blob: 6ae87cca130fb9b39a51f1bd761ad3097c1bdffd
1 | /* |
2 | * HEVC video decoder |
3 | * |
4 | * Copyright (C) 2012 - 2013 Guillaume Martres |
5 | * |
6 | * This file is part of FFmpeg. |
7 | * |
8 | * FFmpeg is free software; you can redistribute it and/or |
9 | * modify it under the terms of the GNU Lesser General Public |
10 | * License as published by the Free Software Foundation; either |
11 | * version 2.1 of the License, or (at your option) any later version. |
12 | * |
13 | * FFmpeg is distributed in the hope that it will be useful, |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
16 | * Lesser General Public License for more details. |
17 | * |
18 | * You should have received a copy of the GNU Lesser General Public |
19 | * License along with FFmpeg; if not, write to the Free Software |
20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
21 | */ |
22 | |
23 | #include "libavutil/pixdesc.h" |
24 | |
25 | #include "bit_depth_template.c" |
26 | #include "hevcpred.h" |
27 | |
28 | #define POS(x, y) src[(x) + stride * (y)] |
29 | |
30 | static av_always_inline void FUNC(intra_pred)(HEVCContext *s, int x0, int y0, |
31 | int log2_size, int c_idx) |
32 | { |
33 | #define PU(x) \ |
34 | ((x) >> s->ps.sps->log2_min_pu_size) |
35 | #define MVF(x, y) \ |
36 | (s->ref->tab_mvf[(x) + (y) * min_pu_width]) |
37 | #define MVF_PU(x, y) \ |
38 | MVF(PU(x0 + ((x) << hshift)), PU(y0 + ((y) << vshift))) |
39 | #define IS_INTRA(x, y) \ |
40 | (MVF_PU(x, y).pred_flag == PF_INTRA) |
41 | #define MIN_TB_ADDR_ZS(x, y) \ |
42 | s->ps.pps->min_tb_addr_zs[(y) * (s->ps.sps->tb_mask+2) + (x)] |
43 | #define EXTEND(ptr, val, len) \ |
44 | do { \ |
45 | pixel4 pix = PIXEL_SPLAT_X4(val); \ |
46 | for (i = 0; i < (len); i += 4) \ |
47 | AV_WN4P(ptr + i, pix); \ |
48 | } while (0) |
49 | |
50 | #define EXTEND_RIGHT_CIP(ptr, start, length) \ |
51 | for (i = start; i < (start) + (length); i += 4) \ |
52 | if (!IS_INTRA(i, -1)) \ |
53 | AV_WN4P(&ptr[i], a); \ |
54 | else \ |
55 | a = PIXEL_SPLAT_X4(ptr[i+3]) |
56 | #define EXTEND_LEFT_CIP(ptr, start, length) \ |
57 | for (i = start; i > (start) - (length); i--) \ |
58 | if (!IS_INTRA(i - 1, -1)) \ |
59 | ptr[i - 1] = ptr[i] |
60 | #define EXTEND_UP_CIP(ptr, start, length) \ |
61 | for (i = (start); i > (start) - (length); i -= 4) \ |
62 | if (!IS_INTRA(-1, i - 3)) \ |
63 | AV_WN4P(&ptr[i - 3], a); \ |
64 | else \ |
65 | a = PIXEL_SPLAT_X4(ptr[i - 3]) |
66 | #define EXTEND_DOWN_CIP(ptr, start, length) \ |
67 | for (i = start; i < (start) + (length); i += 4) \ |
68 | if (!IS_INTRA(-1, i)) \ |
69 | AV_WN4P(&ptr[i], a); \ |
70 | else \ |
71 | a = PIXEL_SPLAT_X4(ptr[i + 3]) |
72 | |
73 | HEVCLocalContext *lc = s->HEVClc; |
74 | int i; |
75 | int hshift = s->ps.sps->hshift[c_idx]; |
76 | int vshift = s->ps.sps->vshift[c_idx]; |
77 | int size = (1 << log2_size); |
78 | int size_in_luma_h = size << hshift; |
