cavsdsp_mmx.c
Go to the documentation of this file.
1 /*
2  * Chinese AVS video (AVS1-P2, JiZhun profile) decoder.
3  * Copyright (c) 2006 Stefan Gehrer <stefan.gehrer@gmx.de>
4  *
5  * MMX-optimized DSP functions, based on H.264 optimizations by
6  * Michael Niedermayer and Loren Merritt
7  *
8  * This file is part of Libav.
9  *
10  * Libav is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU Lesser General Public
12  * License as published by the Free Software Foundation; either
13  * version 2.1 of the License, or (at your option) any later version.
14  *
15  * Libav is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18  * Lesser General Public License for more details.
19  *
20  * You should have received a copy of the GNU Lesser General Public
21  * License along with Libav; if not, write to the Free Software
22  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23  */
24 
25 #include "libavutil/common.h"
26 #include "libavutil/cpu.h"
27 #include "libavutil/x86_cpu.h"
28 #include "libavcodec/dsputil.h"
29 #include "libavcodec/cavsdsp.h"
30 #include "dsputil_mmx.h"
31 
32 /*****************************************************************************
33  *
34  * inverse transform
35  *
36  ****************************************************************************/
37 
38 static inline void cavs_idct8_1d(int16_t *block, uint64_t bias)
39 {
40  __asm__ volatile(
41  "movq 112(%0), %%mm4 \n\t" /* mm4 = src7 */
42  "movq 16(%0), %%mm5 \n\t" /* mm5 = src1 */
43  "movq 80(%0), %%mm2 \n\t" /* mm2 = src5 */
44  "movq 48(%0), %%mm7 \n\t" /* mm7 = src3 */
45  "movq %%mm4, %%mm0 \n\t"
46  "movq %%mm5, %%mm3 \n\t"
47  "movq %%mm2, %%mm6 \n\t"
48  "movq %%mm7, %%mm1 \n\t"
49 
50  "paddw %%mm4, %%mm4 \n\t" /* mm4 = 2*src7 */
51  "paddw %%mm3, %%mm3 \n\t" /* mm3 = 2*src1 */
52  "paddw %%mm6, %%mm6 \n\t" /* mm6 = 2*src5 */
53  "paddw %%mm1, %%mm1 \n\t" /* mm1 = 2*src3 */
54  "paddw %%mm4, %%mm0 \n\t" /* mm0 = 3*src7 */
55  "paddw %%mm3, %%mm5 \n\t" /* mm5 = 3*src1 */
56  "paddw %%mm6, %%mm2 \n\t" /* mm2 = 3*src5 */
57  "paddw %%mm1, %%mm7 \n\t" /* mm7 = 3*src3 */
58  "psubw %%mm4, %%mm5 \n\t" /* mm5 = 3*src1 - 2*src7 = a0 */
59  "paddw %%mm6, %%mm7 \n\t" /* mm7 = 3*src3 + 2*src5 = a1 */
60  "psubw %%mm2, %%mm1 \n\t" /* mm1 = 2*src3 - 3*src5 = a2 */
61  "paddw %%mm0, %%mm3 \n\t" /* mm3 = 2*src1 + 3*src7 = a3 */
62 
63  "movq %%mm5, %%mm4 \n\t"
64  "movq %%mm7, %%mm6 \n\t"
65  "movq %%mm3, %%mm0 \n\t"
66  "movq %%mm1, %%mm2 \n\t"
67  SUMSUB_BA( %%mm7, %%mm5 ) /* mm7 = a0 + a1 mm5 = a0 - a1 */
68  "paddw %%mm3, %%mm7 \n\t" /* mm7 = a0 + a1 + a3 */
69  "paddw %%mm1, %%mm5 \n\t" /* mm5 = a0 - a1 + a2 */
70  "paddw %%mm7, %%mm7 \n\t"
71  "paddw %%mm5, %%mm5 \n\t"
72  "paddw %%mm6, %%mm7 \n\t" /* mm7 = b4 */
73  "paddw %%mm4, %%mm5 \n\t" /* mm5 = b5 */
74 
75  SUMSUB_BA( %%mm1, %%mm3 ) /* mm1 = a3 + a2 mm3 = a3 - a2 */
