ARM: NEON scalarproduct_int16 and scalarproduct_and_madd_int16
[ffmpeg.git] / libavcodec / arm / dsputil_init_neon.c
1 /*
2  * ARM NEON optimised DSP functions
3  * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21
22 #include <stdint.h>
23
24 #include "libavcodec/avcodec.h"
25 #include "libavcodec/dsputil.h"
26 #include "dsputil_arm.h"
27
28 void ff_simple_idct_neon(DCTELEM *data);
29 void ff_simple_idct_put_neon(uint8_t *dest, int line_size, DCTELEM *data);
30 void ff_simple_idct_add_neon(uint8_t *dest, int line_size, DCTELEM *data);
31
32 void ff_vp3_idct_neon(DCTELEM *data);
33 void ff_vp3_idct_put_neon(uint8_t *dest, int line_size, DCTELEM *data);
34 void ff_vp3_idct_add_neon(uint8_t *dest, int line_size, DCTELEM *data);
35
36 void ff_put_pixels16_neon(uint8_t *, const uint8_t *, int, int);
37 void ff_put_pixels16_x2_neon(uint8_t *, const uint8_t *, int, int);
38 void ff_put_pixels16_y2_neon(uint8_t *, const uint8_t *, int, int);
39 void ff_put_pixels16_xy2_neon(uint8_t *, const uint8_t *, int, int);
40 void ff_put_pixels8_neon(uint8_t *, const uint8_t *, int, int);
41 void ff_put_pixels8_x2_neon(uint8_t *, const uint8_t *, int, int);
42 void ff_put_pixels8_y2_neon(uint8_t *, const uint8_t *, int, int);
43 void ff_put_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int);
44 void ff_put_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
45 void ff_put_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
46 void ff_put_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
47 void ff_put_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
48 void ff_put_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
49 void ff_put_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
50
51 void ff_avg_pixels16_neon(uint8_t *, const uint8_t *, int, int);
52 void ff_avg_pixels8_neon(uint8_t *, const uint8_t *, int, int);
53
54 void ff_add_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
55 void ff_put_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
56 void ff_put_signed_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
57
58 void ff_put_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int);
59 void ff_put_h264_qpel16_mc10_neon(uint8_t *, uint8_t *, int);
60 void ff_put_h264_qpel16_mc20_neon(uint8_t *, uint8_t *, int);
61 void ff_put_h264_qpel16_mc30_neon(uint8_t *, uint8_t *, int);
62 void ff_put_h264_qpel16_mc01_neon(uint8_t *, uint8_t *, int);
63 void ff_put_h264_qpel16_mc11_neon(uint8_t *, uint8_t *, int);
64 void ff_put_h264_qpel16_mc21_neon(uint8_t *, uint8_t *, int);
65 void ff_put_h264_qpel16_mc31_neon(uint8_t *, uint8_t *, int);
66 void ff_put_h264_qpel16_mc02_neon(uint8_t *, uint8_t *, int);
67 void ff_put_h264_qpel16_mc12_neon(uint8_t *, uint8_t *, int);
68 void ff_put_h264_qpel16_mc22_neon(uint8_t *, uint8_t *, int);
69 void ff_put_h264_qpel16_mc32_neon(uint8_t *, uint8_t *, int);
70 void ff_put_h264_qpel16_mc03_neon(uint8_t *, uint8_t *, int);
71 void ff_put_h264_qpel16_mc13_neon(uint8_t *, uint8_t *, int);
72 void ff_put_h264_qpel16_mc23_neon(uint8_t *, uint8_t *, int);
73 void ff_put_h264_qpel16_mc33_neon(uint8_t *, uint8_t *, int);
74
75 void ff_put_h264_qpel8_mc00_neon(uint8_t *, uint8_t *, int);
76 void ff_put_h264_qpel8_mc10_neon(uint8_t *, uint8_t *, int);
77 void ff_put_h264_qpel8_mc20_neon(uint8_t *, uint8_t *, int);
78 void ff_put_h264_qpel8_mc30_neon(uint8_t *, uint8_t *, int);
79 void ff_put_h264_qpel8_mc01_neon(uint8_t *, uint8_t *, int);
80 void ff_put_h264_qpel8_mc11_neon(uint8_t *, uint8_t *, int);
