11975
|
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/x86_cpu.h"
|
|
24 #include "libavcodec/vp8dsp.h"
|
|
25
|
11991
|
26 #if HAVE_YASM
|
|
27
|
11975
|
28 /*
|
|
29 * MC functions
|
|
30 */
|
|
31 extern void ff_put_vp8_epel4_h4_mmxext(uint8_t *dst, int dststride,
|
|
32 uint8_t *src, int srcstride,
|
|
33 int height, int mx, int my);
|
|
34 extern void ff_put_vp8_epel4_h6_mmxext(uint8_t *dst, int dststride,
|
|
35 uint8_t *src, int srcstride,
|
|
36 int height, int mx, int my);
|
|
37 extern void ff_put_vp8_epel4_v4_mmxext(uint8_t *dst, int dststride,
|
|
38 uint8_t *src, int srcstride,
|
|
39 int height, int mx, int my);
|
|
40 extern void ff_put_vp8_epel4_v6_mmxext(uint8_t *dst, int dststride,
|
|
41 uint8_t *src, int srcstride,
|
|
42 int height, int mx, int my);
|
|
43
|
|
44 extern void ff_put_vp8_epel8_h4_sse2 (uint8_t *dst, int dststride,
|
|
45 uint8_t *src, int srcstride,
|
|
46 int height, int mx, int my);
|
|
47 extern void ff_put_vp8_epel8_h6_sse2 (uint8_t *dst, int dststride,
|
|
48 uint8_t *src, int srcstride,
|
|
49 int height, int mx, int my);
|
|
50 extern void ff_put_vp8_epel8_v4_sse2 (uint8_t *dst, int dststride,
|
|
51 uint8_t *src, int srcstride,
|
|
52 int height, int mx, int my);
|
|
53 extern void ff_put_vp8_epel8_v6_sse2 (uint8_t *dst, int dststride,
|
|
54 uint8_t *src, int srcstride,
|
|
55 int height, int mx, int my);
|
|
56
|
|
57 extern void ff_put_vp8_epel8_h4_ssse3 (uint8_t *dst, int dststride,
|
|
58 uint8_t *src, int srcstride,
|
|
59 int height, int mx, int my);
|
|
60 extern void ff_put_vp8_epel8_h6_ssse3 (uint8_t *dst, int dststride,
|
|
61 uint8_t *src, int srcstride,
|
|
62 int height, int mx, int my);
|
|
63 extern void ff_put_vp8_epel8_v4_ssse3 (uint8_t *dst, int dststride,
|
|
64 uint8_t *src, int srcstride,
|
|
65 int height, int mx, int my);
|
|
66 extern void ff_put_vp8_epel8_v6_ssse3 (uint8_t *dst, int dststride,
|
|
67 uint8_t *src, int srcstride,
|
|
68 int height, int mx, int my);
|
|
69
|
11991
|
70 extern void ff_put_vp8_bilinear4_h_mmxext(uint8_t *dst, int dststride,
|
|
71 uint8_t *src, int srcstride,
|
|
72 int height, int mx, int my);
|
|
73 extern void ff_put_vp8_bilinear8_h_sse2 (uint8_t *dst, int dststride,
|
|
74 uint8_t *src, int srcstride,
|
|
75 int height, int mx, int my);
|
|
76
|
|
77 extern void ff_put_vp8_bilinear4_v_mmxext(uint8_t *dst, int dststride,
|
|
78 uint8_t *src, int srcstride,
|
|
79 int height, int mx, int my);
|
|
80 extern void ff_put_vp8_bilinear8_v_sse2 (uint8_t *dst, int dststride,
|
|
81 uint8_t *src, int srcstride,
|
|
82 int height, int mx, int my);
|
|
83 extern void ff_put_vp8_bilinear8_v_ssse3 (uint8_t *dst, int dststride,
|
|
84 uint8_t *src, int srcstride,
|
|
85 int height, int mx, int my);
|
|
86 extern void ff_put_vp8_bilinear8_h_ssse3 (uint8_t *dst, int dststride,
|
|
87 uint8_t *src, int srcstride,
|
|
88 int height, int mx, int my);
|
|
89
|
|
90 #define TAP_W16(OPT, FILTERTYPE, TAPTYPE) \
|
|
91 static void ff_put_vp8_ ## FILTERTYPE ## 16_ ## TAPTYPE ## _ ## OPT( \
|
|
92 uint8_t *dst, int dststride, uint8_t *src, \
|
|
93 int srcstride, int height, int mx, int my) \
|
11975
|
94 { \
|
11991
|
95 ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
|
|
96 dst, dststride, src, srcstride, height, mx, my); \
|
|
97 ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
|
|