76  "psubw %%mm1, %%mm4 \n\t" /* mm4 = a0 - a2 - a3 */
77  "movq %%mm4, %%mm1 \n\t" /* mm1 = a0 - a2 - a3 */
78  "psubw %%mm6, %%mm3 \n\t" /* mm3 = a3 - a2 - a1 */
79  "paddw %%mm1, %%mm1 \n\t"
80  "paddw %%mm3, %%mm3 \n\t"
81  "psubw %%mm2, %%mm1 \n\t" /* mm1 = b7 */
82  "paddw %%mm0, %%mm3 \n\t" /* mm3 = b6 */
83 
84  "movq 32(%0), %%mm2 \n\t" /* mm2 = src2 */
85  "movq 96(%0), %%mm6 \n\t" /* mm6 = src6 */
86  "movq %%mm2, %%mm4 \n\t"
87  "movq %%mm6, %%mm0 \n\t"
88  "psllw $2, %%mm4 \n\t" /* mm4 = 4*src2 */
89  "psllw $2, %%mm6 \n\t" /* mm6 = 4*src6 */
90  "paddw %%mm4, %%mm2 \n\t" /* mm2 = 5*src2 */
91  "paddw %%mm6, %%mm0 \n\t" /* mm0 = 5*src6 */
92  "paddw %%mm2, %%mm2 \n\t"
93  "paddw %%mm0, %%mm0 \n\t"
94  "psubw %%mm0, %%mm4 \n\t" /* mm4 = 4*src2 - 10*src6 = a7 */
95  "paddw %%mm2, %%mm6 \n\t" /* mm6 = 4*src6 + 10*src2 = a6 */
96 
97  "movq (%0), %%mm2 \n\t" /* mm2 = src0 */
98  "movq 64(%0), %%mm0 \n\t" /* mm0 = src4 */
99  SUMSUB_BA( %%mm0, %%mm2 ) /* mm0 = src0+src4 mm2 = src0-src4 */
100  "psllw $3, %%mm0 \n\t"
101  "psllw $3, %%mm2 \n\t"
102  "paddw %1, %%mm0 \n\t" /* add rounding bias */
103  "paddw %1, %%mm2 \n\t" /* add rounding bias */
104 
105  SUMSUB_BA( %%mm6, %%mm0 ) /* mm6 = a4 + a6 mm0 = a4 - a6 */
106  SUMSUB_BA( %%mm4, %%mm2 ) /* mm4 = a5 + a7 mm2 = a5 - a7 */
107  SUMSUB_BA( %%mm7, %%mm6 ) /* mm7 = dst0 mm6 = dst7 */
108  SUMSUB_BA( %%mm5, %%mm4 ) /* mm5 = dst1 mm4 = dst6 */
109  SUMSUB_BA( %%mm3, %%mm2 ) /* mm3 = dst2 mm2 = dst5 */
110  SUMSUB_BA( %%mm1, %%mm0 ) /* mm1 = dst3 mm0 = dst4 */
111  :: "r"(block), "m"(bias)
112  );
113 }
114 
115 static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
116 {
117  int i;
118  DECLARE_ALIGNED(8, int16_t, b2)[64];
119 
120  for(i=0; i<2; i++){
121  DECLARE_ALIGNED(8, uint64_t, tmp);
122 
123  cavs_idct8_1d(block+4*i, ff_pw_4.a);
124 
125  __asm__ volatile(
126  "psraw $3, %%mm7 \n\t"
127  "psraw $3, %%mm6 \n\t"
128  "psraw $3, %%mm5 \n\t"
129  "psraw $3, %%mm4 \n\t"
130  "psraw $3, %%mm3 \n\t"
131  "psraw $3, %%mm2 \n\t"
132  "psraw $3, %%mm1 \n\t"
133  "psraw $3, %%mm0 \n\t"
134  "movq %%mm7, %0 \n\t"
135  TRANSPOSE4( %%mm0, %%mm2, %%mm4, %%mm6, %%mm7 )
136  "movq %%mm0, 8(%1) \n\t"
137  "movq %%mm6, 24(%1) \n\t"
138  "movq %%mm7, 40(%1) \n\t"
139  "movq %%mm4, 56(%1) \n\t"
140  "movq %0, %%mm7 \n\t"
141  TRANSPOSE4( %%mm7, %%mm5, %%mm3, %%mm1, %%mm0 )
142  "movq %%mm7, (%1) \n\t"
143  "movq %%mm1, 16(%1) \n\t"
144  "movq %%mm0, 32(%1) \n\t"
145  "movq %%mm3, 48(%1) \n\t"
146  : "=m"(tmp)
147  : "r"(b2+32*i)
148  : "memory"
149  );
150  }
151 
152  for(i=0; i<2; i++){
153  cavs_idct8_1d(b2+4*i, ff_pw_64.a);
154 
155  __asm__ volatile(
156  "psraw $7, %%mm7 \n\t"
157  "psraw $7, %%mm6 \n\t"
158  "psraw $7, %%mm5 \n\t"
159  "psraw $7, %%mm4 \n\t"
160  "psraw $7, %%mm3 \n\t"
161  "psraw $7, %%mm2 \n\t"
162  "psraw $7, %%mm1 \n\t"
