Commit | Line | Data |
---|---|---|
2ba45a60 DM |
1 | /* |
2 | * quarterpel DSP functions | |
3 | * Copyright (c) 2000, 2001 Fabrice Bellard | |
4 | * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at> | |
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 <stddef.h> | |
24 | #include <stdint.h> | |
25 | ||
26 | #include "config.h" | |
27 | #include "libavutil/attributes.h" | |
28 | #include "libavutil/cpu.h" | |
29 | #include "libavutil/x86/cpu.h" | |
30 | #include "libavcodec/pixels.h" | |
31 | #include "libavcodec/qpeldsp.h" | |
32 | #include "fpel.h" | |
33 | ||
34 | void ff_put_pixels8_l2_mmxext(uint8_t *dst, | |
35 | const uint8_t *src1, const uint8_t *src2, | |
36 | int dstStride, int src1Stride, int h); | |
37 | void ff_put_no_rnd_pixels8_l2_mmxext(uint8_t *dst, | |
38 | const uint8_t *src1, const uint8_t *src2, | |
39 | int dstStride, int src1Stride, int h); | |
40 | void ff_avg_pixels8_l2_mmxext(uint8_t *dst, | |
41 | const uint8_t *src1, const uint8_t *src2, | |
42 | int dstStride, int src1Stride, int h); | |
43 | void ff_put_pixels16_l2_mmxext(uint8_t *dst, | |
44 | const uint8_t *src1, const uint8_t *src2, | |
45 | int dstStride, int src1Stride, int h); | |
46 | void ff_avg_pixels16_l2_mmxext(uint8_t *dst, | |
47 | const uint8_t *src1, const uint8_t *src2, | |
48 | int dstStride, int src1Stride, int h); | |
49 | void ff_put_no_rnd_pixels16_l2_mmxext(uint8_t *dst, | |
50 | const uint8_t *src1, const uint8_t *src2, | |
51 | int dstStride, int src1Stride, int h); | |
52 | void ff_put_mpeg4_qpel16_h_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
53 | int dstStride, int srcStride, int h); | |
54 | void ff_avg_mpeg4_qpel16_h_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
55 | int dstStride, int srcStride, int h); | |
56 | void ff_put_no_rnd_mpeg4_qpel16_h_lowpass_mmxext(uint8_t *dst, | |
57 | const uint8_t *src, | |
58 | int dstStride, int srcStride, | |
59 | int h); | |
60 | void ff_put_mpeg4_qpel8_h_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
61 | int dstStride, int srcStride, int h); | |
62 | void ff_avg_mpeg4_qpel8_h_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
63 | int dstStride, int srcStride, int h); | |
64 | void ff_put_no_rnd_mpeg4_qpel8_h_lowpass_mmxext(uint8_t *dst, | |
65 | const uint8_t *src, | |
66 | int dstStride, int srcStride, | |
67 | int h); | |
68 | void ff_put_mpeg4_qpel16_v_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
69 | int dstStride, int srcStride); | |
70 | void ff_avg_mpeg4_qpel16_v_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
71 | int dstStride, int srcStride); | |
72 | void ff_put_no_rnd_mpeg4_qpel16_v_lowpass_mmxext(uint8_t *dst, | |
73 | const uint8_t *src, | |
74 | int dstStride, int srcStride); | |
75 | void ff_put_mpeg4_qpel8_v_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
76 | int dstStride, int srcStride); | |
77 | void ff_avg_mpeg4_qpel8_v_lowpass_mmxext(uint8_t *dst, const uint8_t *src, | |
78 | int dstStride, int srcStride); | |
79 | void ff_put_no_rnd_mpeg4_qpel8_v_lowpass_mmxext(uint8_t *dst, | |
80 | const uint8_t *src, | |
81 | int dstStride, int srcStride); | |
