FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
vp8dsp_init.c
Go to the documentation of this file.
1 /*
2  * VP8 DSP functions x86-optimized
3  * Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com>
4  * Copyright (c) 2010 Jason Garrett-Glaser <darkshikari@gmail.com>
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/cpu.h"
24 #include "libavutil/mem.h"
25 #include "libavutil/x86/asm.h"
26 #include "libavcodec/vp8dsp.h"
27 
28 #if HAVE_YASM
29 
30 /*
31  * MC functions
32  */
33 extern void ff_put_vp8_epel4_h4_mmxext(uint8_t *dst, ptrdiff_t dststride,
34  uint8_t *src, ptrdiff_t srcstride,
35  int height, int mx, int my);
36 extern void ff_put_vp8_epel4_h6_mmxext(uint8_t *dst, ptrdiff_t dststride,
37  uint8_t *src, ptrdiff_t srcstride,
38  int height, int mx, int my);
39 extern void ff_put_vp8_epel4_v4_mmxext(uint8_t *dst, ptrdiff_t dststride,
40  uint8_t *src, ptrdiff_t srcstride,
41  int height, int mx, int my);
42 extern void ff_put_vp8_epel4_v6_mmxext(uint8_t *dst, ptrdiff_t dststride,
43  uint8_t *src, ptrdiff_t srcstride,
44  int height, int mx, int my);
45 
46 extern void ff_put_vp8_epel8_h4_sse2 (uint8_t *dst, ptrdiff_t dststride,
47  uint8_t *src, ptrdiff_t srcstride,
48  int height, int mx, int my);
49 extern void ff_put_vp8_epel8_h6_sse2 (uint8_t *dst, ptrdiff_t dststride,
50  uint8_t *src, ptrdiff_t srcstride,
51  int height, int mx, int my);
52 extern void ff_put_vp8_epel8_v4_sse2 (uint8_t *dst, ptrdiff_t dststride,
53  uint8_t *src, ptrdiff_t srcstride,
54  int height, int mx, int my);
55 extern void ff_put_vp8_epel8_v6_sse2 (uint8_t *dst, ptrdiff_t dststride,
56  uint8_t *src, ptrdiff_t srcstride,
57  int height, int mx, int my);
58 
59 extern void ff_put_vp8_epel4_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
60  uint8_t *src, ptrdiff_t srcstride,
61  int height, int mx, int my);
62 extern void ff_put_vp8_epel4_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
63  uint8_t *src, ptrdiff_t srcstride,
64  int height, int mx, int my);
65 extern void ff_put_vp8_epel4_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
66  uint8_t *src, ptrdiff_t srcstride,
67  int height, int mx, int my);
68 extern void ff_put_vp8_epel4_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
69  uint8_t *src, ptrdiff_t srcstride,
70  int height, int mx, int my);
71 extern void ff_put_vp8_epel8_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
72  uint8_t *src, ptrdiff_t srcstride,
73  int height, int mx, int my);
74 extern void ff_put_vp8_epel8_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
75  uint8_t *src, ptrdiff_t srcstride,
76  int height, int mx, int my);
77 extern void ff_put_vp8_epel8_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride,
78  uint8_t *src, ptrdiff_t srcstride,
79  int height, int mx, int my);
80 extern void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride,
81  uint8_t *src, ptrdiff_t srcstride,
82  int height, int mx, int my);
83 
84 extern void ff_put_vp8_bilinear4_h_mmxext(uint8_t *dst, ptrdiff_t dststride,
85  uint8_t *src, ptrdiff_t srcstride,
86  int height, int mx, int my);
87 extern void ff_put_vp8_bilinear8_h_sse2 (uint8_t *dst, ptrdiff_t dststride,
88  uint8_t *src, ptrdiff_t srcstride,
89  int height, int mx, int my);
90 extern void ff_put_vp8_bilinear4_h_ssse3 (uint8_t *dst, ptrdiff_t dststride,
91  uint8_t *src, ptrdiff_t srcstride,
92  int height, int mx, int my);
93 extern void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, ptrdiff_t dststride,
94  uint8_t *src, ptrdiff_t srcstride,
