407dfd2cae0102ee26a5f6bad5180d9909361787
[ffmpeg.git] / libavcodec / arm / vp8dsp_init_arm.c
1 /*
2  * This file is part of FFmpeg.
3  *
4  * FFmpeg is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU Lesser General Public
6  * License as published by the Free Software Foundation; either
7  * version 2.1 of the License, or (at your option) any later version.
8  *
9  * FFmpeg is distributed in the hope that it will be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12  * Lesser General Public License for more details.
13  *
14  * You should have received a copy of the GNU Lesser General Public
15  * License along with FFmpeg; if not, write to the Free Software
16  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17  */
18
19 #include <stdint.h>
20 #include "libavcodec/vp8dsp.h"
21
22 void ff_vp8_luma_dc_wht_dc_armv6(DCTELEM block[4][4][16], DCTELEM dc[16]);
23
24 #define idct_funcs(opt) \
25 void ff_vp8_luma_dc_wht_ ## opt(DCTELEM block[4][4][16], DCTELEM dc[16]); \
26 void ff_vp8_idct_add_ ## opt(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride); \
27 void ff_vp8_idct_dc_add_ ## opt(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride); \
28 void ff_vp8_idct_dc_add4y_ ## opt(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride); \
29 void ff_vp8_idct_dc_add4uv_ ## opt(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride)
30
31 idct_funcs(neon);
32 idct_funcs(armv6);
33
34 void ff_vp8_v_loop_filter16_neon(uint8_t *dst, ptrdiff_t stride,
35                                  int flim_E, int flim_I, int hev_thresh);
36 void ff_vp8_h_loop_filter16_neon(uint8_t *dst, ptrdiff_t stride,
37                                  int flim_E, int flim_I, int hev_thresh);
38 void ff_vp8_v_loop_filter8uv_neon(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
39                                   int flim_E, int flim_I, int hev_thresh);
40 void ff_vp8_h_loop_filter8uv_neon(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
41                                   int flim_E, int flim_I, int hev_thresh);
42
43 void ff_vp8_v_loop_filter16_inner_neon(uint8_t *dst, ptrdiff_t stride,
44                                        int flim_E, int flim_I, int hev_thresh);
45 void ff_vp8_h_loop_filter16_inner_neon(uint8_t *dst, ptrdiff_t stride,
46                                        int flim_E, int flim_I, int hev_thresh);
47 void ff_vp8_v_loop_filter8uv_inner_neon(uint8_t *dstU, uint8_t *dstV,
48                                         ptrdiff_t stride, int flim_E, int flim_I,
49                                         int hev_thresh);
50 void ff_vp8_h_loop_filter8uv_inner_neon(uint8_t *dstU, uint8_t *dstV,
51                                         ptrdiff_t stride, int flim_E, int flim_I,
52                                         int hev_thresh);
53
54 void ff_vp8_v_loop_filter_inner_armv6(uint8_t *dst, ptrdiff_t stride,
55                                       int flim_E, int flim_I,
56                                       int hev_thresh, int count);
57 void ff_vp8_h_loop_filter_inner_armv6(uint8_t *dst, ptrdiff_t stride,
58                                       int flim_E, int flim_I,
59                                       int hev_thresh, int count);
60 void ff_vp8_v_loop_filter_armv6(uint8_t *dst, ptrdiff_t stride,
61                                 int flim_E, int flim_I,
62                                 int hev_thresh, int count);
63 void ff_vp8_h_loop_filter_armv6(uint8_t *dst, ptrdiff_t stride,
64                                 int flim_E, int flim_I,
65                                 int hev_thresh, int count);
66
67 static void ff_vp8_v_loop_filter16_armv6(uint8_t *dst, ptrdiff_t stride,
68                                          int flim_E, int flim_I, int hev_thresh)
69 {
70     ff_vp8_v_loop_filter_armv6(dst, stride, flim_E, flim_I, hev_thresh, 4);
71 }
72
73 static void ff_vp8_h_loop_filter16_armv6(uint8_t *dst, ptrdiff_t stride,