81 void ff_put_h264_qpel8_mc21_neon(uint8_t *, uint8_t *, int);
82 void ff_put_h264_qpel8_mc31_neon(uint8_t *, uint8_t *, int);
83 void ff_put_h264_qpel8_mc02_neon(uint8_t *, uint8_t *, int);
84 void ff_put_h264_qpel8_mc12_neon(uint8_t *, uint8_t *, int);
85 void ff_put_h264_qpel8_mc22_neon(uint8_t *, uint8_t *, int);
86 void ff_put_h264_qpel8_mc32_neon(uint8_t *, uint8_t *, int);
87 void ff_put_h264_qpel8_mc03_neon(uint8_t *, uint8_t *, int);
88 void ff_put_h264_qpel8_mc13_neon(uint8_t *, uint8_t *, int);
89 void ff_put_h264_qpel8_mc23_neon(uint8_t *, uint8_t *, int);
90 void ff_put_h264_qpel8_mc33_neon(uint8_t *, uint8_t *, int);
91
92 void ff_avg_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int);
93 void ff_avg_h264_qpel16_mc10_neon(uint8_t *, uint8_t *, int);
94 void ff_avg_h264_qpel16_mc20_neon(uint8_t *, uint8_t *, int);
95 void ff_avg_h264_qpel16_mc30_neon(uint8_t *, uint8_t *, int);
96 void ff_avg_h264_qpel16_mc01_neon(uint8_t *, uint8_t *, int);
97 void ff_avg_h264_qpel16_mc11_neon(uint8_t *, uint8_t *, int);
98 void ff_avg_h264_qpel16_mc21_neon(uint8_t *, uint8_t *, int);
99 void ff_avg_h264_qpel16_mc31_neon(uint8_t *, uint8_t *, int);
100 void ff_avg_h264_qpel16_mc02_neon(uint8_t *, uint8_t *, int);
101 void ff_avg_h264_qpel16_mc12_neon(uint8_t *, uint8_t *, int);
102 void ff_avg_h264_qpel16_mc22_neon(uint8_t *, uint8_t *, int);
103 void ff_avg_h264_qpel16_mc32_neon(uint8_t *, uint8_t *, int);
104 void ff_avg_h264_qpel16_mc03_neon(uint8_t *, uint8_t *, int);
105 void ff_avg_h264_qpel16_mc13_neon(uint8_t *, uint8_t *, int);
106 void ff_avg_h264_qpel16_mc23_neon(uint8_t *, uint8_t *, int);
107 void ff_avg_h264_qpel16_mc33_neon(uint8_t *, uint8_t *, int);
108
109 void ff_avg_h264_qpel8_mc00_neon(uint8_t *, uint8_t *, int);
110 void ff_avg_h264_qpel8_mc10_neon(uint8_t *, uint8_t *, int);
111 void ff_avg_h264_qpel8_mc20_neon(uint8_t *, uint8_t *, int);
112 void ff_avg_h264_qpel8_mc30_neon(uint8_t *, uint8_t *, int);
113 void ff_avg_h264_qpel8_mc01_neon(uint8_t *, uint8_t *, int);
114 void ff_avg_h264_qpel8_mc11_neon(uint8_t *, uint8_t *, int);
115 void ff_avg_h264_qpel8_mc21_neon(uint8_t *, uint8_t *, int);
116 void ff_avg_h264_qpel8_mc31_neon(uint8_t *, uint8_t *, int);
117 void ff_avg_h264_qpel8_mc02_neon(uint8_t *, uint8_t *, int);
118 void ff_avg_h264_qpel8_mc12_neon(uint8_t *, uint8_t *, int);
119 void ff_avg_h264_qpel8_mc22_neon(uint8_t *, uint8_t *, int);
120 void ff_avg_h264_qpel8_mc32_neon(uint8_t *, uint8_t *, int);
121 void ff_avg_h264_qpel8_mc03_neon(uint8_t *, uint8_t *, int);
122 void ff_avg_h264_qpel8_mc13_neon(uint8_t *, uint8_t *, int);
123 void ff_avg_h264_qpel8_mc23_neon(uint8_t *, uint8_t *, int);
124 void ff_avg_h264_qpel8_mc33_neon(uint8_t *, uint8_t *, int);
125
126 void ff_put_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
127 void ff_put_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
128 void ff_put_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
129
130 void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
131 void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
132 void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
133
134 void ff_h264_v_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
135                                      int beta, int8_t *tc0);
136 void ff_h264_h_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
137                                      int beta, int8_t *tc0);