79 | int size_in_tbs_h = size_in_luma_h >> s->ps.sps->log2_min_tb_size; |
80 | int size_in_luma_v = size << vshift; |
81 | int size_in_tbs_v = size_in_luma_v >> s->ps.sps->log2_min_tb_size; |
82 | int x = x0 >> hshift; |
83 | int y = y0 >> vshift; |
84 | int x_tb = (x0 >> s->ps.sps->log2_min_tb_size) & s->ps.sps->tb_mask; |
85 | int y_tb = (y0 >> s->ps.sps->log2_min_tb_size) & s->ps.sps->tb_mask; |
86 | |
87 | int cur_tb_addr = MIN_TB_ADDR_ZS(x_tb, y_tb); |
88 | |
89 | ptrdiff_t stride = s->frame->linesize[c_idx] / sizeof(pixel); |
90 | pixel *src = (pixel*)s->frame->data[c_idx] + x + y * stride; |
91 | |
92 | int min_pu_width = s->ps.sps->min_pu_width; |
93 | |
94 | enum IntraPredMode mode = c_idx ? lc->tu.intra_pred_mode_c : |
95 | lc->tu.intra_pred_mode; |
96 | pixel4 a; |
97 | pixel left_array[2 * MAX_TB_SIZE + 1]; |
98 | pixel filtered_left_array[2 * MAX_TB_SIZE + 1]; |
99 | pixel top_array[2 * MAX_TB_SIZE + 1]; |
100 | pixel filtered_top_array[2 * MAX_TB_SIZE + 1]; |
101 | |
102 | pixel *left = left_array + 1; |
103 | pixel *top = top_array + 1; |
104 | pixel *filtered_left = filtered_left_array + 1; |
105 | pixel *filtered_top = filtered_top_array + 1; |
106 | int cand_bottom_left = lc->na.cand_bottom_left && cur_tb_addr > MIN_TB_ADDR_ZS( x_tb - 1, (y_tb + size_in_tbs_v) & s->ps.sps->tb_mask); |
107 | int cand_left = lc->na.cand_left; |
108 | int cand_up_left = lc->na.cand_up_left; |
109 | int cand_up = lc->na.cand_up; |
110 | int cand_up_right = lc->na.cand_up_right && cur_tb_addr > MIN_TB_ADDR_ZS((x_tb + size_in_tbs_h) & s->ps.sps->tb_mask, y_tb - 1); |
111 | |
112 | int bottom_left_size = (FFMIN(y0 + 2 * size_in_luma_v, s->ps.sps->height) - |
113 | (y0 + size_in_luma_v)) >> vshift; |
114 | int top_right_size = (FFMIN(x0 + 2 * size_in_luma_h, s->ps.sps->width) - |
115 | (x0 + size_in_luma_h)) >> hshift; |
116 | |
117 | if (s->ps.pps->constrained_intra_pred_flag == 1) { |
118 | int size_in_luma_pu_v = PU(size_in_luma_v); |
119 | int size_in_luma_pu_h = PU(size_in_luma_h); |
120 | int on_pu_edge_x = !av_mod_uintp2(x0, s->ps.sps->log2_min_pu_size); |
121 | int on_pu_edge_y = !av_mod_uintp2(y0, s->ps.sps->log2_min_pu_size); |
122 | if (!size_in_luma_pu_h) |
123 | size_in_luma_pu_h++; |
124 | if (cand_bottom_left == 1 && on_pu_edge_x) { |
125 | int x_left_pu = PU(x0 - 1); |
126 | int y_bottom_pu = PU(y0 + size_in_luma_v); |
127 | int max = FFMIN(size_in_luma_pu_v, s->ps.sps->min_pu_height - y_bottom_pu); |
128 | cand_bottom_left = 0; |
129 | for (i = 0; i < max; i += 2) |
130 | cand_bottom_left |= (MVF(x_left_pu, y_bottom_pu + i).pred_flag == PF_INTRA); |
131 | } |
132 | if (cand_left == 1 && on_pu_edge_x) { |
133 | int x_left_pu = PU(x0 - 1); |
134 | int y_left_pu = PU(y0); |
135 | int max = FFMIN(size_in_luma_pu_v, s->ps.sps->min_pu_height - y_left_pu); |
136 | cand_left = 0; |
137 | for (i = 0; i < max; i += 2) |