74                                          int flim_E, int flim_I, int hev_thresh)
75 {
76     ff_vp8_h_loop_filter_armv6(dst, stride, flim_E, flim_I, hev_thresh, 4);
77 }
78
79 static void ff_vp8_v_loop_filter8uv_armv6(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
80                                           int flim_E, int flim_I, int hev_thresh)
81 {
82     ff_vp8_v_loop_filter_armv6(dstU, stride, flim_E, flim_I, hev_thresh, 2);
83     ff_vp8_v_loop_filter_armv6(dstV, stride, flim_E, flim_I, hev_thresh, 2);
84 }
85
86 static void ff_vp8_h_loop_filter8uv_armv6(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
87                                           int flim_E, int flim_I, int hev_thresh)
88 {
89     ff_vp8_h_loop_filter_armv6(dstU, stride, flim_E, flim_I, hev_thresh, 2);
90     ff_vp8_h_loop_filter_armv6(dstV, stride, flim_E, flim_I, hev_thresh, 2);
91 }
92
93 static void ff_vp8_v_loop_filter16_inner_armv6(uint8_t *dst, ptrdiff_t stride,
94                                                int flim_E, int flim_I, int hev_thresh)
95 {
96     ff_vp8_v_loop_filter_inner_armv6(dst, stride, flim_E, flim_I, hev_thresh, 4);
97 }
98
99 static void ff_vp8_h_loop_filter16_inner_armv6(uint8_t *dst, ptrdiff_t stride,
100                                                int flim_E, int flim_I, int hev_thresh)
101 {
102     ff_vp8_h_loop_filter_inner_armv6(dst, stride, flim_E, flim_I, hev_thresh, 4);
103 }
104
105 static void ff_vp8_v_loop_filter8uv_inner_armv6(uint8_t *dstU, uint8_t *dstV,
106                                                 ptrdiff_t stride, int flim_E, int flim_I,
107                                                 int hev_thresh)
108 {
109     ff_vp8_v_loop_filter_inner_armv6(dstU, stride, flim_E, flim_I, hev_thresh, 2);
110     ff_vp8_v_loop_filter_inner_armv6(dstV, stride, flim_E, flim_I, hev_thresh, 2);
111 }
112
113 static void ff_vp8_h_loop_filter8uv_inner_armv6(uint8_t *dstU, uint8_t *dstV,
114                                                 ptrdiff_t stride, int flim_E, int flim_I,
115                                                 int hev_thresh)
116 {
117     ff_vp8_h_loop_filter_inner_armv6(dstU, stride, flim_E, flim_I, hev_thresh, 2);
118     ff_vp8_h_loop_filter_inner_armv6(dstV, stride, flim_E, flim_I, hev_thresh, 2);
119 }
120
121 #define simple_lf_funcs(opt) \
122 void ff_vp8_v_loop_filter16_simple_ ## opt(uint8_t *dst, ptrdiff_t stride, int flim); \
123 void ff_vp8_h_loop_filter16_simple_ ## opt(uint8_t *dst, ptrdiff_t stride, int flim)
124
125 simple_lf_funcs(neon);
126 simple_lf_funcs(armv6);
127
128 #define VP8_MC_OPT(n, opt)                                               \
129     void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride,       \
130                                 uint8_t *src, ptrdiff_t srcstride,       \
131                                 int h, int x, int y)
132
133 #define VP8_MC(n) \
134     VP8_MC_OPT(n, neon)
135
136 #define VP8_EPEL(w)                             \
137     VP8_MC(epel ## w ## _h4);                   \
138     VP8_MC(epel ## w ## _h6);                   \
139     VP8_MC(epel ## w ## _h4v4);                 \
140     VP8_MC(epel ## w ## _h6v4);                 \
141     VP8_MC(epel ## w ## _v4);                   \
142     VP8_MC(epel ## w ## _v6);                   \
143     VP8_MC(epel ## w ## _h4v6);                 \
144     VP8_MC(epel ## w ## _h6v6)
145
146 VP8_EPEL(16);
147 VP8_MC(pixels16);
148 VP8_MC_OPT(pixels16, armv6);
149 VP8_EPEL(8);
150 VP8_MC(pixels8);
151 VP8_MC_OPT(pixels8,  armv6);
152 VP8_EPEL(4);
153 VP8_MC_OPT(pixels4,  armv6);
154
155 VP8_MC(bilin16_h);
156 VP8_MC(bilin16_v);
157 VP8_MC(bilin16_hv);
158 VP8_MC(bilin8_h);
159 VP8_MC(bilin8_v);
160 VP8_MC(bilin8_hv);
161 VP8_MC(bilin4_h);
162 VP8_MC(bilin4_v);
163 VP8_MC(bilin4_hv);
164
165 #define VP8_V6_MC(n) \
