1 |
|
|
/* |
2 |
|
|
* VP8 DSP functions x86-optimized |
3 |
|
|
* Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com> |
4 |
|
|
* Copyright (c) 2010 Fiona Glaser <fiona@x264.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/attributes.h" |
24 |
|
|
#include "libavutil/cpu.h" |
25 |
|
|
#include "libavutil/mem.h" |
26 |
|
|
#include "libavutil/x86/cpu.h" |
27 |
|
|
#include "libavcodec/vp8dsp.h" |
28 |
|
|
|
29 |
|
|
#if HAVE_X86ASM |
30 |
|
|
|
31 |
|
|
/* |
32 |
|
|
* MC functions |
33 |
|
|
*/ |
34 |
|
|
void ff_put_vp8_epel4_h4_mmxext(uint8_t *dst, ptrdiff_t dststride, |
35 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
36 |
|
|
int height, int mx, int my); |
37 |
|
|
void ff_put_vp8_epel4_h6_mmxext(uint8_t *dst, ptrdiff_t dststride, |
38 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
39 |
|
|
int height, int mx, int my); |
40 |
|
|
void ff_put_vp8_epel4_v4_mmxext(uint8_t *dst, ptrdiff_t dststride, |
41 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
42 |
|
|
int height, int mx, int my); |
43 |
|
|
void ff_put_vp8_epel4_v6_mmxext(uint8_t *dst, ptrdiff_t dststride, |
44 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
45 |
|
|
int height, int mx, int my); |
46 |
|
|
|
47 |
|
|
void ff_put_vp8_epel8_h4_sse2 (uint8_t *dst, ptrdiff_t dststride, |
48 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
49 |
|
|
int height, int mx, int my); |
50 |
|
|
void ff_put_vp8_epel8_h6_sse2 (uint8_t *dst, ptrdiff_t dststride, |
51 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
52 |
|
|
int height, int mx, int my); |
53 |
|
|
void ff_put_vp8_epel8_v4_sse2 (uint8_t *dst, ptrdiff_t dststride, |
54 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
55 |
|
|
int height, int mx, int my); |
56 |
|
|
void ff_put_vp8_epel8_v6_sse2 (uint8_t *dst, ptrdiff_t dststride, |
57 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
58 |
|
|
int height, int mx, int my); |
59 |
|
|
|
60 |
|
|
void ff_put_vp8_epel4_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
61 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
62 |
|
|
int height, int mx, int my); |
63 |
|
|
void ff_put_vp8_epel4_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
64 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
65 |
|
|
int height, int mx, int my); |
66 |
|
|
void ff_put_vp8_epel4_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
67 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
68 |
|
|
int height, int mx, int my); |
69 |
|
|
void ff_put_vp8_epel4_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
70 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
71 |
|
|
int height, int mx, int my); |
72 |
|
|
void ff_put_vp8_epel8_h4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
73 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
74 |
|
|
int height, int mx, int my); |
75 |
|
|
void ff_put_vp8_epel8_h6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
76 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
77 |
|
|
int height, int mx, int my); |
78 |
|
|
void ff_put_vp8_epel8_v4_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