82 | #define ff_put_no_rnd_pixels16_mmxext ff_put_pixels16_mmx | |
83 | #define ff_put_no_rnd_pixels8_mmxext ff_put_pixels8_mmx | |
84 | ||
85 | #if HAVE_YASM | |
86 | ||
87 | #define ff_put_pixels16_mmxext ff_put_pixels16_mmx | |
88 | #define ff_put_pixels8_mmxext ff_put_pixels8_mmx | |
89 | ||
90 | #define QPEL_OP(OPNAME, RND, MMX) \ | |
91 | static void OPNAME ## qpel8_mc00_ ## MMX(uint8_t *dst, \ | |
92 | const uint8_t *src, \ | |
93 | ptrdiff_t stride) \ | |
94 | { \ | |
95 | ff_ ## OPNAME ## pixels8_ ## MMX(dst, src, stride, 8); \ | |
96 | } \ | |
97 | \ | |
98 | static void OPNAME ## qpel8_mc10_ ## MMX(uint8_t *dst, \ | |
99 | const uint8_t *src, \ | |
100 | ptrdiff_t stride) \ | |
101 | { \ | |
102 | uint64_t temp[8]; \ | |
103 | uint8_t *const half = (uint8_t *) temp; \ | |
104 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(half, src, 8, \ | |
105 | stride, 8); \ | |
106 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, src, half, \ | |
107 | stride, stride, 8); \ | |
108 | } \ | |
109 | \ | |
110 | static void OPNAME ## qpel8_mc20_ ## MMX(uint8_t *dst, \ | |
111 | const uint8_t *src, \ | |
112 | ptrdiff_t stride) \ | |
113 | { \ | |
114 | ff_ ## OPNAME ## mpeg4_qpel8_h_lowpass_ ## MMX(dst, src, stride, \ | |
115 | stride, 8); \ | |
116 | } \ | |
117 | \ | |
118 | static void OPNAME ## qpel8_mc30_ ## MMX(uint8_t *dst, \ | |
119 | const uint8_t *src, \ | |
120 | ptrdiff_t stride) \ | |
121 | { \ | |
122 | uint64_t temp[8]; \ | |
123 | uint8_t *const half = (uint8_t *) temp; \ | |
124 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(half, src, 8, \ | |
125 | stride, 8); \ | |
126 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, src + 1, half, stride, \ | |
127 | stride, 8); \ | |
128 | } \ | |
129 | \ | |
130 | static void OPNAME ## qpel8_mc01_ ## MMX(uint8_t *dst, \ | |
131 | const uint8_t *src, \ | |
132 | ptrdiff_t stride) \ | |
133 | { \ | |
134 | uint64_t temp[8]; \ | |
135 | uint8_t *const half = (uint8_t *) temp; \ | |
136 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, \ | |
137 | 8, stride); \ | |
138 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, src, half, \ | |
139 | stride, stride, 8); \ | |
140 | } \ | |
141 | \ | |
142 | static void OPNAME ## qpel8_mc02_ ## MMX(uint8_t *dst, \ | |
143 | const uint8_t *src, \ | |
144 | ptrdiff_t stride) \ | |
145 | { \ | |
146 | ff_ ## OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, src, \ | |
147 | stride, stride); \ | |
148 | } \ | |
149 | \ | |
150 | static void OPNAME ## qpel8_mc03_ ## MMX(uint8_t *dst, \ | |
151 | const uint8_t *src, \ | |
152 | ptrdiff_t stride) \ | |
153 | { \ | |
154 | uint64_t temp[8]; \ | |
155 | uint8_t *const half = (uint8_t *) temp; \ | |
156 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, \ | |
157 | 8, stride); \ | |
158 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, src + stride, half, stride,\ | |
159 | stride, 8); \ | |
160 | } \ | |
161 | \ | |
162 | static void OPNAME ## qpel8_mc11_ ## MMX(uint8_t *dst, \ | |
163 | const uint8_t *src, \ | |
164 | ptrdiff_t stride) \ | |
165 | { \ | |
