ARM: NEON optimised H.264 biweighted prediction
[ffmpeg.git] / libavcodec / arm / dsputil_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
27 void ff_put_pixels16_neon(uint8_t *, const uint8_t *, int, int);
28 void ff_put_pixels16_x2_neon(uint8_t *, const uint8_t *, int, int);
29 void ff_put_pixels16_y2_neon(uint8_t *, const uint8_t *, int, int);
30 void ff_put_pixels16_xy2_neon(uint8_t *, const uint8_t *, int, int);
31 void ff_put_pixels8_neon(uint8_t *, const uint8_t *, int, int);
32 void ff_put_pixels8_x2_neon(uint8_t *, const uint8_t *, int, int);
33 void ff_put_pixels8_y2_neon(uint8_t *, const uint8_t *, int, int);
34 void ff_put_pixels8_xy2_neon(uint8_t *, const uint8_t *, int, int);
35 void ff_put_pixels16_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
36 void ff_put_pixels16_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
37 void ff_put_pixels16_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
38 void ff_put_pixels8_x2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
39 void ff_put_pixels8_y2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
40 void ff_put_pixels8_xy2_no_rnd_neon(uint8_t *, const uint8_t *, int, int);
41
42 void ff_avg_pixels16_neon(uint8_t *, const uint8_t *, int, int);
43
44 void ff_put_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int);
45 void ff_put_h264_qpel16_mc10_neon(uint8_t *, uint8_t *, int);
46 void ff_put_h264_qpel16_mc20_neon(uint8_t *, uint8_t *, int);
47 void ff_put_h264_qpel16_mc30_neon(uint8_t *, uint8_t *, int);
48 void ff_put_h264_qpel16_mc01_neon(uint8_t *, uint8_t *, int);
49 void ff_put_h264_qpel16_mc11_neon(uint8_t *, uint8_t *, int);
50 void ff_put_h264_qpel16_mc21_neon(uint8_t *, uint8_t *, int);
51 void ff_put_h264_qpel16_mc31_neon(uint8_t *, uint8_t *, int);
52 void ff_put_h264_qpel16_mc02_neon(uint8_t *, uint8_t *, int);
53 void ff_put_h264_qpel16_mc12_neon(uint8_t *, uint8_t *, int);
54 void ff_put_h264_qpel16_mc22_neon(uint8_t *, uint8_t *, int);
55 void ff_put_h264_qpel16_mc32_neon(uint8_t *, uint8_t *, int);
56 void ff_put_h264_qpel16_mc03_neon(uint8_t *, uint8_t *, int);
57 void ff_put_h264_qpel16_mc13_neon(uint8_t *, uint8_t *, int);
58 void ff_put_h264_qpel16_mc23_neon(uint8_t *, uint8_t *, int);
59 void ff_put_h264_qpel16_mc33_neon(uint8_t *, uint8_t *, int);
60
61 void ff_put_h264_qpel8_mc00_neon(uint8_t *, uint8_t *, int);
62 void ff_put_h264_qpel8_mc10_neon(uint8_t *, uint8_t *, int);
63 void ff_put_h264_qpel8_mc20_neon(uint8_t *, uint8_t *, int);
64 void ff_put_h264_qpel8_mc30_neon(uint8_t *, uint8_t *, int);
65 void ff_put_h264_qpel8_mc01_neon(uint8_t *, uint8_t *, int);
66 void ff_put_h264_qpel8_mc11_neon(uint8_t *, uint8_t *, int);
67 void ff_put_h264_qpel8_mc21_neon(uint8_t *, uint8_t *, int);
68 void ff_put_h264_qpel8_mc31_neon(uint8_t *, uint8_t *, int);
69 void ff_put_h264_qpel8_mc02_neon(uint8_t *, uint8_t *, int);
70 void ff_put_h264_qpel8_mc12_neon(uint8_t *, uint8_t *, int);
71 void ff_put_h264_qpel8_mc22_neon(uint8_t *, uint8_t *, int);
72 void ff_put_h264_qpel8_mc32_neon(uint8_t *, uint8_t *, int);
73 void ff_put_h264_qpel8_mc03_neon(uint8_t *, uint8_t *, int);
74 void ff_put_h264_qpel8_mc13_neon(uint8_t *, uint8_t *, int);
75 void ff_put_h264_qpel8_mc23_neon(uint8_t *, uint8_t *, int);
76 void ff_put_h264_qpel8_mc33_neon(uint8_t *, uint8_t *, int);
77
78 void ff_avg_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int);
79
80 void ff_put_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
81 void ff_put_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
82
83 void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
84 void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
85
86 void ff_h264_v_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
87                                      int beta, int8_t *tc0);
88 void ff_h264_h_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
89                                      int beta, int8_t *tc0);
90 void ff_h264_v_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
91                                        int beta, int8_t *tc0);
92 void ff_h264_h_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
93                                        int beta, int8_t *tc0);
94