95  int height, int mx, int my);
96 
97 extern void ff_put_vp8_bilinear4_v_mmxext(uint8_t *dst, ptrdiff_t dststride,
98  uint8_t *src, ptrdiff_t srcstride,
99  int height, int mx, int my);
100 extern void ff_put_vp8_bilinear8_v_sse2 (uint8_t *dst, ptrdiff_t dststride,
101  uint8_t *src, ptrdiff_t srcstride,
102  int height, int mx, int my);
103 extern void ff_put_vp8_bilinear4_v_ssse3 (uint8_t *dst, ptrdiff_t dststride,
104  uint8_t *src, ptrdiff_t srcstride,
105  int height, int mx, int my);
106 extern void ff_put_vp8_bilinear8_v_ssse3 (uint8_t *dst, ptrdiff_t dststride,
107  uint8_t *src, ptrdiff_t srcstride,
108  int height, int mx, int my);
109 
110 
111 extern void ff_put_vp8_pixels8_mmx (uint8_t *dst, ptrdiff_t dststride,
112  uint8_t *src, ptrdiff_t srcstride,
113  int height, int mx, int my);
114 extern void ff_put_vp8_pixels16_mmx(uint8_t *dst, ptrdiff_t dststride,
115  uint8_t *src, ptrdiff_t srcstride,
116  int height, int mx, int my);
117 extern void ff_put_vp8_pixels16_sse(uint8_t *dst, ptrdiff_t dststride,
118  uint8_t *src, ptrdiff_t srcstride,
119  int height, int mx, int my);
120 
121 #define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \
122 static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \
123  uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
124  ptrdiff_t srcstride, int height, int mx, int my) \
125 { \
126  ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
127  dst, dststride, src, srcstride, height, mx, my); \
128  ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
129  dst + 8, dststride, src + 8, srcstride, height, mx, my); \
130 }
131 #define TAP_W8(OPT, FILTERTYPE, TAPTYPE) \
132 static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
133  uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
134  ptrdiff_t srcstride, int height, int mx, int my) \
135 { \
136  ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
137  dst, dststride, src, srcstride, height, mx, my); \
138  ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
139  dst + 4, dststride, src + 4, srcstride, height, mx, my); \
140 }
141 
142 #if ARCH_X86_32
143 TAP_W8 (mmxext, epel, h4)
144 TAP_W8 (mmxext, epel, h6)
145 TAP_W16(mmxext, epel, h6)
146 TAP_W8 (mmxext, epel, v4)
147 TAP_W8 (mmxext, epel, v6)
148 TAP_W16(mmxext, epel, v6)
149 TAP_W8 (mmxext, bilinear, h)
150 TAP_W16(mmxext, bilinear, h)
151 TAP_W8 (mmxext, bilinear, v)
152 TAP_W16(mmxext, bilinear, v)
153 #endif
154 
155 TAP_W16(sse2, epel, h6)
156 TAP_W16(sse2, epel, v6)
157 TAP_W16(sse2, bilinear, h)
158 TAP_W16(sse2, bilinear, v)
159 
160 TAP_W16(ssse3, epel, h6)
161 TAP_W16(ssse3, epel, v6)
162 TAP_W16(ssse3, bilinear, h)
163 TAP_W16(ssse3, bilinear, v)
164 
165 #define HVTAP(OPT, ALIGN, TAPNUMX, TAPNUMY, SIZE, MAXHEIGHT) \
166 static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT( \
167  uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
168  ptrdiff_t srcstride, int height, int mx, int my) \
169 { \
170  DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + TAPNUMY - 1)]; \
171  uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \
172  src -= srcstride * (TAPNUMY / 2 - 1); \
173  ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## _ ## OPT( \
174  tmp, SIZE, src, srcstride, height + TAPNUMY - 1, mx, my); \
175  ff_put_vp8_epel ## SIZE ## _v ## TAPNUMY ## _ ## OPT( \
176  dst, dststride, tmpptr, SIZE, height, mx, my); \
177 }
178 
179 #if ARCH_X86_32
180 #define HVTAPMMX(x, y) \
181 HVTAP(mmxext, 8, x, y, 4, 8) \