163  "psraw $7, %%mm0 \n\t"
164  "movq %%mm7, (%0) \n\t"
165  "movq %%mm5, 16(%0) \n\t"
166  "movq %%mm3, 32(%0) \n\t"
167  "movq %%mm1, 48(%0) \n\t"
168  "movq %%mm0, 64(%0) \n\t"
169  "movq %%mm2, 80(%0) \n\t"
170  "movq %%mm4, 96(%0) \n\t"
171  "movq %%mm6, 112(%0) \n\t"
172  :: "r"(b2+4*i)
173  : "memory"
174  );
175  }
176 
177  ff_add_pixels_clamped_mmx(b2, dst, stride);
178 }
179 
180 /*****************************************************************************
181  *
182  * motion compensation
183  *
184  ****************************************************************************/
185 
186 /* vertical filter [-1 -2 96 42 -7 0] */
187 #define QPEL_CAVSV1(A,B,C,D,E,F,OP,MUL2) \
188  "movd (%0), "#F" \n\t"\
189  "movq "#C", %%mm6 \n\t"\
190  "pmullw %5, %%mm6 \n\t"\
191  "movq "#D", %%mm7 \n\t"\
192  "pmullw "MANGLE(MUL2)", %%mm7\n\t"\
193  "psllw $3, "#E" \n\t"\
194  "psubw "#E", %%mm6 \n\t"\
195  "psraw $3, "#E" \n\t"\
196  "paddw %%mm7, %%mm6 \n\t"\
197  "paddw "#E", %%mm6 \n\t"\
198  "paddw "#B", "#B" \n\t"\
199  "pxor %%mm7, %%mm7 \n\t"\
200  "add %2, %0 \n\t"\
201  "punpcklbw %%mm7, "#F" \n\t"\
202  "psubw "#B", %%mm6 \n\t"\
203  "psraw $1, "#B" \n\t"\
204  "psubw "#A", %%mm6 \n\t"\
205  "paddw %4, %%mm6 \n\t"\
206  "psraw $7, %%mm6 \n\t"\
207  "packuswb %%mm6, %%mm6 \n\t"\
208  OP(%%mm6, (%1), A, d) \
209  "add %3, %1 \n\t"
210 
211 /* vertical filter [ 0 -1 5 5 -1 0] */
212 #define QPEL_CAVSV2(A,B,C,D,E,F,OP,MUL2) \
213  "movd (%0), "#F" \n\t"\
214  "movq "#C", %%mm6 \n\t"\
215  "paddw "#D", %%mm6 \n\t"\
216  "pmullw %5, %%mm6 \n\t"\
217  "add %2, %0 \n\t"\
218  "punpcklbw %%mm7, "#F" \n\t"\
219  "psubw "#B", %%mm6 \n\t"\
220  "psubw "#E", %%mm6 \n\t"\
221  "paddw %4, %%mm6 \n\t"\
222  "psraw $3, %%mm6 \n\t"\
223  "packuswb %%mm6, %%mm6 \n\t"\
224  OP(%%mm6, (%1), A, d) \
225  "add %3, %1 \n\t"
226 
227 /* vertical filter [ 0 -7 42 96 -2 -1] */
228 #define QPEL_CAVSV3(A,B,C,D,E,F,OP,MUL2) \
229  "movd (%0), "#F" \n\t"\
230  "movq "#C", %%mm6 \n\t"\
231  "pmullw "MANGLE(MUL2)", %%mm6\n\t"\
232  "movq "#D", %%mm7 \n\t"\
233  "pmullw %5, %%mm7 \n\t"\
234  "psllw $3, "#B" \n\t"\
235  "psubw "#B", %%mm6 \n\t"\
236  "psraw $3, "#B" \n\t"\
237  "paddw %%mm7, %%mm6 \n\t"\
238  "paddw "#B", %%mm6 \n\t"\
239  "paddw "#E", "#E" \n\t"\
240  "pxor %%mm7, %%mm7 \n\t"\
241  "add %2, %0 \n\t"\
242  "punpcklbw %%mm7, "#F" \n\t"\
243  "psubw "#E", %%mm6 \n\t"\
244  "psraw $1, "#E" \n\t"\
245  "psubw "#F", %%mm6 \n\t"\
246  "paddw %4, %%mm6 \n\t"\
247  "psraw $7, %%mm6 \n\t"\
248  "packuswb %%mm6, %%mm6 \n\t"\
249  OP(%%mm6, (%1), A, d) \
250  "add %3, %1 \n\t"
251 
252 
253 #define QPEL_CAVSVNUM(VOP,OP,ADD,MUL1,MUL2)\
254  int w= 2;\
255  src -= 2*srcStride;\
256  \
257  while(w--){\
258  __asm__ volatile(\
259  "pxor %%mm7, %%mm7 \n\t"\
260  "movd (%0), %%mm0 \n\t"\
261  "add %2, %0 \n\t"\