138 | cand_left |= (MVF(x_left_pu, y_left_pu + i).pred_flag == PF_INTRA); |
139 | } |
140 | if (cand_up_left == 1) { |
141 | int x_left_pu = PU(x0 - 1); |
142 | int y_top_pu = PU(y0 - 1); |
143 | cand_up_left = MVF(x_left_pu, y_top_pu).pred_flag == PF_INTRA; |
144 | } |
145 | if (cand_up == 1 && on_pu_edge_y) { |
146 | int x_top_pu = PU(x0); |
147 | int y_top_pu = PU(y0 - 1); |
148 | int max = FFMIN(size_in_luma_pu_h, s->ps.sps->min_pu_width - x_top_pu); |
149 | cand_up = 0; |
150 | for (i = 0; i < max; i += 2) |
151 | cand_up |= (MVF(x_top_pu + i, y_top_pu).pred_flag == PF_INTRA); |
152 | } |
153 | if (cand_up_right == 1 && on_pu_edge_y) { |
154 | int y_top_pu = PU(y0 - 1); |
155 | int x_right_pu = PU(x0 + size_in_luma_h); |
156 | int max = FFMIN(size_in_luma_pu_h, s->ps.sps->min_pu_width - x_right_pu); |
157 | cand_up_right = 0; |
158 | for (i = 0; i < max; i += 2) |
159 | cand_up_right |= (MVF(x_right_pu + i, y_top_pu).pred_flag == PF_INTRA); |
160 | } |
161 | memset(left, 128, 2 * MAX_TB_SIZE*sizeof(pixel)); |
162 | memset(top , 128, 2 * MAX_TB_SIZE*sizeof(pixel)); |
163 | top[-1] = 128; |
164 | } |
165 | if (cand_up_left) { |
166 | left[-1] = POS(-1, -1); |
167 | top[-1] = left[-1]; |
168 | } |
169 | if (cand_up) |
170 | memcpy(top, src - stride, size * sizeof(pixel)); |
171 | if (cand_up_right) { |
172 | memcpy(top + size, src - stride + size, size * sizeof(pixel)); |
173 | EXTEND(top + size + top_right_size, POS(size + top_right_size - 1, -1), |
174 | size - top_right_size); |
175 | } |
176 | if (cand_left) |
177 | for (i = 0; i < size; i++) |
178 | left[i] = POS(-1, i); |
179 | if (cand_bottom_left) { |
180 | for (i = size; i < size + bottom_left_size; i++) |
181 | left[i] = POS(-1, i); |
182 | EXTEND(left + size + bottom_left_size, POS(-1, size + bottom_left_size - 1), |
183 | size - bottom_left_size); |
184 | } |
185 | |
186 | if (s->ps.pps->constrained_intra_pred_flag == 1) { |
187 | if (cand_bottom_left || cand_left || cand_up_left || cand_up || cand_up_right) { |
188 | int size_max_x = x0 + ((2 * size) << hshift) < s->ps.sps->width ? |
189 | 2 * size : (s->ps.sps->width - x0) >> hshift; |
190 | int size_max_y = y0 + ((2 * size) << vshift) < s->ps.sps->height ? |
191 | 2 * size : (s->ps.sps->height - y0) >> vshift; |
192 | int j = size + (cand_bottom_left? bottom_left_size: 0) -1; |
193 | if (!cand_up_right) { |
194 | size_max_x = x0 + ((size) << hshift) < s->ps.sps->width ? |
195 | size : (s->ps.sps->width - x0) >> hshift; |
196 | } |
197 | if (!cand_bottom_left) { |
198 | size_max_y = y0 + (( size) << vshift) < s->ps.sps->height ? |
199 | size : (s->ps.sps->height - y0) >> vshift; |
200 | } |
201 | if (cand_bottom_left || cand_left || cand_up_left) { |
202 | while (j > -1 && !IS_INTRA(-1, j)) |
203 | j--; |
204 | if (!IS_INTRA(-1, j)) { |
205 | j = 0; |
206 | while (j < size_max_x && !IS_INTRA(j, -1)) |
207 | j++; |