79 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
80 |
|
|
int height, int mx, int my); |
81 |
|
|
void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
82 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
83 |
|
|
int height, int mx, int my); |
84 |
|
|
|
85 |
|
|
void ff_put_vp8_bilinear4_h_mmxext(uint8_t *dst, ptrdiff_t dststride, |
86 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
87 |
|
|
int height, int mx, int my); |
88 |
|
|
void ff_put_vp8_bilinear8_h_sse2 (uint8_t *dst, ptrdiff_t dststride, |
89 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
90 |
|
|
int height, int mx, int my); |
91 |
|
|
void ff_put_vp8_bilinear4_h_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
92 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
93 |
|
|
int height, int mx, int my); |
94 |
|
|
void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
95 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
96 |
|
|
int height, int mx, int my); |
97 |
|
|
|
98 |
|
|
void ff_put_vp8_bilinear4_v_mmxext(uint8_t *dst, ptrdiff_t dststride, |
99 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
100 |
|
|
int height, int mx, int my); |
101 |
|
|
void ff_put_vp8_bilinear8_v_sse2 (uint8_t *dst, ptrdiff_t dststride, |
102 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
103 |
|
|
int height, int mx, int my); |
104 |
|
|
void ff_put_vp8_bilinear4_v_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
105 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
106 |
|
|
int height, int mx, int my); |
107 |
|
|
void ff_put_vp8_bilinear8_v_ssse3 (uint8_t *dst, ptrdiff_t dststride, |
108 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
109 |
|
|
int height, int mx, int my); |
110 |
|
|
|
111 |
|
|
|
112 |
|
|
void ff_put_vp8_pixels8_mmx (uint8_t *dst, ptrdiff_t dststride, |
113 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
114 |
|
|
int height, int mx, int my); |
115 |
|
|
void ff_put_vp8_pixels16_mmx(uint8_t *dst, ptrdiff_t dststride, |
116 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
117 |
|
|
int height, int mx, int my); |
118 |
|
|
void ff_put_vp8_pixels16_sse(uint8_t *dst, ptrdiff_t dststride, |
119 |
|
|
uint8_t *src, ptrdiff_t srcstride, |
120 |
|
|
int height, int mx, int my); |
121 |
|
|
|
122 |
|
|
#define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \ |
123 |
|
|
static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \ |
124 |
|
|
uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
125 |
|
|
ptrdiff_t srcstride, int height, int mx, int my) \ |
126 |
|
|
{ \ |
127 |
|
|
ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
128 |
|
|
dst, dststride, src, srcstride, height, mx, my); \ |
129 |
|
|
ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
130 |
|
|
dst + 8, dststride, src + 8, srcstride, height, mx, my); \ |
131 |
|
|
} |
132 |
|
|
#define TAP_W8(OPT, FILTERTYPE, TAPTYPE) \ |
133 |
|
|
static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \ |
134 |
|
|
uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
135 |
|
|
ptrdiff_t srcstride, int height, int mx, int my) \ |
136 |
|
|
{ \ |
137 |
|
|
ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \ |
138 |
|
|
dst, dststride, src, srcstride, height, mx, my); \ |
139 |
|
|
ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \ |
140 |
|
|
dst + 4, dststride, src + 4, srcstride, height, mx, my); \ |
141 |
|
|
} |
142 |
|
|
|
143 |
|
|
#if ARCH_X86_32 |
144 |
|
|
TAP_W8 (mmxext, epel, h4) |
145 |
|
|
TAP_W8 (mmxext, epel, h6) |
146 |
|
|
TAP_W16(mmxext, epel, h6) |
147 |
|
|
TAP_W8 (mmxext, epel, v4) |
148 |
|
|
TAP_W8 (mmxext, epel, v6) |
149 |
|
|
TAP_W16(mmxext, epel, v6) |
150 |
|
|
TAP_W8 (mmxext, bilinear, h) |
151 |
|
|
TAP_W16(mmxext, bilinear, h) |
152 |
|
|
TAP_W8 (mmxext, bilinear, v) |
153 |
|
|
TAP_W16(mmxext, bilinear, v) |
154 |
|
|
#endif |
155 |
|
|
|
156 |
|
4 |
TAP_W16(sse2, epel, h6) |
157 |
|
4 |
TAP_W16(sse2, epel, v6) |
158 |
|
4 |
TAP_W16(sse2, bilinear, h) |
159 |
|
4 |
TAP_W16(sse2, bilinear, v) |
160 |
|
|
|
161 |
|
2 |
TAP_W16(ssse3, epel, h6) |
162 |
|
2 |
TAP_W16(ssse3, epel, v6) |
163 |
|
2 |
TAP_W16(ssse3, bilinear, h) |
164 |
|
2 |
TAP_W16(ssse3, bilinear, v) |
165 |
|
|
|
166 |
|
|
#define HVTAP(OPT, ALIGN, TAPNUMX, TAPNUMY, SIZE, MAXHEIGHT) \ |
167 |
|
|
static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT( \ |
168 |
|
|
uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
169 |
|
|
ptrdiff_t srcstride, int height, int mx, int my) \ |
170 |
|
|
{ \ |
171 |
|
|
LOCAL_ALIGNED(ALIGN, uint8_t, tmp, [SIZE * (MAXHEIGHT + TAPNUMY - 1)]); \ |
172 |
|
|
uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \ |
173 |
|
|
src -= srcstride * (TAPNUMY / 2 - 1); \ |
174 |
|
|
ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## _ ## OPT( \ |
175 |
|
|
tmp, SIZE, src, srcstride, height + TAPNUMY - 1, mx, my); \ |
176 |
|
|
ff_put_vp8_epel ## SIZE ## _v ## TAPNUMY ## _ ## OPT( \ |
177 |
|
|
dst, dststride, tmpptr, SIZE, height, mx, my); \ |
178 |
|
|
} |
179 |
|
|
|
180 |
|
|
#if ARCH_X86_32 |
181 |
|
|
#define HVTAPMMX(x, y) \ |
182 |
|
|
HVTAP(mmxext, 8, x, y, 4, 8) \ |
183 |
|
|
HVTAP(mmxext, 8, x, y, 8, 16) |
184 |
|
|
|
185 |
|
|
HVTAP(mmxext, 8, 6, 6, 16, 16) |
186 |
|
|
#else |
187 |
|
|
#define HVTAPMMX(x, y) \ |
188 |
|
|
HVTAP(mmxext, 8, x, y, 4, 8) |
189 |
|
|
#endif |
190 |
|
|
|
191 |
|
2 |
HVTAPMMX(4, 4) |
192 |
|
2 |
HVTAPMMX(4, 6) |
193 |
|
2 |
HVTAPMMX(6, 4) |
194 |
|
2 |
HVTAPMMX(6, 6) |
195 |
|
|
|
196 |
|
|
#define HVTAPSSE2(x, y, w) \ |
197 |
|
|
HVTAP(sse2, 16, x, y, w, 16) \ |
198 |
|
|
HVTAP(ssse3, 16, x, y, w, 16) |
199 |
|
|
|
200 |
|
6 |
HVTAPSSE2(4, 4, 8) |
201 |
|
3 |
HVTAPSSE2(4, 6, 8) |
202 |
|
3 |
HVTAPSSE2(6, 4, 8) |
203 |
|
3 |
HVTAPSSE2(6, 6, 8) |
204 |
|
3 |
HVTAPSSE2(6, 6, 16) |
205 |
|
|
|
206 |
|
1 |
HVTAP(ssse3, 16, 4, 4, 4, 8) |
207 |
|
1 |
HVTAP(ssse3, 16, 4, 6, 4, 8) |
208 |
|
1 |
HVTAP(ssse3, 16, 6, 4, 4, 8) |
209 |
|
1 |
HVTAP(ssse3, 16, 6, 6, 4, 8) |
210 |
|
|
|
211 |
|
|
#define HVBILIN(OPT, ALIGN, SIZE, MAXHEIGHT) \ |
212 |
|
|
static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \ |
213 |
|
|
uint8_t *dst, ptrdiff_t dststride, uint8_t *src, \ |
214 |
|
|
ptrdiff_t srcstride, int height, int mx, int my) \ |
215 |
|
|
{ \ |
216 |
|
|