166 void ff_put_vp8_##n##_armv6(uint8_t *dst, int dststride, uint8_t *src, \
167                             int srcstride, int w, int h, int mxy)
168
169 VP8_V6_MC(epel_v6);
170 VP8_V6_MC(epel_h6);
171 VP8_V6_MC(epel_v4);
172 VP8_V6_MC(epel_h4);
173 VP8_V6_MC(bilin_v);
174 VP8_V6_MC(bilin_h);
175
176 #define VP8_EPEL_HV(SIZE, TAPNUMX, TAPNUMY, NAME, HNAME, VNAME, MAXHEIGHT) \
177 static void ff_put_vp8_##NAME##SIZE##_##HNAME##VNAME##_armv6( \
178                                         uint8_t *dst, int dststride, uint8_t *src, \
179                                         int srcstride, int h, int mx, int my) \
180 { \
181     DECLARE_ALIGNED(4, uint8_t, tmp)[SIZE * (MAXHEIGHT + TAPNUMY - 1)]; \
182     uint8_t *tmpptr = tmp + SIZE * (TAPNUMY / 2 - 1); \
183     src -= srcstride * (TAPNUMY / 2 - 1); \
184     ff_put_vp8_ ## NAME ## _ ## HNAME ## _armv6(tmp, SIZE,      src,    srcstride, \
185                                                 SIZE, h + TAPNUMY - 1,  mx); \
186     ff_put_vp8_ ## NAME ## _ ## VNAME ## _armv6(dst, dststride, tmpptr, SIZE, \
187                                                 SIZE, h,                my); \
188 }
189
190 VP8_EPEL_HV(16, 6, 6, epel,  h6, v6, 16);
191 VP8_EPEL_HV(16, 2, 2, bilin, h,  v,  16);
192 VP8_EPEL_HV(8,  6, 6, epel,  h6, v6, 16);
193 VP8_EPEL_HV(8,  4, 6, epel,  h4, v6, 16);
194 VP8_EPEL_HV(8,  6, 4, epel,  h6, v4, 16);
195 VP8_EPEL_HV(8,  4, 4, epel,  h4, v4, 16);
196 VP8_EPEL_HV(8,  2, 2, bilin, h,  v,  16);
197 VP8_EPEL_HV(4,  6, 6, epel,  h6, v6, 8);
198 VP8_EPEL_HV(4,  4, 6, epel,  h4, v6, 8);
199 VP8_EPEL_HV(4,  6, 4, epel,  h6, v4, 8);
200 VP8_EPEL_HV(4,  4, 4, epel,  h4, v4, 8);
201 VP8_EPEL_HV(4,  2, 2, bilin, h,  v,  8);
202
203 extern void put_vp8_epel4_v6_c(uint8_t *dst, int d, uint8_t *src, int s, int h, int mx, int my);
204 #undef printf
205 #define VP8_EPEL_H_OR_V(SIZE, NAME, HV) \
206 static void ff_put_vp8_##NAME##SIZE##_##HV##_armv6( \
207                                         uint8_t *dst, int dststride, uint8_t *src, \
208                                         int srcstride, int h, int mx, int my) \
209 { \
210     ff_put_vp8_## NAME ## _ ## HV ## _armv6(dst, dststride, src, srcstride, \
211                                             SIZE, h, mx | my); \
212 }
213
214 VP8_EPEL_H_OR_V(4,  epel,  h6);
215 VP8_EPEL_H_OR_V(4,  epel,  h4);
216 VP8_EPEL_H_OR_V(4,  epel,  v6);
217 VP8_EPEL_H_OR_V(4,  epel,  v4);
218 VP8_EPEL_H_OR_V(4,  bilin, v);
219 VP8_EPEL_H_OR_V(4,  bilin, h);
220 VP8_EPEL_H_OR_V(8,  epel,  h6);
221 VP8_EPEL_H_OR_V(8,  epel,  h4);
222 VP8_EPEL_H_OR_V(8,  epel,  v6);
223 VP8_EPEL_H_OR_V(8,  epel,  v4);
224 VP8_EPEL_H_OR_V(8,  bilin, v);
225 VP8_EPEL_H_OR_V(8,  bilin, h);
226 VP8_EPEL_H_OR_V(16, epel,  h6);
227 VP8_EPEL_H_OR_V(16, epel,  v6);
228 VP8_EPEL_H_OR_V(16, bilin, v);
229 VP8_EPEL_H_OR_V(16, bilin, h);
230
231 av_cold void ff_vp8dsp_init_arm(VP8DSPContext *dsp)
232 {
233 #define set_func_ptrs(opt) \
234         dsp->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_##opt; \
235         dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_armv6; \
236  \
237         dsp->vp8_idct_add       = ff_vp8_idct_add_##opt; \
238         dsp->vp8_idct_dc_add    = ff_vp8_idct_dc_add_##opt; \
239         dsp->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_##opt; \
240         dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_##opt; \
241  \
242         dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_##opt; \
243         dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_##opt; \
244         dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_##opt; \
245         dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_##opt; \
246  \
247         dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_##opt; \
248         dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_##opt; \
249         dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_##opt; \