182 HVTAP(mmxext, 8, x, y, 8, 16)
183 
184 HVTAP(mmxext, 8, 6, 6, 16, 16)
185 #else
186 #define HVTAPMMX(x, y) \
187 HVTAP(mmxext, 8, x, y, 4, 8)
188 #endif
189 
190 HVTAPMMX(4, 4)
191 HVTAPMMX(4, 6)
192 HVTAPMMX(6, 4)
193 HVTAPMMX(6, 6)
194 
195 #define HVTAPSSE2(x, y, w) \
196 HVTAP(sse2, 16, x, y, w, 16) \
197 HVTAP(ssse3, 16, x, y, w, 16)
198 
199 HVTAPSSE2(4, 4, 8)
200 HVTAPSSE2(4, 6, 8)
201 HVTAPSSE2(6, 4, 8)
202 HVTAPSSE2(6, 6, 8)
203 HVTAPSSE2(6, 6, 16)
204 
205 HVTAP(ssse3, 16, 4, 4, 4, 8)
206 HVTAP(ssse3, 16, 4, 6, 4, 8)
207 HVTAP(ssse3, 16, 6, 4, 4, 8)
208 HVTAP(ssse3, 16, 6, 6, 4, 8)
209 
210 #define HVBILIN(OPT, ALIGN, SIZE, MAXHEIGHT) \
211 static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \
212  uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \
213  ptrdiff_t srcstride, int height, int mx, int my) \
214 { \
215  DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + 2)]; \
216  ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT( \
217  tmp, SIZE, src, srcstride, height + 1, mx, my); \
218  ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT( \
219  dst, dststride, tmp, SIZE, height, mx, my); \
220 }
221 
222 HVBILIN(mmxext, 8, 4, 8)
223 #if ARCH_X86_32
224 HVBILIN(mmxext, 8, 8, 16)
225 HVBILIN(mmxext, 8, 16, 16)
226 #endif
227 HVBILIN(sse2, 8, 8, 16)
228 HVBILIN(sse2, 8, 16, 16)
229 HVBILIN(ssse3, 8, 4, 8)
230 HVBILIN(ssse3, 8, 8, 16)
231 HVBILIN(ssse3, 8, 16, 16)
232 
233 extern void ff_vp8_idct_dc_add_mmx(uint8_t *dst, int16_t block[16],
234  ptrdiff_t stride);
235 extern void ff_vp8_idct_dc_add_sse4(uint8_t *dst, int16_t block[16],
236  ptrdiff_t stride);
237 extern void ff_vp8_idct_dc_add4y_mmx(uint8_t *dst, int16_t block[4][16],
238  ptrdiff_t stride);
239 extern void ff_vp8_idct_dc_add4y_sse2(uint8_t *dst, int16_t block[4][16],
240  ptrdiff_t stride);
241 extern void ff_vp8_idct_dc_add4uv_mmx(uint8_t *dst, int16_t block[2][16],
242  ptrdiff_t stride);
243 extern void ff_vp8_luma_dc_wht_mmx(int16_t block[4][4][16], int16_t dc[16]);
244 extern void ff_vp8_luma_dc_wht_sse(int16_t block[4][4][16], int16_t dc[16]);
245 extern void ff_vp8_idct_add_mmx(uint8_t *dst, int16_t block[16],
246  ptrdiff_t stride);
247 extern void ff_vp8_idct_add_sse(uint8_t *dst, int16_t block[16],
248  ptrdiff_t stride);
249 
250 #define DECLARE_LOOP_FILTER(NAME)\
251 extern void ff_vp8_v_loop_filter_simple_ ## NAME(uint8_t *dst, \
252  ptrdiff_t stride, \
253  int flim);\
254 extern void ff_vp8_h_loop_filter_simple_ ## NAME(uint8_t *dst, \
255  ptrdiff_t stride, \
256  int flim);\
257 extern void ff_vp8_v_loop_filter16y_inner_ ## NAME (uint8_t *dst, \
258  ptrdiff_t stride,\
259  int e, int i, int hvt);\
260 extern void ff_vp8_h_loop_filter16y_inner_ ## NAME (uint8_t *dst, \
261  ptrdiff_t stride,\
262  int e, int i, int hvt);\
263 extern void ff_vp8_v_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \
264  uint8_t *dstV,\
265  ptrdiff_t s, \
266  int e, int i, int hvt);\
267 extern void ff_vp8_h_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \
268  uint8_t *dstV,\
269  ptrdiff_t s, \
270  int e, int i, int hvt);\
271 extern void ff_vp8_v_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \
272  ptrdiff_t stride,\
273  int e, int i, int hvt);\
274 extern void ff_vp8_h_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \
275  ptrdiff_t stride,\
276  int e, int i, int hvt);\
277 extern void ff_vp8_v_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \
278  uint8_t *dstV,\
279  ptrdiff_t s, \
280  int e, int i, int hvt);\