LOCAL_ALIGNED(ALIGN, uint8_t, tmp, [SIZE * (MAXHEIGHT + 2)]); \ |
217 |
|
|
ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT( \ |
218 |
|
|
tmp, SIZE, src, srcstride, height + 1, mx, my); \ |
219 |
|
|
ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT( \ |
220 |
|
|
dst, dststride, tmp, SIZE, height, mx, my); \ |
221 |
|
|
} |
222 |
|
|
|
223 |
|
2 |
HVBILIN(mmxext, 8, 4, 8) |
224 |
|
|
#if ARCH_X86_32 |
225 |
|
|
HVBILIN(mmxext, 8, 8, 16) |
226 |
|
|
HVBILIN(mmxext, 8, 16, 16) |
227 |
|
|
#endif |
228 |
|
2 |
HVBILIN(sse2, 8, 8, 16) |
229 |
|
2 |
HVBILIN(sse2, 8, 16, 16) |
230 |
|
1 |
HVBILIN(ssse3, 8, 4, 8) |
231 |
|
1 |
HVBILIN(ssse3, 8, 8, 16) |
232 |
|
1 |
HVBILIN(ssse3, 8, 16, 16) |
233 |
|
|
|
234 |
|
|
void ff_vp8_idct_dc_add_mmx(uint8_t *dst, int16_t block[16], |
235 |
|
|
ptrdiff_t stride); |
236 |
|
|
void ff_vp8_idct_dc_add_sse2(uint8_t *dst, int16_t block[16], |
237 |
|
|
ptrdiff_t stride); |
238 |
|
|
void ff_vp8_idct_dc_add_sse4(uint8_t *dst, int16_t block[16], |
239 |
|
|
ptrdiff_t stride); |
240 |
|
|
void ff_vp8_idct_dc_add4y_mmx(uint8_t *dst, int16_t block[4][16], |
241 |
|
|
ptrdiff_t stride); |
242 |
|
|
void ff_vp8_idct_dc_add4y_sse2(uint8_t *dst, int16_t block[4][16], |
243 |
|
|
ptrdiff_t stride); |
244 |
|
|
void ff_vp8_idct_dc_add4uv_mmx(uint8_t *dst, int16_t block[2][16], |
245 |
|
|
ptrdiff_t stride); |
246 |
|
|
void ff_vp8_luma_dc_wht_mmx(int16_t block[4][4][16], int16_t dc[16]); |
247 |
|
|
void ff_vp8_luma_dc_wht_sse(int16_t block[4][4][16], int16_t dc[16]); |
248 |
|
|
void ff_vp8_idct_add_mmx(uint8_t *dst, int16_t block[16], ptrdiff_t stride); |
249 |
|
|
void ff_vp8_idct_add_sse(uint8_t *dst, int16_t block[16], ptrdiff_t stride); |
250 |
|
|
|
251 |
|
|
#define DECLARE_LOOP_FILTER(NAME) \ |
252 |
|
|
void ff_vp8_v_loop_filter_simple_ ## NAME(uint8_t *dst, \ |
253 |
|
|
ptrdiff_t stride, \ |
254 |
|
|
int flim); \ |
255 |
|
|
void ff_vp8_h_loop_filter_simple_ ## NAME(uint8_t *dst, \ |
256 |
|
|
ptrdiff_t stride, \ |
257 |
|
|
int flim); \ |
258 |
|
|
void ff_vp8_v_loop_filter16y_inner_ ## NAME (uint8_t *dst, \ |
259 |
|
|
ptrdiff_t stride, \ |
260 |
|
|
int e, int i, int hvt); \ |
261 |
|
|
void ff_vp8_h_loop_filter16y_inner_ ## NAME (uint8_t *dst, \ |
262 |
|
|
ptrdiff_t stride, \ |
263 |
|
|
int e, int i, int hvt); \ |
264 |
|
|
void ff_vp8_v_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \ |
265 |
|
|
uint8_t *dstV, \ |
266 |
|
|
ptrdiff_t s, \ |
267 |
|
|
int e, int i, int hvt); \ |
268 |
|
|
void ff_vp8_h_loop_filter8uv_inner_ ## NAME (uint8_t *dstU, \ |
269 |
|
|
uint8_t *dstV, \ |
270 |
|
|
ptrdiff_t s, \ |
271 |
|
|
int e, int i, int hvt); \ |
272 |
|
|
void ff_vp8_v_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \ |
273 |
|
|
ptrdiff_t stride, \ |
274 |
|
|
int e, int i, int hvt); \ |
275 |
|
|
void ff_vp8_h_loop_filter16y_mbedge_ ## NAME(uint8_t *dst, \ |
276 |
|
|
ptrdiff_t stride, \ |
277 |
|
|
int e, int i, int hvt); \ |
278 |
|
|
void ff_vp8_v_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \ |
279 |
|
|
uint8_t *dstV, \ |
280 |
|
|
ptrdiff_t s, \ |
281 |
|
|
int e, int i, int hvt); \ |
282 |
|
|
void ff_vp8_h_loop_filter8uv_mbedge_ ## NAME(uint8_t *dstU, \ |