166 | uint64_t half[8 + 9]; \ | |
167 | uint8_t *const halfH = (uint8_t *) half + 64; \ | |
168 | uint8_t *const halfHV = (uint8_t *) half; \ | |
169 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
170 | stride, 9); \ | |
171 | ff_put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, \ | |
172 | stride, 9); \ | |
173 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\ | |
174 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, \ | |
175 | stride, 8, 8); \ | |
176 | } \ | |
177 | \ | |
178 | static void OPNAME ## qpel8_mc31_ ## MMX(uint8_t *dst, \ | |
179 | const uint8_t *src, \ | |
180 | ptrdiff_t stride) \ | |
181 | { \ | |
182 | uint64_t half[8 + 9]; \ | |
183 | uint8_t *const halfH = (uint8_t *) half + 64; \ | |
184 | uint8_t *const halfHV = (uint8_t *) half; \ | |
185 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
186 | stride, 9); \ | |
187 | ff_put ## RND ## pixels8_l2_ ## MMX(halfH, src + 1, halfH, 8, \ | |
188 | stride, 9); \ | |
189 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\ | |
190 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, \ | |
191 | stride, 8, 8); \ | |
192 | } \ | |
193 | \ | |
194 | static void OPNAME ## qpel8_mc13_ ## MMX(uint8_t *dst, \ | |
195 | const uint8_t *src, \ | |
196 | ptrdiff_t stride) \ | |
197 | { \ | |
198 | uint64_t half[8 + 9]; \ | |
199 | uint8_t *const halfH = (uint8_t *) half + 64; \ | |
200 | uint8_t *const halfHV = (uint8_t *) half; \ | |
201 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
202 | stride, 9); \ | |
203 | ff_put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, 8, \ | |
204 | stride, 9); \ | |
205 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\ | |
206 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, halfH + 8, halfHV, \ | |
207 | stride, 8, 8); \ | |
208 | } \ | |
209 | \ | |
210 | static void OPNAME ## qpel8_mc33_ ## MMX(uint8_t *dst, \ | |
211 | const uint8_t *src, \ | |
212 | ptrdiff_t stride) \ | |
213 | { \ | |
214 | uint64_t half[8 + 9]; \ | |
215 | uint8_t *const halfH = (uint8_t *) half + 64; \ | |
216 | uint8_t *const halfHV = (uint8_t *) half; \ | |
217 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
218 | stride, 9); \ | |
219 | ff_put ## RND ## pixels8_l2_ ## MMX(halfH, src + 1, halfH, 8, \ | |
220 | stride, 9); \ | |
221 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\ | |
222 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, halfH + 8, halfHV, \ | |
223 | stride, 8, 8); \ | |
224 | } \ | |
225 | \ | |
226 | static void OPNAME ## qpel8_mc21_ ## MMX(uint8_t *dst, \ | |
227 | const uint8_t *src, \ | |
228 | ptrdiff_t stride) \ | |
229 | { \ | |
230 | uint64_t half[8 + 9]; \ | |
231 | uint8_t *const halfH = (uint8_t *) half + 64; \ | |
232 | uint8_t *const halfHV = (uint8_t *) half; \ | |
233 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
234 | stride, 9); \ | |
235 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\ | |
236 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, halfH, halfHV, \ | |
237 | stride, 8, 8); \ | |
238 | } \ | |
239 | \ | |
240 | static void OPNAME ## qpel8_mc23_ ## MMX(uint8_t *dst, \ | |