250         dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_##opt; \
251  \
252         dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_##opt; \
253         dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_##opt; \
254  \
255         dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_##opt; \
256         dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_##opt; \
257         dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_##opt; \
258         dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_##opt; \
259  \
260         dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_##opt; \
261         dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_##opt; \
262         dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_##opt; \
263         dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_##opt; \
264         dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_##opt; \
265         dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_##opt; \
266         dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_##opt; \
267         dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_##opt; \
268         dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_##opt; \
269  \
270         dsp->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_armv6; \
271         dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_##opt; \
272         dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_##opt; \
273         dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_##opt; \
274         dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_##opt; \
275         dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_##opt; \
276         dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_##opt; \
277         dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_##opt; \
278         dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_##opt; \
279  \
280         dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_##opt; \
281         dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilin16_h_##opt; \
282         dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilin16_v_##opt; \
283         dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilin16_hv_##opt; \
284  \
285         dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_##opt; \
286         dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilin8_h_##opt; \
287         dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilin8_h_##opt; \
288         dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilin8_v_##opt; \
289         dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilin8_hv_##opt; \
290         dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilin8_hv_##opt; \
291         dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilin8_v_##opt; \
292         dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilin8_hv_##opt; \
293         dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilin8_hv_##opt; \
294  \
295         dsp->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_armv6; \
296         dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilin4_h_##opt; \
297         dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilin4_h_##opt; \
298         dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilin4_v_##opt; \
299         dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilin4_hv_##opt; \
300         dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilin4_hv_##opt; \
301         dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilin4_v_##opt; \
302         dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_##opt; \
303         dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_##opt
304     if (HAVE_NEON) {
305         set_func_ptrs(neon);
306     } else if (HAVE_ARMV6) {
307         set_func_ptrs(armv6);
308     }
309 }