283 |
|
|
uint8_t *dstV, \ |
284 |
|
|
ptrdiff_t s, \ |
285 |
|
|
int e, int i, int hvt); |
286 |
|
|
|
287 |
|
|
DECLARE_LOOP_FILTER(mmx) |
288 |
|
|
DECLARE_LOOP_FILTER(mmxext) |
289 |
|
|
DECLARE_LOOP_FILTER(sse2) |
290 |
|
|
DECLARE_LOOP_FILTER(ssse3) |
291 |
|
|
DECLARE_LOOP_FILTER(sse4) |
292 |
|
|
|
293 |
|
|
#endif /* HAVE_X86ASM */ |
294 |
|
|
|
295 |
|
|
#define VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) \ |
296 |
|
|
c->put_vp8_epel_pixels_tab[IDX][0][2] = ff_put_vp8_epel ## SIZE ## _h6_ ## OPT; \ |
297 |
|
|
c->put_vp8_epel_pixels_tab[IDX][2][0] = ff_put_vp8_epel ## SIZE ## _v6_ ## OPT; \ |
298 |
|
|
c->put_vp8_epel_pixels_tab[IDX][2][2] = ff_put_vp8_epel ## SIZE ## _h6v6_ ## OPT |
299 |
|
|
|
300 |
|
|
#define VP8_MC_FUNC(IDX, SIZE, OPT) \ |
301 |
|
|
c->put_vp8_epel_pixels_tab[IDX][0][1] = ff_put_vp8_epel ## SIZE ## _h4_ ## OPT; \ |
302 |
|
|
c->put_vp8_epel_pixels_tab[IDX][1][0] = ff_put_vp8_epel ## SIZE ## _v4_ ## OPT; \ |
303 |
|
|
c->put_vp8_epel_pixels_tab[IDX][1][1] = ff_put_vp8_epel ## SIZE ## _h4v4_ ## OPT; \ |
304 |
|
|
c->put_vp8_epel_pixels_tab[IDX][1][2] = ff_put_vp8_epel ## SIZE ## _h6v4_ ## OPT; \ |
305 |
|
|
c->put_vp8_epel_pixels_tab[IDX][2][1] = ff_put_vp8_epel ## SIZE ## _h4v6_ ## OPT; \ |
306 |
|
|
VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) |
307 |
|
|
|
308 |
|
|
#define VP8_BILINEAR_MC_FUNC(IDX, SIZE, OPT) \ |
309 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][0][1] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \ |
310 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][0][2] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \ |
311 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][1][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \ |
312 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][1][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
313 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][1][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
314 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][2][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \ |
315 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][2][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \ |
316 |
|
|
c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT |
317 |
|
|
|
318 |
|
|
|
319 |
|
73 |
av_cold void ff_vp78dsp_init_x86(VP8DSPContext *c) |
320 |
|
|
{ |
321 |
|
|
#if HAVE_X86ASM |
322 |
|
73 |
int cpu_flags = av_get_cpu_flags(); |
323 |
|
|
|
324 |
✓✓ |
73 |
if (EXTERNAL_MMX(cpu_flags)) { |
325 |
|
|
#if ARCH_X86_32 |
326 |
|
|
c->put_vp8_epel_pixels_tab[0][0][0] = |
327 |
|
|
c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx; |
328 |
|
|
#endif |
329 |
|
20 |
c->put_vp8_epel_pixels_tab[1][0][0] = |
330 |
|
20 |
c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_mmx; |
331 |
|
|
} |
332 |
|
|
|
333 |
|
|
/* note that 4-tap width=16 functions are missing because w=16 |
334 |
|
|
* is only used for luma, and luma is always a copy or sixtap. */ |
335 |
✓✓ |
73 |
if (EXTERNAL_MMXEXT(cpu_flags)) { |
336 |
|
19 |
VP8_MC_FUNC(2, 4, mmxext); |