208 | EXTEND_LEFT_CIP(top, j, j + 1); |
209 | left[-1] = top[-1]; |
210 | } |
211 | } else { |
212 | j = 0; |
213 | while (j < size_max_x && !IS_INTRA(j, -1)) |
214 | j++; |
215 | if (j > 0) |
216 | if (x0 > 0) { |
217 | EXTEND_LEFT_CIP(top, j, j + 1); |
218 | } else { |
219 | EXTEND_LEFT_CIP(top, j, j); |
220 | top[-1] = top[0]; |
221 | } |
222 | left[-1] = top[-1]; |
223 | } |
224 | left[-1] = top[-1]; |
225 | if (cand_bottom_left || cand_left) { |
226 | a = PIXEL_SPLAT_X4(left[-1]); |
227 | EXTEND_DOWN_CIP(left, 0, size_max_y); |
228 | } |
229 | if (!cand_left) |
230 | EXTEND(left, left[-1], size); |
231 | if (!cand_bottom_left) |
232 | EXTEND(left + size, left[size - 1], size); |
233 | if (x0 != 0 && y0 != 0) { |
234 | a = PIXEL_SPLAT_X4(left[size_max_y - 1]); |
235 | EXTEND_UP_CIP(left, size_max_y - 1, size_max_y); |
236 | if (!IS_INTRA(-1, - 1)) |
237 | left[-1] = left[0]; |
238 | } else if (x0 == 0) { |
239 | EXTEND(left, 0, size_max_y); |
240 | } else { |
241 | a = PIXEL_SPLAT_X4(left[size_max_y - 1]); |
242 | EXTEND_UP_CIP(left, size_max_y - 1, size_max_y); |
243 | } |
244 | top[-1] = left[-1]; |
245 | if (y0 != 0) { |
246 | a = PIXEL_SPLAT_X4(left[-1]); |
247 | EXTEND_RIGHT_CIP(top, 0, size_max_x); |
248 | } |
249 | } |
250 | } |
251 | // Infer the unavailable samples |
252 | if (!cand_bottom_left) { |
253 | if (cand_left) { |
254 | EXTEND(left + size, left[size - 1], size); |
255 | } else if (cand_up_left) { |
256 | EXTEND(left, left[-1], 2 * size); |
257 | cand_left = 1; |
258 | } else if (cand_up) { |
259 | left[-1] = top[0]; |
260 | EXTEND(left, left[-1], 2 * size); |
261 | cand_up_left = 1; |
262 | cand_left = 1; |
263 | } else if (cand_up_right) { |
264 | EXTEND(top, top[size], size); |
265 | left[-1] = top[size]; |
266 | EXTEND(left, left[-1], 2 * size); |
267 | cand_up = 1; |
268 | cand_up_left = 1; |
269 | cand_left = 1; |
270 | } else { // No samples available |
271 | left[-1] = (1 << (BIT_DEPTH - 1)); |
272 | EXTEND(top, left[-1], 2 * size); |
273 | EXTEND(left, left[-1], 2 * size); |
274 | } |
275 | } |
276 | |
277 | if (!cand_left) |
278 | EXTEND(left, left[size], size); |
279 | if (!cand_up_left) { |
280 | left[-1] = left[0]; |
281 | } |
282 | if (!cand_up) |
283 | EXTEND(top, left[-1], size); |
284 | if (!cand_up_right) |
285 | EXTEND(top + size, top[size - 1], size); |
286 | |
287 | top[-1] = left[-1]; |
288 | |
289 | // Filtering process |
290 | if (!s->ps.sps->intra_smoothing_disabled_flag && (c_idx == 0 || s->ps.sps->chroma_format_idc == 3)) { |
291 | if (mode != INTRA_DC && size != 4){ |
292 | int intra_hor_ver_dist_thresh[] = { 7, 1, 0 }; |
293 | int min_dist_vert_hor = FFMIN(FFABS((int)(mode - 26U)), |
294 | FFABS((int)(mode - 10U))); |
295 | if (min_dist_vert_hor > intra_hor_ver_dist_thresh[log2_size - 3]) { |
296 | int threshold = 1 << (BIT_DEPTH - 5); |
297 | if (s->ps.sps->sps_strong_intra_smoothing_enable_flag && c_idx == 0 && |