138 void ff_h264_v_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
139                                        int beta, int8_t *tc0);
140 void ff_h264_h_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
141                                        int beta, int8_t *tc0);
142
143 void ff_weight_h264_pixels_16x16_neon(uint8_t *ds, int stride, int log2_den,
144                                       int weight, int offset);
145 void ff_weight_h264_pixels_16x8_neon(uint8_t *ds, int stride, int log2_den,
146                                      int weight, int offset);
147 void ff_weight_h264_pixels_8x16_neon(uint8_t *ds, int stride, int log2_den,
148                                      int weight, int offset);
149 void ff_weight_h264_pixels_8x8_neon(uint8_t *ds, int stride, int log2_den,
150                                     int weight, int offset);
151 void ff_weight_h264_pixels_8x4_neon(uint8_t *ds, int stride, int log2_den,
152                                     int weight, int offset);
153 void ff_weight_h264_pixels_4x8_neon(uint8_t *ds, int stride, int log2_den,
154                                     int weight, int offset);
155 void ff_weight_h264_pixels_4x4_neon(uint8_t *ds, int stride, int log2_den,
156                                     int weight, int offset);
157 void ff_weight_h264_pixels_4x2_neon(uint8_t *ds, int stride, int log2_den,
158                                     int weight, int offset);
159
160 void ff_biweight_h264_pixels_16x16_neon(uint8_t *dst, uint8_t *src, int stride,
161                                         int log2_den, int weightd, int weights,
162                                         int offset);
163 void ff_biweight_h264_pixels_16x8_neon(uint8_t *dst, uint8_t *src, int stride,
164                                        int log2_den, int weightd, int weights,
165                                        int offset);
166 void ff_biweight_h264_pixels_8x16_neon(uint8_t *dst, uint8_t *src, int stride,
167                                        int log2_den, int weightd, int weights,
168                                        int offset);
169 void ff_biweight_h264_pixels_8x8_neon(uint8_t *dst, uint8_t *src, int stride,
170                                       int log2_den, int weightd, int weights,
171                                       int offset);
172 void ff_biweight_h264_pixels_8x4_neon(uint8_t *dst, uint8_t *src, int stride,
173                                       int log2_den, int weightd, int weights,
174                                       int offset);
175 void ff_biweight_h264_pixels_4x8_neon(uint8_t *dst, uint8_t *src, int stride,
176                                       int log2_den, int weightd, int weights,
177                                       int offset);
178 void ff_biweight_h264_pixels_4x4_neon(uint8_t *dst, uint8_t *src, int stride,
179                                       int log2_den, int weightd, int weights,
180                                       int offset);
181 void ff_biweight_h264_pixels_4x2_neon(uint8_t *dst, uint8_t *src, int stride,
182                                       int log2_den, int weightd, int weights,
183                                       int offset);
184
185 void ff_h264_idct_add_neon(uint8_t *dst, DCTELEM *block, int stride);
186 void ff_h264_idct_dc_add_neon(uint8_t *dst, DCTELEM *block, int stride);
187 void ff_h264_idct_add16_neon(uint8_t *dst, const int *block_offset,
188                              DCTELEM *block, int stride,