337 |
|
19 |
VP8_BILINEAR_MC_FUNC(2, 4, mmxext); |
338 |
|
|
#if ARCH_X86_32 |
339 |
|
|
VP8_LUMA_MC_FUNC(0, 16, mmxext); |
340 |
|
|
VP8_MC_FUNC(1, 8, mmxext); |
341 |
|
|
VP8_BILINEAR_MC_FUNC(0, 16, mmxext); |
342 |
|
|
VP8_BILINEAR_MC_FUNC(1, 8, mmxext); |
343 |
|
|
#endif |
344 |
|
|
} |
345 |
|
|
|
346 |
✓✓ |
73 |
if (EXTERNAL_SSE(cpu_flags)) { |
347 |
|
18 |
c->put_vp8_epel_pixels_tab[0][0][0] = |
348 |
|
18 |
c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; |
349 |
|
|
} |
350 |
|
|
|
351 |
✓✓✗✓
✗✗ |
73 |
if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { |
352 |
|
17 |
VP8_LUMA_MC_FUNC(0, 16, sse2); |
353 |
|
17 |
VP8_MC_FUNC(1, 8, sse2); |
354 |
|
17 |
VP8_BILINEAR_MC_FUNC(0, 16, sse2); |
355 |
|
17 |
VP8_BILINEAR_MC_FUNC(1, 8, sse2); |
356 |
|
|
} |
357 |
|
|
|
358 |
✓✓ |
73 |
if (EXTERNAL_SSSE3(cpu_flags)) { |
359 |
|
15 |
VP8_LUMA_MC_FUNC(0, 16, ssse3); |
360 |
|
15 |
VP8_MC_FUNC(1, 8, ssse3); |
361 |
|
15 |
VP8_MC_FUNC(2, 4, ssse3); |
362 |
|
15 |
VP8_BILINEAR_MC_FUNC(0, 16, ssse3); |
363 |
|
15 |
VP8_BILINEAR_MC_FUNC(1, 8, ssse3); |
364 |
|
15 |
VP8_BILINEAR_MC_FUNC(2, 4, ssse3); |
365 |
|
|
} |
366 |
|
|
#endif /* HAVE_X86ASM */ |
367 |
|
73 |
} |
368 |
|
|
|
369 |
|
135 |
av_cold void ff_vp8dsp_init_x86(VP8DSPContext *c) |
370 |
|
|
{ |
371 |
|
|
#if HAVE_X86ASM |
372 |
|
135 |
int cpu_flags = av_get_cpu_flags(); |
373 |
|
|
|
374 |
✓✓ |
135 |
if (EXTERNAL_MMX(cpu_flags)) { |
375 |
|
80 |
c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx; |
376 |
|
|
#if ARCH_X86_32 |
377 |
|
|
c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx; |
378 |
|
|
c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_mmx; |
379 |
|
|
c->vp8_idct_add = ff_vp8_idct_add_mmx; |
380 |
|
|
c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmx; |
381 |
|
|
|
382 |
|
|
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmx; |
383 |
|
|
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmx; |
384 |
|
|
|
385 |
|
|
c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmx; |
386 |
|
|
c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmx; |
387 |
|
|
c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmx; |
388 |
|
|
c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmx; |
389 |
|
|
|
390 |
|
|
c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmx; |
391 |
|
|
c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_mmx; |
392 |
|
|
c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_mmx; |
393 |
|
|
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_mmx; |
394 |
|
|
#endif |
395 |
|
|
} |
396 |
|
|
|
397 |
|
|
/* note that 4-tap width=16 functions are missing because w=16 |
398 |
|
|
* is only used for luma, and luma is always a copy or sixtap. */ |
399 |
|
135 |
if (EXTERNAL_MMXEXT(cpu_flags)) { |
400 |
|
|
#if ARCH_X86_32 |
401 |
|
|
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_mmxext; |
402 |
|
|
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_mmxext; |
403 |
|
|
|
404 |
|
|
c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_mmxext; |
405 |
|
|
c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_mmxext; |
406 |
|
|