241 | const uint8_t *src, \ | |
242 | ptrdiff_t stride) \ | |
243 | { \ | |
244 | uint64_t half[8 + 9]; \ | |
245 | uint8_t *const halfH = (uint8_t *) half + 64; \ | |
246 | uint8_t *const halfHV = (uint8_t *) half; \ | |
247 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
248 | stride, 9); \ | |
249 | ff_put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\ | |
250 | ff_ ## OPNAME ## pixels8_l2_ ## MMX(dst, halfH + 8, halfHV, \ | |
251 | stride, 8, 8); \ | |
252 | } \ | |
253 | \ | |
254 | static void OPNAME ## qpel8_mc12_ ## MMX(uint8_t *dst, \ | |
255 | const uint8_t *src, \ | |
256 | ptrdiff_t stride) \ | |
257 | { \ | |
258 | uint64_t half[8 + 9]; \ | |
259 | uint8_t *const halfH = (uint8_t *) half; \ | |
260 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
261 | stride, 9); \ | |
262 | ff_put ## RND ## pixels8_l2_ ## MMX(halfH, src, halfH, \ | |
263 | 8, stride, 9); \ | |
264 | ff_ ## OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, \ | |
265 | stride, 8); \ | |
266 | } \ | |
267 | \ | |
268 | static void OPNAME ## qpel8_mc32_ ## MMX(uint8_t *dst, \ | |
269 | const uint8_t *src, \ | |
270 | ptrdiff_t stride) \ | |
271 | { \ | |
272 | uint64_t half[8 + 9]; \ | |
273 | uint8_t *const halfH = (uint8_t *) half; \ | |
274 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
275 | stride, 9); \ | |
276 | ff_put ## RND ## pixels8_l2_ ## MMX(halfH, src + 1, halfH, 8, \ | |
277 | stride, 9); \ | |
278 | ff_ ## OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, \ | |
279 | stride, 8); \ | |
280 | } \ | |
281 | \ | |
282 | static void OPNAME ## qpel8_mc22_ ## MMX(uint8_t *dst, \ | |
283 | const uint8_t *src, \ | |
284 | ptrdiff_t stride) \ | |
285 | { \ | |
286 | uint64_t half[9]; \ | |
287 | uint8_t *const halfH = (uint8_t *) half; \ | |
288 | ff_put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, \ | |
289 | stride, 9); \ | |
290 | ff_ ## OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, \ | |
291 | stride, 8); \ | |
292 | } \ | |
293 | \ | |
294 | static void OPNAME ## qpel16_mc00_ ## MMX(uint8_t *dst, \ | |
295 | const uint8_t *src, \ | |
296 | ptrdiff_t stride) \ | |
297 | { \ | |
298 | ff_ ## OPNAME ## pixels16_ ## MMX(dst, src, stride, 16); \ | |
299 | } \ | |
300 | \ | |
301 | static void OPNAME ## qpel16_mc10_ ## MMX(uint8_t *dst, \ | |
302 | const uint8_t *src, \ | |
303 | ptrdiff_t stride) \ | |
304 | { \ | |
305 | uint64_t temp[32]; \ | |
306 | uint8_t *const half = (uint8_t *) temp; \ | |
307 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(half, src, 16, \ | |
308 | stride, 16); \ | |
309 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, src, half, stride, \ | |
310 | stride, 16); \ | |
311 | } \ | |
312 | \ | |
313 | static void OPNAME ## qpel16_mc20_ ## MMX(uint8_t *dst, \ | |
314 | const uint8_t *src, \ | |
315 | ptrdiff_t stride) \ | |
316 | { \ | |
317 | ff_ ## OPNAME ## mpeg4_qpel16_h_lowpass_ ## MMX(dst, src, \ | |
318 | stride, stride, 16);\ | |
319 | } \ | |
320 | \ | |
321 | static void OPNAME ## qpel16_mc30_ ## MMX(uint8_t *dst, \ | |
322 | const uint8_t *src, \ | |
323 | ptrdiff_t stride) \ | |