98 dst + 8, dststride, src + 8, srcstride, height, mx, my); \
|
11975
|
99 }
|
11991
|
100 #define TAP_W8(OPT, FILTERTYPE, TAPTYPE) \
|
|
101 static void ff_put_vp8_ ## FILTERTYPE ## 8_ ## TAPTYPE ## _ ## OPT( \
|
|
102 uint8_t *dst, int dststride, uint8_t *src, \
|
|
103 int srcstride, int height, int mx, int my) \
|
11975
|
104 { \
|
11991
|
105 ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
|
|
106 dst, dststride, src, srcstride, height, mx, my); \
|
|
107 ff_put_vp8_ ## FILTERTYPE ## 4_ ## TAPTYPE ## _ ## OPT( \
|
|
108 dst + 4, dststride, src + 4, srcstride, height, mx, my); \
|
11975
|
109 }
|
|
110
|
11991
|
111 TAP_W8 (mmxext, epel, h4)
|
|
112 TAP_W8 (mmxext, epel, h6)
|
|
113 TAP_W16(mmxext, epel, h6)
|
|
114 TAP_W8 (mmxext, epel, v4)
|
|
115 TAP_W8 (mmxext, epel, v6)
|
|
116 TAP_W16(mmxext, epel, v6)
|
|
117 TAP_W8 (mmxext, bilinear, h)
|
|
118 TAP_W16(mmxext, bilinear, h)
|
|
119 TAP_W8 (mmxext, bilinear, v)
|
|
120 TAP_W16(mmxext, bilinear, v)
|
11975
|
121
|
11991
|
122 TAP_W16(sse2, epel, h6)
|
|
123 TAP_W16(sse2, epel, v6)
|
|
124 TAP_W16(sse2, bilinear, h)
|
|
125 TAP_W16(sse2, bilinear, v)
|
11975
|
126
|
11991
|
127 TAP_W16(ssse3, epel, h6)
|
|
128 TAP_W16(ssse3, epel, v6)
|
|
129 TAP_W16(ssse3, bilinear, h)
|
|
130 TAP_W16(ssse3, bilinear, v)
|
11975
|
131
|
|
132 #define HVTAP(OPT, ALIGN, TAPNUMX, TAPNUMY, SIZE, MAXHEIGHT) \
|
11991
|
133 static void ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## v ## TAPNUMY ## _ ## OPT( \
|
|
134 uint8_t *dst, int dststride, uint8_t *src, \
|
|
135 int srcstride, int height, int mx, int my) \
|
11975
|
136 { \
|
|
137 DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + TAPNUMY - 1)]; \
|
|
138 uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \
|
|
139 src -= srcstride * (TAPNUMY / 2 - 1); \
|
11991
|
140 ff_put_vp8_epel ## SIZE ## _h ## TAPNUMX ## _ ## OPT( \
|
|
141 tmp, SIZE, src, srcstride, height + TAPNUMY - 1, mx, my); \
|
|
142 ff_put_vp8_epel ## SIZE ## _v ## TAPNUMY ## _ ## OPT( \
|
|
143 dst, dststride, tmpptr, SIZE, height, mx, my); \
|
11975
|
144 }
|
|
145
|
|
146 #define HVTAPMMX(x, y) \
|
|
147 HVTAP(mmxext, 8, x, y, 4, 8) \
|
|
148 HVTAP(mmxext, 8, x, y, 8, 16)
|
|
149
|
|
150 HVTAPMMX(4, 4)
|
|
151 HVTAPMMX(4, 6)
|
|
152 HVTAPMMX(6, 4)
|
|
153 HVTAPMMX(6, 6)
|
|
154 HVTAP(mmxext, 8, 6, 6, 16, 16)
|
|
155
|
|
156 #define HVTAPSSE2(x, y, w) \
|
|
157 HVTAP(sse2, 16, x, y, w, 16) \
|
|
158 HVTAP(ssse3, 16, x, y, w, 16)
|
|
159
|
|
160 HVTAPSSE2(4, 4, 8)
|
|
161 HVTAPSSE2(4, 6, 8)
|
|
162 HVTAPSSE2(6, 4, 8)
|
|
163 HVTAPSSE2(6, 6, 8)
|
|
164 HVTAPSSE2(6, 6, 16)
|
11991
|
165
|
|
166 #define HVBILIN(OPT, ALIGN, SIZE, MAXHEIGHT) \
|
|
167 static void ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT( \
|
|
168 uint8_t *dst, int dststride, uint8_t *src, \
|
|
169 int srcstride, int height, int mx, int my) \
|
|
170 { \
|
|
171 DECLARE_ALIGNED(ALIGN, uint8_t, tmp)[SIZE * (MAXHEIGHT + 2)]; \
|
|
172 ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT( \
|
|
173 tmp, SIZE, src, srcstride, height + 1, mx, my); \
|
|
174 ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT( \
|
|
175 dst, dststride, tmp, SIZE, height, mx, my); \
|
|
176 }
|
|
177
|
|
178 HVBILIN(mmxext, 8, 4, 8)
|
|
179 HVBILIN(mmxext, 8, 8, 16)
|
|
180 HVBILIN(mmxext, 8, 16, 16)
|
|
181 HVBILIN(sse2, 8, 8, 16)
|
|
182 HVBILIN(sse2, 8, 16, 16)
|
|