c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_mmxext; |
407 |
|
|
c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_mmxext; |
408 |
|
|
|
409 |
|
|
c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_mmxext; |
410 |
|
|
c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_mmxext; |
411 |
|
|
c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_mmxext; |
412 |
|
|
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_mmxext; |
413 |
|
|
#endif |
414 |
|
|
} |
415 |
|
|
|
416 |
✓✓ |
135 |
if (EXTERNAL_SSE(cpu_flags)) { |
417 |
|
68 |
c->vp8_idct_add = ff_vp8_idct_add_sse; |
418 |
|
68 |
c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse; |
419 |
|
|
} |
420 |
|
|
|
421 |
✓✓✗✓
✗✗ |
135 |
if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { |
422 |
|
62 |
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; |
423 |
|
|
|
424 |
|
62 |
c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2; |
425 |
|
62 |
c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_sse2; |
426 |
|
|
|
427 |
|
62 |
c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_sse2; |
428 |
|
62 |
c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_sse2; |
429 |
|
|
} |
430 |
|
|
|
431 |
✓✓ |
135 |
if (EXTERNAL_SSE2(cpu_flags)) { |
432 |
|
62 |
c->vp8_idct_dc_add = ff_vp8_idct_dc_add_sse2; |
433 |
|
62 |
c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_sse2; |
434 |
|
|
|
435 |
|
62 |
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2; |
436 |
|
|
|
437 |
|
62 |
c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_sse2; |
438 |
|
62 |
c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_sse2; |
439 |
|
|
|
440 |
|
62 |
c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_sse2; |
441 |
|
62 |
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_sse2; |
442 |
|
|
} |
443 |
|
|
|
444 |
✓✓ |
135 |
if (EXTERNAL_SSSE3(cpu_flags)) { |
445 |
|
50 |
c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_ssse3; |
446 |
|
50 |
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_ssse3; |
447 |
|
|
|
448 |
|
50 |
c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_ssse3; |
449 |
|
50 |
c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_ssse3; |
450 |
|
50 |
c->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_ssse3; |
451 |
|
50 |
c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_ssse3; |
452 |
|
|
|
453 |
|
50 |
c->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16y_mbedge_ssse3; |
454 |
|
50 |
c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_ssse3; |
455 |
|
50 |
c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_ssse3; |
456 |
|
50 |
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_ssse3; |
457 |
|
|
} |
458 |
|
|
|
459 |
✓✓ |
135 |
if (EXTERNAL_SSE4(cpu_flags)) { |
460 |
|
44 |
c->vp8_idct_dc_add = ff_vp8_idct_dc_add_sse4; |
461 |
|
|
|
462 |
|
44 |
c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse4; |
463 |
|
44 |
c->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16y_mbedge_sse4; |
464 |
|
44 |
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_sse4; |
465 |
|
|
} |
466 |
|
|
#endif /* HAVE_X86ASM */ |
467 |
|
135 |
} |