324 | { \ | |
325 | uint64_t temp[32]; \ | |
326 | uint8_t *const half = (uint8_t*) temp; \ | |
327 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(half, src, 16, \ | |
328 | stride, 16); \ | |
329 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, src + 1, half, \ | |
330 | stride, stride, 16); \ | |
331 | } \ | |
332 | \ | |
333 | static void OPNAME ## qpel16_mc01_ ## MMX(uint8_t *dst, \ | |
334 | const uint8_t *src, \ | |
335 | ptrdiff_t stride) \ | |
336 | { \ | |
337 | uint64_t temp[32]; \ | |
338 | uint8_t *const half = (uint8_t *) temp; \ | |
339 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16, \ | |
340 | stride); \ | |
341 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, src, half, stride, \ | |
342 | stride, 16); \ | |
343 | } \ | |
344 | \ | |
345 | static void OPNAME ## qpel16_mc02_ ## MMX(uint8_t *dst, \ | |
346 | const uint8_t *src, \ | |
347 | ptrdiff_t stride) \ | |
348 | { \ | |
349 | ff_ ## OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, src, \ | |
350 | stride, stride); \ | |
351 | } \ | |
352 | \ | |
353 | static void OPNAME ## qpel16_mc03_ ## MMX(uint8_t *dst, \ | |
354 | const uint8_t *src, \ | |
355 | ptrdiff_t stride) \ | |
356 | { \ | |
357 | uint64_t temp[32]; \ | |
358 | uint8_t *const half = (uint8_t *) temp; \ | |
359 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16, \ | |
360 | stride); \ | |
361 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, src+stride, half, \ | |
362 | stride, stride, 16); \ | |
363 | } \ | |
364 | \ | |
365 | static void OPNAME ## qpel16_mc11_ ## MMX(uint8_t *dst, \ | |
366 | const uint8_t *src, \ | |
367 | ptrdiff_t stride) \ | |
368 | { \ | |
369 | uint64_t half[16 * 2 + 17 * 2]; \ | |
370 | uint8_t *const halfH = (uint8_t *) half + 256; \ | |
371 | uint8_t *const halfHV = (uint8_t *) half; \ | |
372 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
373 | stride, 17); \ | |
374 | ff_put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16, \ | |
375 | stride, 17); \ | |
376 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, \ | |
377 | 16, 16); \ | |
378 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, \ | |
379 | stride, 16, 16); \ | |
380 | } \ | |
381 | \ | |
382 | static void OPNAME ## qpel16_mc31_ ## MMX(uint8_t *dst, \ | |
383 | const uint8_t *src, \ | |
384 | ptrdiff_t stride) \ | |
385 | { \ | |
386 | uint64_t half[16 * 2 + 17 * 2]; \ | |
387 | uint8_t *const halfH = (uint8_t *) half + 256; \ | |
388 | uint8_t *const halfHV = (uint8_t *) half; \ | |
389 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
390 | stride, 17); \ | |
391 | ff_put ## RND ## pixels16_l2_ ## MMX(halfH, src + 1, halfH, 16, \ | |
392 | stride, 17); \ | |
393 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, \ | |
394 | 16, 16); \ | |
395 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, \ | |
396 | stride, 16, 16); \ | |
397 | } \ | |
398 | \ | |
399 | static void OPNAME ## qpel16_mc13_ ## MMX(uint8_t *dst, \ | |
400 | const uint8_t *src, \ | |
401 | ptrdiff_t stride) \ | |
402 | { \ | |
403 | uint64_t half[16 * 2 + 17 * 2]; \ | |
