0]);
249 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 2,
250 h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][h->chroma_qp[1]][0]);
251 for (j = 1; j < 3; j++) {
252 for (i = j * 16; i < j * 16 + 4; i++)
253 if (h->non_zero_count_cache[scan8[i]] || h->mb[i * 16]) {
254 uint8_t *const ptr = dest[j - 1] + block_offset[i];
255 ff_svq3_add_idct_c(ptr, h->mb + i * 16,
256 uvlinesize,
257 ff_h264_chroma_qp[0][h->qscale + 12] - 12, 2);
258 }
259 }
260 }
261 }
262 }
263 }
264 }
265
266 #if !SIMPLE || BITS == 8
267
268 #undef CHROMA_IDC
269 #define CHROMA_IDC 3
270 #include "h264_mc_template.c"
271
272 static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
273 {
274 const int mb_x = h->mb_x;
275 const int mb_y = h->mb_y;
276 const int mb_xy = h->mb_xy;
277 const int mb_type = h->cur_pic.mb_type[mb_xy];
278 uint8_t *dest[3];
279 int linesize;
280 int i, j, p;
281 int *block_offset = &h->block_offset[0];
282 const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass);
283 const int plane_count = (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) ? 3 : 1;
284
285 for (p = 0; p < plane_count; p++) {
286 dest[p] = h->cur_pic.f.data[p] +
287 ((mb_x << PIXEL_SHIFT) + mb_y * h->linesize) * 16;
288 h->vdsp.prefetch(dest[p] + (h->mb_x & 3) * 4 * h->linesize + (64 << PIXEL_SHIFT),
289 h->linesize, 4);
290 }
291
292 h->list_counts[mb_xy] = h->list_count;
293
294 if (!SIMPLE && MB_FIELD(h)) {
295 linesize = h->mb_linesize = h->mb_uvlinesize = h->linesize * 2;
296 block_offset = &h->block_offset[48];
297 if (mb_y & 1) // FIXME move out of this function?
298 for (p = 0; p < 3; p++)
299 dest[p] -= h->linesize * 15;
300 if (FRAME_MBAFF(h)) {
301 int list;
302 for (list = 0; list < h->list_count; list++) {
303 if (!USES_LIST(mb_type, list))
304 continue;
305 if (IS_16X16(mb_type)) {
306 int8_t *ref = &h->ref_cache[list][scan8[0]];
307 fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (h->mb_y & 1), 1);
308 } else {
309 for (i = 0; i < 16; i += 4) {
310 int ref = h->ref_cache[list][scan8[i]];
311 if (ref >= 0)
312 fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
313 8, (16 + ref) ^ (h->mb_y & 1), 1);
314 }
315 }
316 }
317 }
318 } else {
319 linesize = h->mb_linesize = h->mb_uvlinesize = h->linesize;
320 }
321
322 if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
323 if (PIXEL_SHIFT) {
324 const int bit_depth = h->sps.bit_depth_luma;
325 GetBitContext gb;
326 init_get_bits(&gb, h->intra_pcm_ptr, 768 * bit_depth);
327
328 for (p = 0; p < plane_count; p++)
329 for (i = 0; i < 16; i++) {
330 uint16_t *tmp = (uint16_t *)(dest[p] + i * linesize);
331 for (j = 0; j < 16; j++)
332 tmp[j] = get_bits(&gb, bit_depth);
333 }
334 } else {
335 for (p = 0; p < plane_count; p++)
336 for (i = 0; i < 16; i++)
337 memcpy(dest[p] + i * linesize,
338 h->intra_pcm_ptr + p * 256 + i * 16, 16);
339 }
340 } else {
341 if (IS_INTRA(mb_type)) {
342 if (h->deblocking_filter)
343 xchg_mb_border(h, dest[0], dest[1], dest[2], linesize,
344 linesize, 1, 1, SIMPLE, PIXEL_SHIFT);
345
346 for (p = 0; p < plane_count; p++)
347 hl_decode_mb_predict_luma(h, mb_type, 1, SIMPLE,
348 transform_bypass, PIXEL_SHIFT,
349 block_offset, linesize, dest[p], p);
350
351 if (h->deblocking_filter)
352 xchg_mb_border(h, dest[0], dest[1], dest[2], linesize,
353 linesize, 0, 1, SIMPLE, PIXEL_SHIFT);
354 } else {
355 FUNC(hl_motion_444)(h, dest[0], dest[1], dest[2],
356 h->qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
357 h->qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
358 h->h264dsp.weight_h264_pixels_tab,
359 h->h264dsp.biweight_h264_pixels_tab);
360 }
361
362 for (p = 0; p < plane_count; p++)
363 hl_decode_mb_idct_luma(h, mb_type, 1, SIMPLE, transform_bypass,
364 PIXEL_SHIFT, block_offset, linesize,
365 dest[p], p);
366 }
367 }
368
369 #endif