189                              const uint8_t nnzc[6*8]);
190 void ff_h264_idct_add16intra_neon(uint8_t *dst, const int *block_offset,
191                                   DCTELEM *block, int stride,
192                                   const uint8_t nnzc[6*8]);
193 void ff_h264_idct_add8_neon(uint8_t **dest, const int *block_offset,
194                             DCTELEM *block, int stride,
195                             const uint8_t nnzc[6*8]);
196
197 void ff_vp3_v_loop_filter_neon(uint8_t *, int, int *);
198 void ff_vp3_h_loop_filter_neon(uint8_t *, int, int *);
199
200 void ff_vector_fmul_neon(float *dst, const float *src, int len);
201 void ff_vector_fmul_window_neon(float *dst, const float *src0,
202                                 const float *src1, const float *win,
203                                 float add_bias, int len);
204 void ff_vector_fmul_scalar_neon(float *dst, const float *src, float mul,
205                                 int len);
206 void ff_vector_fmul_sv_scalar_2_neon(float *dst, const float *src,
207                                      const float **vp, float mul, int len);
208 void ff_vector_fmul_sv_scalar_4_neon(float *dst, const float *src,
209                                      const float **vp, float mul, int len);
210 void ff_sv_fmul_scalar_2_neon(float *dst, const float **vp, float mul,
211                               int len);
212 void ff_sv_fmul_scalar_4_neon(float *dst, const float **vp, float mul,
213                               int len);
214 void ff_butterflies_float_neon(float *v1, float *v2, int len);
215 float ff_scalarproduct_float_neon(const float *v1, const float *v2, int len);
216 void ff_int32_to_float_fmul_scalar_neon(float *dst, const int *src,
217                                         float mul, int len);
218 void ff_vector_fmul_reverse_neon(float *dst, const float *src0,
219                                  const float *src1, int len);
220 void ff_vector_fmul_add_neon(float *dst, const float *src0, const float *src1,
221                              const float *src2, int len);
222
223 void ff_vector_clipf_neon(float *dst, const float *src, float min, float max,
224                           int len);
225 void ff_float_to_int16_neon(int16_t *, const float *, long);
226 void ff_float_to_int16_interleave_neon(int16_t *, const float **, long, int);
227
228 void ff_vorbis_inverse_coupling_neon(float *mag, float *ang, int blocksize);
229
230 int32_t ff_scalarproduct_int16_neon(int16_t *v1, int16_t *v2, int len,
231                                     int shift);
232 int32_t ff_scalarproduct_and_madd_int16_neon(int16_t *v1, int16_t *v2,
233                                              int16_t *v3, int len, int mul);
234
235 void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
236 {
237     if (!avctx->lowres) {
238         if (avctx->idct_algo == FF_IDCT_AUTO ||
239             avctx->idct_algo == FF_IDCT_SIMPLENEON) {
240             c->idct_put              = ff_simple_idct_put_neon;
241             c->idct_add              = ff_simple_idct_add_neon;
242             c->idct                  = ff_simple_idct_neon;
243             c->idct_permutation_type = FF_PARTTRANS_IDCT_PERM;
244         } else if ((CONFIG_VP3_DECODER || CONFIG_VP5_DECODER ||
245                     CONFIG_VP6_DECODER) &&
246                    avctx->idct_algo == FF_IDCT_VP3) {
247             c->idct_put              = ff_vp3_idct_put_neon;
248             c->idct_add              = ff_vp3_idct_add_neon;