298 | log2_size == 5 && |
299 | FFABS(top[-1] + top[63] - 2 * top[31]) < threshold && |
300 | FFABS(left[-1] + left[63] - 2 * left[31]) < threshold) { |
301 | // We can't just overwrite values in top because it could be |
302 | // a pointer into src |
303 | filtered_top[-1] = top[-1]; |
304 | filtered_top[63] = top[63]; |
305 | for (i = 0; i < 63; i++) |
306 | filtered_top[i] = ((64 - (i + 1)) * top[-1] + |
307 | (i + 1) * top[63] + 32) >> 6; |
308 | for (i = 0; i < 63; i++) |
309 | left[i] = ((64 - (i + 1)) * left[-1] + |
310 | (i + 1) * left[63] + 32) >> 6; |
311 | top = filtered_top; |
312 | } else { |
313 | filtered_left[2 * size - 1] = left[2 * size - 1]; |
314 | filtered_top[2 * size - 1] = top[2 * size - 1]; |
315 | for (i = 2 * size - 2; i >= 0; i--) |
316 | filtered_left[i] = (left[i + 1] + 2 * left[i] + |
317 | left[i - 1] + 2) >> 2; |
318 | filtered_top[-1] = |
319 | filtered_left[-1] = (left[0] + 2 * left[-1] + top[0] + 2) >> 2; |
320 | for (i = 2 * size - 2; i >= 0; i--) |
321 | filtered_top[i] = (top[i + 1] + 2 * top[i] + |
322 | top[i - 1] + 2) >> 2; |
323 | left = filtered_left; |
324 | top = filtered_top; |
325 | } |
326 | } |
327 | } |
328 | } |
329 | |
330 | switch (mode) { |
331 | case INTRA_PLANAR: |
332 | s->hpc.pred_planar[log2_size - 2]((uint8_t *)src, (uint8_t *)top, |
333 | (uint8_t *)left, stride); |
334 | break; |
335 | case INTRA_DC: |
336 | s->hpc.pred_dc((uint8_t *)src, (uint8_t *)top, |
337 | (uint8_t *)left, stride, log2_size, c_idx); |
338 | break; |
339 | default: |
340 | s->hpc.pred_angular[log2_size - 2]((uint8_t *)src, (uint8_t *)top, |
341 | (uint8_t *)left, stride, c_idx, |
342 | mode); |
343 | break; |
344 | } |
345 | } |
346 | |
347 | #define INTRA_PRED(size) \ |
348 | static void FUNC(intra_pred_ ## size)(HEVCContext *s, int x0, int y0, int c_idx) \ |
349 | { \ |
350 | FUNC(intra_pred)(s, x0, y0, size, c_idx); \ |
351 | } |
352 | |
353 | INTRA_PRED(2) |
354 | INTRA_PRED(3) |
355 | INTRA_PRED(4) |
356 | INTRA_PRED(5) |
357 | |
358 | #undef INTRA_PRED |
359 | |
360 | static av_always_inline void FUNC(pred_planar)(uint8_t *_src, const uint8_t *_top, |
361 | const uint8_t *_left, ptrdiff_t stride, |
362 | int trafo_size) |
363 | { |
364 | int x, y; |
365 | pixel *src = (pixel *)_src; |
366 | const pixel *top = (const pixel *)_top; |
367 | const pixel *left = (const pixel *)_left; |
368 | int size = 1 << trafo_size; |
369 | for (y = 0; y < size; y++) |
370 | for (x = 0; x < size; x++) |
371 | POS(x, y) = ((size - 1 - x) * left[y] + (x + 1) * top[size] + |
372 | (size - 1 - y) * top[x] + (y + 1) * left[size] + size) >> (trafo_size + 1); |
373 | } |
374 | |
375 | #define PRED_PLANAR(size)\ |
376 | static void FUNC(pred_planar_ ## size)(uint8_t *src, const uint8_t *top, \ |
377 | const uint8_t *left, ptrdiff_t stride) \ |
378 | { \ |
379 | FUNC(pred_planar)(src, top, left, stride, size + 2); \ |