262  "movd (%0), %%mm1 \n\t"\
263  "add %2, %0 \n\t"\
264  "movd (%0), %%mm2 \n\t"\
265  "add %2, %0 \n\t"\
266  "movd (%0), %%mm3 \n\t"\
267  "add %2, %0 \n\t"\
268  "movd (%0), %%mm4 \n\t"\
269  "add %2, %0 \n\t"\
270  "punpcklbw %%mm7, %%mm0 \n\t"\
271  "punpcklbw %%mm7, %%mm1 \n\t"\
272  "punpcklbw %%mm7, %%mm2 \n\t"\
273  "punpcklbw %%mm7, %%mm3 \n\t"\
274  "punpcklbw %%mm7, %%mm4 \n\t"\
275  VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, MUL2)\
276  VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, MUL2)\
277  VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, MUL2)\
278  VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, MUL2)\
279  VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, MUL2)\
280  VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, MUL2)\
281  VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, MUL2)\
282  VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, MUL2)\
283  \
284  : "+a"(src), "+c"(dst)\
285  : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride), "m"(ADD), "m"(MUL1)\
286  : "memory"\
287  );\
288  if(h==16){\
289  __asm__ volatile(\
290  VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, MUL2)\
291  VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, MUL2)\
292  VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, MUL2)\
293  VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, MUL2)\
294  VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, MUL2)\
295  VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, MUL2)\
296  VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, MUL2)\
297  VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, MUL2)\
298  \
299  : "+a"(src), "+c"(dst)\
300  : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride), "m"(ADD), "m"(MUL1)\
301  : "memory"\
302  );\
303  }\
304  src += 4-(h+5)*srcStride;\
305  dst += 4-h*dstStride;\
306  }
307 
308 #define QPEL_CAVS(OPNAME, OP, MMX)\
309 static void OPNAME ## cavs_qpel8_h_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
310  int h=8;\
311  __asm__ volatile(\
312  "pxor %%mm7, %%mm7 \n\t"\
313  "movq %5, %%mm6 \n\t"\
314  "1: \n\t"\
315  "movq (%0), %%mm0 \n\t"\
316  "movq 1(%0), %%mm2 \n\t"\
317  "movq %%mm0, %%mm1 \n\t"\
318  "movq %%mm2, %%mm3 \n\t"\
319  "punpcklbw %%mm7, %%mm0 \n\t"\
320  "punpckhbw %%mm7, %%mm1 \n\t"\
321  "punpcklbw %%mm7, %%mm2 \n\t"\
322  "punpckhbw %%mm7, %%mm3 \n\t"\
323  "paddw %%mm2, %%mm0 \n\t"\
324  "paddw %%mm3, %%mm1 \n\t"\
325  "pmullw %%mm6, %%mm0 \n\t"\
326  "pmullw %%mm6, %%mm1 \n\t"\
327  "movq -1(%0), %%mm2 \n\t"\
328  "movq 2(%0), %%mm4 \n\t"\
329  "movq %%mm2, %%mm3 \n\t"\
330  "movq %%mm4, %%mm5 \n\t"\
331  "punpcklbw %%mm7, %%mm2 \n\t"\
332  "punpckhbw %%mm7, %%mm3 \n\t"\
333  "punpcklbw %%mm7, %%mm4 \n\t"\
334  "punpckhbw %%mm7, %%mm5 \n\t"\
335  "paddw %%mm4, %%mm2 \n\t"\