249             c->idct                  = ff_vp3_idct_neon;
250             c->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
251         }
252     }
253
254     c->put_pixels_tab[0][0] = ff_put_pixels16_neon;
255     c->put_pixels_tab[0][1] = ff_put_pixels16_x2_neon;
256     c->put_pixels_tab[0][2] = ff_put_pixels16_y2_neon;
257     c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_neon;
258     c->put_pixels_tab[1][0] = ff_put_pixels8_neon;
259     c->put_pixels_tab[1][1] = ff_put_pixels8_x2_neon;
260     c->put_pixels_tab[1][2] = ff_put_pixels8_y2_neon;
261     c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_neon;
262
263     c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_neon;
264     c->put_no_rnd_pixels_tab[0][1] = ff_put_pixels16_x2_no_rnd_neon;
265     c->put_no_rnd_pixels_tab[0][2] = ff_put_pixels16_y2_no_rnd_neon;
266     c->put_no_rnd_pixels_tab[0][3] = ff_put_pixels16_xy2_no_rnd_neon;
267     c->put_no_rnd_pixels_tab[1][0] = ff_put_pixels8_neon;
268     c->put_no_rnd_pixels_tab[1][1] = ff_put_pixels8_x2_no_rnd_neon;
269     c->put_no_rnd_pixels_tab[1][2] = ff_put_pixels8_y2_no_rnd_neon;
270     c->put_no_rnd_pixels_tab[1][3] = ff_put_pixels8_xy2_no_rnd_neon;
271
272     c->avg_pixels_tab[0][0] = ff_avg_pixels16_neon;
273     c->avg_pixels_tab[1][0] = ff_avg_pixels8_neon;
274
275     c->add_pixels_clamped = ff_add_pixels_clamped_neon;
276     c->put_pixels_clamped = ff_put_pixels_clamped_neon;
277     c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_neon;
278
279     if (CONFIG_H264_DECODER) {
280         c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon;
281         c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon;
282         c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_neon;
283
284         c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_neon;
285         c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_neon;
286         c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_neon;
287
288         c->put_h264_qpel_pixels_tab[0][ 0] = ff_put_h264_qpel16_mc00_neon;
289         c->put_h264_qpel_pixels_tab[0][ 1] = ff_put_h264_qpel16_mc10_neon;
290         c->put_h264_qpel_pixels_tab[0][ 2] = ff_put_h264_qpel16_mc20_neon;
291         c->put_h264_qpel_pixels_tab[0][ 3] = ff_put_h264_qpel16_mc30_neon;
292         c->put_h264_qpel_pixels_tab[0][ 4] = ff_put_h264_qpel16_mc01_neon;
293         c->put_h264_qpel_pixels_tab[0][ 5] = ff_put_h264_qpel16_mc11_neon;
294         c->put_h264_qpel_pixels_tab[0][ 6] = ff_put_h264_qpel16_mc21_neon;
295         c->put_h264_qpel_pixels_tab[0][ 7] = ff_put_h264_qpel16_mc31_neon;
296         c->put_h264_qpel_pixels_tab[0][ 8] = ff_put_h264_qpel16_mc02_neon;
297         c->put_h264_qpel_pixels_tab[0][ 9] = ff_put_h264_qpel16_mc12_neon;
298         c->put_h264_qpel_pixels_tab[0][10] = ff_put_h264_qpel16_mc22_neon;
299         c->put_h264_qpel_pixels_tab[0][11] = ff_put_h264_qpel16_mc32_neon;
300         c->put_h264_qpel_pixels_tab[0][12] = ff_put_h264_qpel16_mc03_neon;
301         c->put_h264_qpel_pixels_tab[0][13] = ff_put_h264_qpel16_mc13_neon;
302         c->put_h264_qpel_pixels_tab[0][14] = ff_put_h264_qpel16_mc23_neon;
303         c->put_h264_qpel_pixels_tab[0][15] = ff_put_h264_qpel16_mc33_neon;
304
305         c->put_h264_qpel_pixels_tab[1][ 0] = ff_put_h264_qpel8_mc00_neon;
306         c->put_h264_qpel_pixels_tab[1][ 1] = ff_put_h264_qpel8_mc10_neon;
307         c->put_h264_qpel_pixels_tab[1][ 2] = ff_put_h264_qpel8_mc20_neon;