404 | uint8_t *const halfH = (uint8_t *) half + 256; \ | |
405 | uint8_t *const halfHV = (uint8_t *) half; \ | |
406 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
407 | stride, 17); \ | |
408 | ff_put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16, \ | |
409 | stride, 17); \ | |
410 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, \ | |
411 | 16, 16); \ | |
412 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, halfH + 16, halfHV, \ | |
413 | stride, 16, 16); \ | |
414 | } \ | |
415 | \ | |
416 | static void OPNAME ## qpel16_mc33_ ## MMX(uint8_t *dst, \ | |
417 | const uint8_t *src, \ | |
418 | ptrdiff_t stride) \ | |
419 | { \ | |
420 | uint64_t half[16 * 2 + 17 * 2]; \ | |
421 | uint8_t *const halfH = (uint8_t *) half + 256; \ | |
422 | uint8_t *const halfHV = (uint8_t *) half; \ | |
423 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
424 | stride, 17); \ | |
425 | ff_put ## RND ## pixels16_l2_ ## MMX(halfH, src + 1, halfH, 16, \ | |
426 | stride, 17); \ | |
427 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, \ | |
428 | 16, 16); \ | |
429 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, halfH + 16, halfHV, \ | |
430 | stride, 16, 16); \ | |
431 | } \ | |
432 | \ | |
433 | static void OPNAME ## qpel16_mc21_ ## MMX(uint8_t *dst, \ | |
434 | const uint8_t *src, \ | |
435 | ptrdiff_t stride) \ | |
436 | { \ | |
437 | uint64_t half[16 * 2 + 17 * 2]; \ | |
438 | uint8_t *const halfH = (uint8_t *) half + 256; \ | |
439 | uint8_t *const halfHV = (uint8_t *) half; \ | |
440 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
441 | stride, 17); \ | |
442 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, \ | |
443 | 16, 16); \ | |
444 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, halfH, halfHV, \ | |
445 | stride, 16, 16); \ | |
446 | } \ | |
447 | \ | |
448 | static void OPNAME ## qpel16_mc23_ ## MMX(uint8_t *dst, \ | |
449 | const uint8_t *src, \ | |
450 | ptrdiff_t stride) \ | |
451 | { \ | |
452 | uint64_t half[16 * 2 + 17 * 2]; \ | |
453 | uint8_t *const halfH = (uint8_t *) half + 256; \ | |
454 | uint8_t *const halfHV = (uint8_t *) half; \ | |
455 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
456 | stride, 17); \ | |
457 | ff_put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, \ | |
458 | 16, 16); \ | |
459 | ff_ ## OPNAME ## pixels16_l2_ ## MMX(dst, halfH + 16, halfHV, \ | |
460 | stride, 16, 16); \ | |
461 | } \ | |
462 | \ | |
463 | static void OPNAME ## qpel16_mc12_ ## MMX(uint8_t *dst, \ | |
464 | const uint8_t *src, \ | |
465 | ptrdiff_t stride) \ | |
466 | { \ | |
467 | uint64_t half[17 * 2]; \ | |
468 | uint8_t *const halfH = (uint8_t *) half; \ | |
469 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
470 | stride, 17); \ | |
471 | ff_put ## RND ## pixels16_l2_ ## MMX(halfH, src, halfH, 16, \ | |
472 | stride, 17); \ | |
473 | ff_ ## OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, \ | |
474 | stride, 16); \ | |
475 | } \ | |
476 | \ | |
477 | static void OPNAME ## qpel16_mc32_ ## MMX(uint8_t *dst, \ | |
478 | const uint8_t *src, \ | |
479 | ptrdiff_t stride) \ | |
480 | { \ | |