380 | } |
381 | |
382 | PRED_PLANAR(0) |
383 | PRED_PLANAR(1) |
384 | PRED_PLANAR(2) |
385 | PRED_PLANAR(3) |
386 | |
387 | #undef PRED_PLANAR |
388 | |
389 | static void FUNC(pred_dc)(uint8_t *_src, const uint8_t *_top, |
390 | const uint8_t *_left, |
391 | ptrdiff_t stride, int log2_size, int c_idx) |
392 | { |
393 | int i, j, x, y; |
394 | int size = (1 << log2_size); |
395 | pixel *src = (pixel *)_src; |
396 | const pixel *top = (const pixel *)_top; |
397 | const pixel *left = (const pixel *)_left; |
398 | int dc = size; |
399 | pixel4 a; |
400 | for (i = 0; i < size; i++) |
401 | dc += left[i] + top[i]; |
402 | |
403 | dc >>= log2_size + 1; |
404 | |
405 | a = PIXEL_SPLAT_X4(dc); |
406 | |
407 | for (i = 0; i < size; i++) |
408 | for (j = 0; j < size; j+=4) |
409 | AV_WN4P(&POS(j, i), a); |
410 | |
411 | if (c_idx == 0 && size < 32) { |
412 | POS(0, 0) = (left[0] + 2 * dc + top[0] + 2) >> 2; |
413 | for (x = 1; x < size; x++) |
414 | POS(x, 0) = (top[x] + 3 * dc + 2) >> 2; |
415 | for (y = 1; y < size; y++) |
416 | POS(0, y) = (left[y] + 3 * dc + 2) >> 2; |
417 | } |
418 | } |
419 | |
420 | static av_always_inline void FUNC(pred_angular)(uint8_t *_src, |
421 | const uint8_t *_top, |
422 | const uint8_t *_left, |
423 | ptrdiff_t stride, int c_idx, |
424 | int mode, int size) |
425 | { |
426 | int x, y; |
427 | pixel *src = (pixel *)_src; |
428 | const pixel *top = (const pixel *)_top; |
429 | const pixel *left = (const pixel *)_left; |
430 | |
431 | static const int intra_pred_angle[] = { |
432 | 32, 26, 21, 17, 13, 9, 5, 2, 0, -2, -5, -9, -13, -17, -21, -26, -32, |
433 | -26, -21, -17, -13, -9, -5, -2, 0, 2, 5, 9, 13, 17, 21, 26, 32 |
434 | }; |
435 | static const int inv_angle[] = { |
436 | -4096, -1638, -910, -630, -482, -390, -315, -256, -315, -390, -482, |
437 | -630, -910, -1638, -4096 |
438 | }; |
439 | |
440 | int angle = intra_pred_angle[mode - 2]; |
441 | pixel ref_array[3 * MAX_TB_SIZE + 4]; |
442 | pixel *ref_tmp = ref_array + size; |
443 | const pixel *ref; |
444 | int last = (size * angle) >> 5; |
445 | |
446 | if (mode >= 18) { |
447 | ref = top - 1; |
448 | if (angle < 0 && last < -1) { |
449 | for (x = 0; x <= size; x += 4) |
450 | AV_WN4P(&ref_tmp[x], AV_RN4P(&top[x - 1])); |
451 | for (x = last; x <= -1; x++) |
452 | ref_tmp[x] = left[-1 + ((x * inv_angle[mode - 11] + 128) >> 8)]; |
453 | ref = ref_tmp; |
454 | } |
455 | |
456 | for (y = 0; y < size; y++) { |
457 | int idx = ((y + 1) * angle) >> 5; |
458 | int fact = ((y + 1) * angle) & 31; |
459 | if (fact) { |
460 | for (x = 0; x < size; x += 4) { |
461 | POS(x , y) = ((32 - fact) * ref[x + idx + 1] + |
462 | fact * ref[x + idx + 2] + 16) >> 5; |
463 | POS(x + 1, y) = ((32 - fact) * ref[x + 1 + idx + 1] + |
464 | fact * ref[x + 1 + idx + 2] + 16) >> 5; |
465 | POS(x + 2, y) = ((32 - fact) * ref[x + 2 + idx + 1] + |
466 | fact * ref[x + 2 + idx + 2] + 16) >> 5; |