336  "paddw %%mm3, %%mm5 \n\t"\
337  "psubw %%mm2, %%mm0 \n\t"\
338  "psubw %%mm5, %%mm1 \n\t"\
339  "movq %6, %%mm5 \n\t"\
340  "paddw %%mm5, %%mm0 \n\t"\
341  "paddw %%mm5, %%mm1 \n\t"\
342  "psraw $3, %%mm0 \n\t"\
343  "psraw $3, %%mm1 \n\t"\
344  "packuswb %%mm1, %%mm0 \n\t"\
345  OP(%%mm0, (%1),%%mm5, q) \
346  "add %3, %0 \n\t"\
347  "add %4, %1 \n\t"\
348  "decl %2 \n\t"\
349  " jnz 1b \n\t"\
350  : "+a"(src), "+c"(dst), "+m"(h)\
351  : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride), "m"(ff_pw_5), "m"(ff_pw_4)\
352  : "memory"\
353  );\
354 }\
355 \
356 static inline void OPNAME ## cavs_qpel8or16_v1_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
357  QPEL_CAVSVNUM(QPEL_CAVSV1,OP,ff_pw_64,ff_pw_96,ff_pw_42) \
358 }\
359 \
360 static inline void OPNAME ## cavs_qpel8or16_v2_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
361  QPEL_CAVSVNUM(QPEL_CAVSV2,OP,ff_pw_4,ff_pw_5,ff_pw_5) \
362 }\
363 \
364 static inline void OPNAME ## cavs_qpel8or16_v3_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
365  QPEL_CAVSVNUM(QPEL_CAVSV3,OP,ff_pw_64,ff_pw_96,ff_pw_42) \
366 }\
367 \
368 static void OPNAME ## cavs_qpel8_v1_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
369  OPNAME ## cavs_qpel8or16_v1_ ## MMX(dst , src , dstStride, srcStride, 8);\
370 }\
371 static void OPNAME ## cavs_qpel16_v1_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
372  OPNAME ## cavs_qpel8or16_v1_ ## MMX(dst , src , dstStride, srcStride, 16);\
373  OPNAME ## cavs_qpel8or16_v1_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
374 }\
375 \
376 static void OPNAME ## cavs_qpel8_v2_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
377  OPNAME ## cavs_qpel8or16_v2_ ## MMX(dst , src , dstStride, srcStride, 8);\
378 }\
379 static void OPNAME ## cavs_qpel16_v2_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
380  OPNAME ## cavs_qpel8or16_v2_ ## MMX(dst , src , dstStride, srcStride, 16);\
381  OPNAME ## cavs_qpel8or16_v2_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
382 }\
383 \
384 static void OPNAME ## cavs_qpel8_v3_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
385  OPNAME ## cavs_qpel8or16_v3_ ## MMX(dst , src , dstStride, srcStride, 8);\
386 }\
387 static void OPNAME ## cavs_qpel16_v3_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
388  OPNAME ## cavs_qpel8or16_v3_ ## MMX(dst , src , dstStride, srcStride, 16);\
389  OPNAME ## cavs_qpel8or16_v3_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
390 }\
391 \
392 static void OPNAME ## cavs_qpel16_h_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
393  OPNAME ## cavs_qpel8_h_ ## MMX(dst , src , dstStride, srcStride);\
394  OPNAME ## cavs_qpel8_h_ ## MMX(dst+8, src+8, dstStride, srcStride);\
395  src += 8*srcStride;\
396  dst += 8*dstStride;\
397  OPNAME ## cavs_qpel8_h_ ## MMX(dst , src , dstStride, srcStride);\