281 extern void ff_vp8_h_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \
282  uint8_t *dstV,\
283  ptrdiff_t s, \
284  int e, int i, int hvt);
285 
286 DECLARE_LOOP_FILTER(mmx)
287 DECLARE_LOOP_FILTER(mmxext)
288 DECLARE_LOOP_FILTER(sse2)
289 DECLARE_LOOP_FILTER(ssse3)
290 DECLARE_LOOP_FILTER(sse4)
291 
292 #endif /* HAVE_YASM */
293 
294 #define VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) \
295  c->put_vp8_epel_pixels_tab[IDX][0][2] = ff_put_vp8_epel ## SIZE ## _h6_ ## OPT; \
296  c->put_vp8_epel_pixels_tab[IDX][2][0] = ff_put_vp8_epel ## SIZE ## _v6_ ## OPT; \
297  c->put_vp8_epel_pixels_tab[IDX][2][2] = ff_put_vp8_epel ## SIZE ## _h6v6_ ## OPT
298 
299 #define VP8_MC_FUNC(IDX, SIZE, OPT) \
300  c->put_vp8_epel_pixels_tab[IDX][0][1] = ff_put_vp8_epel ## SIZE ## _h4_ ## OPT; \
301  c->put_vp8_epel_pixels_tab[IDX][1][0] = ff_put_vp8_epel ## SIZE ## _v4_ ## OPT; \
302  c->put_vp8_epel_pixels_tab[IDX][1][1] = ff_put_vp8_epel ## SIZE ## _h4v4_ ## OPT; \
303  c->put_vp8_epel_pixels_tab[IDX][1][2] = ff_put_vp8_epel ## SIZE ## _h6v4_ ## OPT; \
304  c->put_vp8_epel_pixels_tab[IDX][2][1] = ff_put_vp8_epel ## SIZE ## _h4v6_ ## OPT; \
305  VP8_LUMA_MC_FUNC(IDX, SIZE, OPT)
306 
307 #define VP8_BILINEAR_MC_FUNC(IDX, SIZE, OPT) \
308  c->put_vp8_bilinear_pixels_tab[IDX][0][1] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
309  c->put_vp8_bilinear_pixels_tab[IDX][0][2] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
310  c->put_vp8_bilinear_pixels_tab[IDX][1][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
311  c->put_vp8_bilinear_pixels_tab[IDX][1][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
312  c->put_vp8_bilinear_pixels_tab[IDX][1][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
313  c->put_vp8_bilinear_pixels_tab[IDX][2][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
314  c->put_vp8_bilinear_pixels_tab[IDX][2][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
315  c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT
316 
317 
319 {
320 #if HAVE_YASM
321  int mm_flags = av_get_cpu_flags();
322 
323  if (mm_flags & AV_CPU_FLAG_MMX) {
324  c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx;
325  c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx;
326 #if ARCH_X86_32
327  c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_mmx;
328  c->vp8_idct_add = ff_vp8_idct_add_mmx;
329  c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmx;
330  c->put_vp8_epel_pixels_tab[0][0][0] =
331  c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx;
332 #endif
333  c->put_vp8_epel_pixels_tab[1][0][0] =
334  c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx;
335 
336 #if ARCH_X86_32
337  c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx;
338  c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx;
339 
340  c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx;
341  c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx;
342  c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx;
343  c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx;
344 
345  c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmx;
346  c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_mmx;
347  c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_mmx;
348  c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_mmx;
349 #endif
350  }
351 
352  /* note that 4-tap width=16 functions are missing because w=16
353  * is only used for luma, and luma is always a copy or sixtap. */
354  if (mm_flags & AV_CPU_FLAG_MMXEXT) {