308         c->put_h264_qpel_pixels_tab[1][ 3] = ff_put_h264_qpel8_mc30_neon;
309         c->put_h264_qpel_pixels_tab[1][ 4] = ff_put_h264_qpel8_mc01_neon;
310         c->put_h264_qpel_pixels_tab[1][ 5] = ff_put_h264_qpel8_mc11_neon;
311         c->put_h264_qpel_pixels_tab[1][ 6] = ff_put_h264_qpel8_mc21_neon;
312         c->put_h264_qpel_pixels_tab[1][ 7] = ff_put_h264_qpel8_mc31_neon;
313         c->put_h264_qpel_pixels_tab[1][ 8] = ff_put_h264_qpel8_mc02_neon;
314         c->put_h264_qpel_pixels_tab[1][ 9] = ff_put_h264_qpel8_mc12_neon;
315         c->put_h264_qpel_pixels_tab[1][10] = ff_put_h264_qpel8_mc22_neon;
316         c->put_h264_qpel_pixels_tab[1][11] = ff_put_h264_qpel8_mc32_neon;
317         c->put_h264_qpel_pixels_tab[1][12] = ff_put_h264_qpel8_mc03_neon;
318         c->put_h264_qpel_pixels_tab[1][13] = ff_put_h264_qpel8_mc13_neon;
319         c->put_h264_qpel_pixels_tab[1][14] = ff_put_h264_qpel8_mc23_neon;
320         c->put_h264_qpel_pixels_tab[1][15] = ff_put_h264_qpel8_mc33_neon;
321
322         c->avg_h264_qpel_pixels_tab[0][ 0] = ff_avg_h264_qpel16_mc00_neon;
323         c->avg_h264_qpel_pixels_tab[0][ 1] = ff_avg_h264_qpel16_mc10_neon;
324         c->avg_h264_qpel_pixels_tab[0][ 2] = ff_avg_h264_qpel16_mc20_neon;
325         c->avg_h264_qpel_pixels_tab[0][ 3] = ff_avg_h264_qpel16_mc30_neon;
326         c->avg_h264_qpel_pixels_tab[0][ 4] = ff_avg_h264_qpel16_mc01_neon;
327         c->avg_h264_qpel_pixels_tab[0][ 5] = ff_avg_h264_qpel16_mc11_neon;
328         c->avg_h264_qpel_pixels_tab[0][ 6] = ff_avg_h264_qpel16_mc21_neon;
329         c->avg_h264_qpel_pixels_tab[0][ 7] = ff_avg_h264_qpel16_mc31_neon;
330         c->avg_h264_qpel_pixels_tab[0][ 8] = ff_avg_h264_qpel16_mc02_neon;
331         c->avg_h264_qpel_pixels_tab[0][ 9] = ff_avg_h264_qpel16_mc12_neon;
332         c->avg_h264_qpel_pixels_tab[0][10] = ff_avg_h264_qpel16_mc22_neon;
333         c->avg_h264_qpel_pixels_tab[0][11] = ff_avg_h264_qpel16_mc32_neon;
334         c->avg_h264_qpel_pixels_tab[0][12] = ff_avg_h264_qpel16_mc03_neon;
335         c->avg_h264_qpel_pixels_tab[0][13] = ff_avg_h264_qpel16_mc13_neon;
336         c->avg_h264_qpel_pixels_tab[0][14] = ff_avg_h264_qpel16_mc23_neon;
337         c->avg_h264_qpel_pixels_tab[0][15] = ff_avg_h264_qpel16_mc33_neon;
338
339         c->avg_h264_qpel_pixels_tab[1][ 0] = ff_avg_h264_qpel8_mc00_neon;
340         c->avg_h264_qpel_pixels_tab[1][ 1] = ff_avg_h264_qpel8_mc10_neon;
341         c->avg_h264_qpel_pixels_tab[1][ 2] = ff_avg_h264_qpel8_mc20_neon;
342         c->avg_h264_qpel_pixels_tab[1][ 3] = ff_avg_h264_qpel8_mc30_neon;
343         c->avg_h264_qpel_pixels_tab[1][ 4] = ff_avg_h264_qpel8_mc01_neon;
344         c->avg_h264_qpel_pixels_tab[1][ 5] = ff_avg_h264_qpel8_mc11_neon;
345         c->avg_h264_qpel_pixels_tab[1][ 6] = ff_avg_h264_qpel8_mc21_neon;
346         c->avg_h264_qpel_pixels_tab[1][ 7] = ff_avg_h264_qpel8_mc31_neon;
347         c->avg_h264_qpel_pixels_tab[1][ 8] = ff_avg_h264_qpel8_mc02_neon;
348         c->avg_h264_qpel_pixels_tab[1][ 9] = ff_avg_h264_qpel8_mc12_neon;
349         c->avg_h264_qpel_pixels_tab[1][10] = ff_avg_h264_qpel8_mc22_neon;
350         c->avg_h264_qpel_pixels_tab[1][11] = ff_avg_h264_qpel8_mc32_neon;
351         c->avg_h264_qpel_pixels_tab[1][12] = ff_avg_h264_qpel8_mc03_neon;
352         c->avg_h264_qpel_pixels_tab[1][13] = ff_avg_h264_qpel8_mc13_neon;
353         c->avg_h264_qpel_pixels_tab[1][14] = ff_avg_h264_qpel8_mc23_neon;