481 | uint64_t half[17 * 2]; \ | |
482 | uint8_t *const halfH = (uint8_t *) half; \ | |
483 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
484 | stride, 17); \ | |
485 | ff_put ## RND ## pixels16_l2_ ## MMX(halfH, src + 1, halfH, 16, \ | |
486 | stride, 17); \ | |
487 | ff_ ## OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, \ | |
488 | stride, 16); \ | |
489 | } \ | |
490 | \ | |
491 | static void OPNAME ## qpel16_mc22_ ## MMX(uint8_t *dst, \ | |
492 | const uint8_t *src, \ | |
493 | ptrdiff_t stride) \ | |
494 | { \ | |
495 | uint64_t half[17 * 2]; \ | |
496 | uint8_t *const halfH = (uint8_t *) half; \ | |
497 | ff_put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, \ | |
498 | stride, 17); \ | |
499 | ff_ ## OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, \ | |
500 | stride, 16); \ | |
501 | } | |
502 | ||
503 | QPEL_OP(put_, _, mmxext) | |
504 | QPEL_OP(avg_, _, mmxext) | |
505 | QPEL_OP(put_no_rnd_, _no_rnd_, mmxext) | |
506 | ||
507 | #endif /* HAVE_YASM */ | |
508 | ||
509 | #define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX) \ | |
510 | do { \ | |
511 | c->PFX ## _pixels_tab[IDX][ 0] = PREFIX ## PFX ## SIZE ## _mc00_ ## CPU; \ | |
512 | c->PFX ## _pixels_tab[IDX][ 1] = PREFIX ## PFX ## SIZE ## _mc10_ ## CPU; \ | |
513 | c->PFX ## _pixels_tab[IDX][ 2] = PREFIX ## PFX ## SIZE ## _mc20_ ## CPU; \ | |
514 | c->PFX ## _pixels_tab[IDX][ 3] = PREFIX ## PFX ## SIZE ## _mc30_ ## CPU; \ | |
515 | c->PFX ## _pixels_tab[IDX][ 4] = PREFIX ## PFX ## SIZE ## _mc01_ ## CPU; \ | |
516 | c->PFX ## _pixels_tab[IDX][ 5] = PREFIX ## PFX ## SIZE ## _mc11_ ## CPU; \ | |
517 | c->PFX ## _pixels_tab[IDX][ 6] = PREFIX ## PFX ## SIZE ## _mc21_ ## CPU; \ | |
518 | c->PFX ## _pixels_tab[IDX][ 7] = PREFIX ## PFX ## SIZE ## _mc31_ ## CPU; \ | |
519 | c->PFX ## _pixels_tab[IDX][ 8] = PREFIX ## PFX ## SIZE ## _mc02_ ## CPU; \ | |
520 | c->PFX ## _pixels_tab[IDX][ 9] = PREFIX ## PFX ## SIZE ## _mc12_ ## CPU; \ | |
521 | c->PFX ## _pixels_tab[IDX][10] = PREFIX ## PFX ## SIZE ## _mc22_ ## CPU; \ | |
522 | c->PFX ## _pixels_tab[IDX][11] = PREFIX ## PFX ## SIZE ## _mc32_ ## CPU; \ | |
523 | c->PFX ## _pixels_tab[IDX][12] = PREFIX ## PFX ## SIZE ## _mc03_ ## CPU; \ | |
524 | c->PFX ## _pixels_tab[IDX][13] = PREFIX ## PFX ## SIZE ## _mc13_ ## CPU; \ | |
525 | c->PFX ## _pixels_tab[IDX][14] = PREFIX ## PFX ## SIZE ## _mc23_ ## CPU; \ | |
526 | c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU; \ | |
527 | } while (0) | |
528 | ||
529 | av_cold void ff_qpeldsp_init_x86(QpelDSPContext *c) | |
530 | { | |
531 | int cpu_flags = av_get_cpu_flags(); | |
532 | ||
533 | if (X86_MMXEXT(cpu_flags)) { | |
534 | #if HAVE_MMXEXT_EXTERNAL | |
535 | SET_QPEL_FUNCS(avg_qpel, 0, 16, mmxext, ); | |
536 | SET_QPEL_FUNCS(avg_qpel, 1, 8, mmxext, ); | |
537 | ||
538 | SET_QPEL_FUNCS(put_qpel, 0, 16, mmxext, ); | |
539 | SET_QPEL_FUNCS(put_qpel, 1, 8, mmxext, ); | |
540 | SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmxext, ); | |
541 | SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmxext, ); | |
542 | #endif /* HAVE_MMXEXT_EXTERNAL */ | |
543 | } | |
544 | } |