183 HVBILIN(ssse3, 8, 8, 16)
|
|
184 HVBILIN(ssse3, 8, 16, 16)
|
11975
|
185
|
|
186 extern void ff_vp8_idct_dc_add_mmx(uint8_t *dst, DCTELEM block[16], int stride);
|
|
187 extern void ff_vp8_idct_dc_add_sse4(uint8_t *dst, DCTELEM block[16], int stride);
|
11991
|
188 #endif
|
|
189
|
|
190 #define VP8_LUMA_MC_FUNC(IDX, SIZE, OPT) \
|
|
191 c->put_vp8_epel_pixels_tab[IDX][0][2] = ff_put_vp8_epel ## SIZE ## _h6_ ## OPT; \
|
|
192 c->put_vp8_epel_pixels_tab[IDX][2][0] = ff_put_vp8_epel ## SIZE ## _v6_ ## OPT; \
|
|
193 c->put_vp8_epel_pixels_tab[IDX][2][2] = ff_put_vp8_epel ## SIZE ## _h6v6_ ## OPT
|
|
194
|
|
195 #define VP8_MC_FUNC(IDX, SIZE, OPT) \
|
|
196 c->put_vp8_epel_pixels_tab[IDX][0][1] = ff_put_vp8_epel ## SIZE ## _h4_ ## OPT; \
|
|
197 c->put_vp8_epel_pixels_tab[IDX][1][0] = ff_put_vp8_epel ## SIZE ## _v4_ ## OPT; \
|
|
198 c->put_vp8_epel_pixels_tab[IDX][1][1] = ff_put_vp8_epel ## SIZE ## _h4v4_ ## OPT; \
|
|
199 c->put_vp8_epel_pixels_tab[IDX][1][2] = ff_put_vp8_epel ## SIZE ## _h6v4_ ## OPT; \
|
|
200 c->put_vp8_epel_pixels_tab[IDX][2][1] = ff_put_vp8_epel ## SIZE ## _h4v6_ ## OPT; \
|
|
201 VP8_LUMA_MC_FUNC(IDX, SIZE, OPT)
|
|
202
|
|
203 #define VP8_BILINEAR_MC_FUNC(IDX, SIZE, OPT) \
|
|
204 c->put_vp8_bilinear_pixels_tab[IDX][0][1] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
|
|
205 c->put_vp8_bilinear_pixels_tab[IDX][0][2] = ff_put_vp8_bilinear ## SIZE ## _h_ ## OPT; \
|
|
206 c->put_vp8_bilinear_pixels_tab[IDX][1][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
|
|
207 c->put_vp8_bilinear_pixels_tab[IDX][1][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
|
|
208 c->put_vp8_bilinear_pixels_tab[IDX][1][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
|
|
209 c->put_vp8_bilinear_pixels_tab[IDX][2][0] = ff_put_vp8_bilinear ## SIZE ## _v_ ## OPT; \
|
|
210 c->put_vp8_bilinear_pixels_tab[IDX][2][1] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT; \
|
|
211 c->put_vp8_bilinear_pixels_tab[IDX][2][2] = ff_put_vp8_bilinear ## SIZE ## _hv_ ## OPT
|
|
212
|
11975
|
213
|
|
214 av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
|
|
215 {
|
|
216 mm_flags = mm_support();
|
|
217
|
11976
|
218 #if HAVE_YASM
|
11975
|
219 if (mm_flags & FF_MM_MMX) {
|
|
220 c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx;
|
|
221 }
|
|
222
|
|
223 /* note that 4-tap width=16 functions are missing because w=16
|
|
224 * is only used for luma, and luma is always a copy or sixtap. */
|
|
225 if (mm_flags & FF_MM_MMXEXT) {
|
11991
|
226 VP8_LUMA_MC_FUNC(0, 16, mmxext);
|
|
227 VP8_MC_FUNC(1, 8, mmxext);
|
|
228 VP8_MC_FUNC(1, 4, mmxext);
|
|
229 VP8_BILINEAR_MC_FUNC(0, 16, mmxext);
|
|
230 VP8_BILINEAR_MC_FUNC(1, 8, mmxext);
|
|
231 VP8_BILINEAR_MC_FUNC(1, 4, mmxext);
|
11975
|
232 }
|
|
233
|
|
234 if (mm_flags & FF_MM_SSE2) {
|
11991
|
235 VP8_LUMA_MC_FUNC(0, 16, sse2);
|
|
236 VP8_MC_FUNC(1, 8, sse2);
|
|
237 VP8_BILINEAR_MC_FUNC(0, 16, sse2);
|
|
238 VP8_BILINEAR_MC_FUNC(1, 8, sse2);
|
11975
|
239 }
|
|
240
|
|
241 if (mm_flags & FF_MM_SSSE3) {
|
11991
|
242 VP8_LUMA_MC_FUNC(0, 16, ssse3);
|
|
243 VP8_MC_FUNC(1, 8, ssse3);
|
|
244 VP8_BILINEAR_MC_FUNC(0, 16, ssse3);
|
|
245 VP8_BILINEAR_MC_FUNC(1, 8, ssse3);
|
11975
|
246 }
|
|
247
|
|
248 if (mm_flags & FF_MM_SSE4) {
|
|
249 c->vp8_idct_dc_add = ff_vp8_idct_dc_add_sse4;
|
|
250 }
|
11976
|
251 #endif
|
11975
|
252 }
|