354         c->avg_h264_qpel_pixels_tab[1][15] = ff_avg_h264_qpel8_mc33_neon;
355
356         c->h264_v_loop_filter_luma   = ff_h264_v_loop_filter_luma_neon;
357         c->h264_h_loop_filter_luma   = ff_h264_h_loop_filter_luma_neon;
358         c->h264_v_loop_filter_chroma = ff_h264_v_loop_filter_chroma_neon;
359         c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma_neon;
360
361         c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels_16x16_neon;
362         c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels_16x8_neon;
363         c->weight_h264_pixels_tab[2] = ff_weight_h264_pixels_8x16_neon;
364         c->weight_h264_pixels_tab[3] = ff_weight_h264_pixels_8x8_neon;
365         c->weight_h264_pixels_tab[4] = ff_weight_h264_pixels_8x4_neon;
366         c->weight_h264_pixels_tab[5] = ff_weight_h264_pixels_4x8_neon;
367         c->weight_h264_pixels_tab[6] = ff_weight_h264_pixels_4x4_neon;
368         c->weight_h264_pixels_tab[7] = ff_weight_h264_pixels_4x2_neon;
369
370         c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels_16x16_neon;
371         c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels_16x8_neon;
372         c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels_8x16_neon;
373         c->biweight_h264_pixels_tab[3] = ff_biweight_h264_pixels_8x8_neon;
374         c->biweight_h264_pixels_tab[4] = ff_biweight_h264_pixels_8x4_neon;
375         c->biweight_h264_pixels_tab[5] = ff_biweight_h264_pixels_4x8_neon;
376         c->biweight_h264_pixels_tab[6] = ff_biweight_h264_pixels_4x4_neon;
377         c->biweight_h264_pixels_tab[7] = ff_biweight_h264_pixels_4x2_neon;
378
379         c->h264_idct_add        = ff_h264_idct_add_neon;
380         c->h264_idct_dc_add     = ff_h264_idct_dc_add_neon;
381         c->h264_idct_add16      = ff_h264_idct_add16_neon;
382         c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
383         c->h264_idct_add8       = ff_h264_idct_add8_neon;
384     }
385
386     if (CONFIG_VP3_DECODER) {
387         c->vp3_v_loop_filter = ff_vp3_v_loop_filter_neon;
388         c->vp3_h_loop_filter = ff_vp3_h_loop_filter_neon;
389     }
390
391     c->vector_fmul                = ff_vector_fmul_neon;
392     c->vector_fmul_window         = ff_vector_fmul_window_neon;
393     c->vector_fmul_scalar         = ff_vector_fmul_scalar_neon;
394     c->butterflies_float          = ff_butterflies_float_neon;
395     c->scalarproduct_float        = ff_scalarproduct_float_neon;
396     c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_neon;
397     c->vector_fmul_reverse        = ff_vector_fmul_reverse_neon;
398     c->vector_fmul_add            = ff_vector_fmul_add_neon;
399     c->vector_clipf               = ff_vector_clipf_neon;
400
401     c->vector_fmul_sv_scalar[0] = ff_vector_fmul_sv_scalar_2_neon;
402     c->vector_fmul_sv_scalar[1] = ff_vector_fmul_sv_scalar_4_neon;
403
404     c->sv_fmul_scalar[0] = ff_sv_fmul_scalar_2_neon;
405     c->sv_fmul_scalar[1] = ff_sv_fmul_scalar_4_neon;
406
407     if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
408         c->float_to_int16            = ff_float_to_int16_neon;
409         c->float_to_int16_interleave = ff_float_to_int16_interleave_neon;
410     }
411
412     if (CONFIG_VORBIS_DECODER)
413         c->vorbis_inverse_coupling = ff_vorbis_inverse_coupling_neon;
414
415     c->scalarproduct_int16 = ff_scalarproduct_int16_neon;
416     c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_neon;
417 }