467 | POS(x + 3, y) = ((32 - fact) * ref[x + 3 + idx + 1] + |
468 | fact * ref[x + 3 + idx + 2] + 16) >> 5; |
469 | } |
470 | } else { |
471 | for (x = 0; x < size; x += 4) |
472 | AV_WN4P(&POS(x, y), AV_RN4P(&ref[x + idx + 1])); |
473 | } |
474 | } |
475 | if (mode == 26 && c_idx == 0 && size < 32) { |
476 | for (y = 0; y < size; y++) |
477 | POS(0, y) = av_clip_pixel(top[0] + ((left[y] - left[-1]) >> 1)); |
478 | } |
479 | } else { |
480 | ref = left - 1; |
481 | if (angle < 0 && last < -1) { |
482 | for (x = 0; x <= size; x += 4) |
483 | AV_WN4P(&ref_tmp[x], AV_RN4P(&left[x - 1])); |
484 | for (x = last; x <= -1; x++) |
485 | ref_tmp[x] = top[-1 + ((x * inv_angle[mode - 11] + 128) >> 8)]; |
486 | ref = ref_tmp; |
487 | } |
488 | |
489 | for (x = 0; x < size; x++) { |
490 | int idx = ((x + 1) * angle) >> 5; |
491 | int fact = ((x + 1) * angle) & 31; |
492 | if (fact) { |
493 | for (y = 0; y < size; y++) { |
494 | POS(x, y) = ((32 - fact) * ref[y + idx + 1] + |
495 | fact * ref[y + idx + 2] + 16) >> 5; |
496 | } |
497 | } else { |
498 | for (y = 0; y < size; y++) |
499 | POS(x, y) = ref[y + idx + 1]; |
500 | } |
501 | } |
502 | if (mode == 10 && c_idx == 0 && size < 32) { |
503 | for (x = 0; x < size; x += 4) { |
504 | POS(x, 0) = av_clip_pixel(left[0] + ((top[x ] - top[-1]) >> 1)); |
505 | POS(x + 1, 0) = av_clip_pixel(left[0] + ((top[x + 1] - top[-1]) >> 1)); |
506 | POS(x + 2, 0) = av_clip_pixel(left[0] + ((top[x + 2] - top[-1]) >> 1)); |
507 | POS(x + 3, 0) = av_clip_pixel(left[0] + ((top[x + 3] - top[-1]) >> 1)); |
508 | } |
509 | } |
510 | } |
511 | } |
512 | |
513 | static void FUNC(pred_angular_0)(uint8_t *src, const uint8_t *top, |
514 | const uint8_t *left, |
515 | ptrdiff_t stride, int c_idx, int mode) |
516 | { |
517 | FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 2); |
518 | } |
519 | |
520 | static void FUNC(pred_angular_1)(uint8_t *src, const uint8_t *top, |
521 | const uint8_t *left, |
522 | ptrdiff_t stride, int c_idx, int mode) |
523 | { |
524 | FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 3); |
525 | } |
526 | |
527 | static void FUNC(pred_angular_2)(uint8_t *src, const uint8_t *top, |
528 | const uint8_t *left, |
529 | ptrdiff_t stride, int c_idx, int mode) |
530 | { |
531 | FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 4); |
532 | } |
533 | |
534 | static void FUNC(pred_angular_3)(uint8_t *src, const uint8_t *top, |
535 | const uint8_t *left, |
536 | ptrdiff_t stride, int c_idx, int mode) |
537 | { |
538 | FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 5); |
539 | } |
540 | |
541 | #undef EXTEND_LEFT_CIP |
542 | #undef EXTEND_RIGHT_CIP |
543 | #undef EXTEND_UP_CIP |
544 | #undef EXTEND_DOWN_CIP |
545 | #undef IS_INTRA |
546 | #undef MVF_PU |
547 | #undef MVF |
548 | #undef PU |
549 | #undef EXTEND |
550 | #undef MIN_TB_ADDR_ZS |
551 | #undef POS |
552 |