355  VP8_MC_FUNC(2, 4, mmxext);
356  VP8_BILINEAR_MC_FUNC(2, 4, mmxext);
357 #if ARCH_X86_32
358  VP8_LUMA_MC_FUNC(0, 16, mmxext);
359  VP8_MC_FUNC(1, 8, mmxext);
360  VP8_BILINEAR_MC_FUNC(0, 16, mmxext);
361  VP8_BILINEAR_MC_FUNC(1, 8, mmxext);
362 
363  c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmxext;
364  c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmxext;
365 
366  c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmxext;
367  c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmxext;
368  c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmxext;
369  c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmxext;
370 
371  c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmxext;
372  c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_mmxext;
373  c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_mmxext;
374  c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_mmxext;
375 #endif
376  }
377 
378  if (mm_flags & AV_CPU_FLAG_SSE) {
379  c->vp8_idct_add = ff_vp8_idct_add_sse;
380  c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse;
381  c->put_vp8_epel_pixels_tab[0][0][0] =
382  c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse;
383  }
384 
385  if (mm_flags & (AV_CPU_FLAG_SSE2|AV_CPU_FLAG_SSE2SLOW)) {
386  VP8_LUMA_MC_FUNC(0, 16, sse2);
387  VP8_MC_FUNC(1, 8, sse2);
388  VP8_BILINEAR_MC_FUNC(0, 16, sse2);
389  VP8_BILINEAR_MC_FUNC(1, 8, sse2);
390 
391  c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2;
392 
393  c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2;
394  c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_sse2;
395 
396  c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_sse2;
397  c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_sse2;
398  }
399 
400  if (mm_flags & AV_CPU_FLAG_SSE2) {
401  c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_sse2;
402 
403  c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2;
404 
405  c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_sse2;
406  c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_sse2;
407 
408  c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_sse2;
409  c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_sse2;
410  }
411 
412  if (mm_flags & AV_CPU_FLAG_SSSE3) {
413  VP8_LUMA_MC_FUNC(0, 16, ssse3);
414  VP8_MC_FUNC(1, 8, ssse3);
415  VP8_MC_FUNC(2, 4, ssse3);
416  VP8_BILINEAR_MC_FUNC(0, 16, ssse3);
417  VP8_BILINEAR_MC_FUNC(1, 8, ssse3);
418  VP8_BILINEAR_MC_FUNC(2, 4, ssse3);
419 
420  c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_ssse3;
421  c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_ssse3;
422 
423  c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_ssse3;
424  c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_ssse3;
425  c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_ssse3;
426  c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_ssse3;
427 
428  c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_ssse3;
429  c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_ssse3;
430  c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_ssse3;
431  c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_ssse3;
432  }
433 
434  if (mm_flags & AV_CPU_FLAG_SSE4) {
435  c->vp8_idct_dc_add = ff_vp8_idct_dc_add_sse4;
436 
437  c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse4;
438  c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_sse4;
439  c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_sse4;
440  }
441 #endif /* HAVE_YASM */
442 }