398  OPNAME ## cavs_qpel8_h_ ## MMX(dst+8, src+8, dstStride, srcStride);\
399 }\
400 
401 #define CAVS_MC(OPNAME, SIZE, MMX) \
402 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
403  OPNAME ## cavs_qpel ## SIZE ## _h_ ## MMX(dst, src, stride, stride);\
404 }\
405 \
406 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
407  OPNAME ## cavs_qpel ## SIZE ## _v1_ ## MMX(dst, src, stride, stride);\
408 }\
409 \
410 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
411  OPNAME ## cavs_qpel ## SIZE ## _v2_ ## MMX(dst, src, stride, stride);\
412 }\
413 \
414 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
415  OPNAME ## cavs_qpel ## SIZE ## _v3_ ## MMX(dst, src, stride, stride);\
416 }\
417 
418 #define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b " \n\t"
419 #define AVG_3DNOW_OP(a,b,temp, size) \
420 "mov" #size " " #b ", " #temp " \n\t"\
421 "pavgusb " #temp ", " #a " \n\t"\
422 "mov" #size " " #a ", " #b " \n\t"
423 #define AVG_MMX2_OP(a,b,temp, size) \
424 "mov" #size " " #b ", " #temp " \n\t"\
425 "pavgb " #temp ", " #a " \n\t"\
426 "mov" #size " " #a ", " #b " \n\t"
427 
428 QPEL_CAVS(put_, PUT_OP, 3dnow)
429 QPEL_CAVS(avg_, AVG_3DNOW_OP, 3dnow)
430 QPEL_CAVS(put_, PUT_OP, mmx2)
431 QPEL_CAVS(avg_, AVG_MMX2_OP, mmx2)
432 
433 CAVS_MC(put_, 8, 3dnow)
434 CAVS_MC(put_, 16,3dnow)
435 CAVS_MC(avg_, 8, 3dnow)
436 CAVS_MC(avg_, 16,3dnow)
437 CAVS_MC(put_, 8, mmx2)
438 CAVS_MC(put_, 16,mmx2)
439 CAVS_MC(avg_, 8, mmx2)
440 CAVS_MC(avg_, 16,mmx2)
441 
443 #define dspfunc(PFX, IDX, NUM) \
444  c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_mmx2; \
445  c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_mmx2; \
446  c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_mmx2; \
447  c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_mmx2; \
448  c->PFX ## _pixels_tab[IDX][12] = ff_ ## PFX ## NUM ## _mc03_mmx2; \
449 
450  dspfunc(put_cavs_qpel, 0, 16);
451  dspfunc(put_cavs_qpel, 1, 8);
452  dspfunc(avg_cavs_qpel, 0, 16);
453  dspfunc(avg_cavs_qpel, 1, 8);
454 #undef dspfunc
455  c->cavs_idct8_add = cavs_idct8_add_mmx;
456 }
457 
459 #define dspfunc(PFX, IDX, NUM) \
460  c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_mmx2; \
461  c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_3dnow; \
462  c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_3dnow; \
463  c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_3dnow; \
464  c->PFX ## _pixels_tab[IDX][12] = ff_ ## PFX ## NUM ## _mc03_3dnow; \
465 
466  dspfunc(put_cavs_qpel, 0, 16);
467  dspfunc(put_cavs_qpel, 1, 8);
468  dspfunc(avg_cavs_qpel, 0, 16);
469  dspfunc(avg_cavs_qpel, 1, 8);
470 #undef dspfunc
472 }
473 
475 {
476  int mm_flags = av_get_cpu_flags();
477 
478  if (mm_flags & AV_CPU_FLAG_MMX2) ff_cavsdsp_init_mmx2 (c, avctx);
479  if (mm_flags & AV_CPU_FLAG_3DNOW) ff_cavsdsp_init_3dnow(c, avctx);
480 }