95 void ff_biweight_h264_pixels_16x16_neon(uint8_t *dst, uint8_t *src, int stride,
96                                         int log2_den, int weightd, int weights,
97                                         int offset);
98 void ff_biweight_h264_pixels_16x8_neon(uint8_t *dst, uint8_t *src, int stride,
99                                        int log2_den, int weightd, int weights,
100                                        int offset);
101 void ff_biweight_h264_pixels_8x16_neon(uint8_t *dst, uint8_t *src, int stride,
102                                        int log2_den, int weightd, int weights,
103                                        int offset);
104 void ff_biweight_h264_pixels_8x8_neon(uint8_t *dst, uint8_t *src, int stride,
105                                       int log2_den, int weightd, int weights,
106                                       int offset);
107 void ff_biweight_h264_pixels_8x4_neon(uint8_t *dst, uint8_t *src, int stride,
108                                       int log2_den, int weightd, int weights,
109                                       int offset);
110 void ff_biweight_h264_pixels_4x8_neon(uint8_t *dst, uint8_t *src, int stride,
111                                       int log2_den, int weightd, int weights,
112                                       int offset);
113 void ff_biweight_h264_pixels_4x4_neon(uint8_t *dst, uint8_t *src, int stride,
114                                       int log2_den, int weightd, int weights,
115                                       int offset);
116 void ff_biweight_h264_pixels_4x2_neon(uint8_t *dst, uint8_t *src, int stride,
117                                       int log2_den, int weightd, int weights,
118                                       int offset);
119
120 void ff_h264_idct_add_neon(uint8_t *dst, DCTELEM *block, int stride);
121 void ff_h264_idct_dc_add_neon(uint8_t *dst, DCTELEM *block, int stride);
122 void ff_h264_idct_add16_neon(uint8_t *dst, const int *block_offset,
123                              DCTELEM *block, int stride,
124                              const uint8_t nnzc[6*8]);
125 void ff_h264_idct_add16intra_neon(uint8_t *dst, const int *block_offset,
126                                   DCTELEM *block, int stride,
127                                   const uint8_t nnzc[6*8]);
128 void ff_h264_idct_add8_neon(uint8_t **dest, const int *block_offset,
129                             DCTELEM *block, int stride,
130                             const uint8_t nnzc[6*8]);
131
132 void ff_float_to_int16_neon(int16_t *, const float *, long);
133 void ff_float_to_int16_interleave_neon(int16_t *, const float **, long, int);
134
135 void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
136 {
137     c->put_pixels_tab[0][0] = ff_put_pixels16_neon;
138     c->put_pixels_tab[0][1] = ff_put_pixels16_x2_neon;
139     c->put_pixels_tab[0][2] = ff_put_pixels16_y2_neon;
140     c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_neon;
141     c->put_pixels_tab[1][0] = ff_put_pixels8_neon;
142     c->put_pixels_tab[1][1] = ff_put_pixels8_x2_neon;
143     c->put_pixels_tab[1][2] = ff_put_pixels8_y2_neon;
144     c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_neon;
145
146     c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_neon;
147     c->put_no_rnd_pixels_tab[0][1] = ff_put_pixels16_x2_no_rnd_neon;
148     c->put_no_rnd_pixels_tab[0][2] = ff_put_pixels16_y2_no_rnd_neon;
149     c->put_no_rnd_pixels_tab[0][3] = ff_put_pixels16_xy2_no_rnd_neon;
150     c->put_no_rnd_pixels_tab[1][0] = ff_put_pixels8_neon;
151     c->put_no_rnd_pixels_tab[1][1] = ff_put_pixels8_x2_no_rnd_neon;
152     c->put_no_rnd_pixels_tab[1][2] = ff_put_pixels8_y2_no_rnd_neon;
153     c->put_no_rnd_pixels_tab[1][3] = ff_put_pixels8_xy2_no_rnd_neon;
154
155     c->avg_pixels_tab[0][0] = ff_avg_pixels16_neon;
156
157     c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon;
158     c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon;
159
160     c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_neon;
161     c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_neon;
162
163     c->put_h264_qpel_pixels_tab[0][ 0] = ff_put_h264_qpel16_mc00_neon;
164     c->put_h264_qpel_pixels_tab[0][ 1] = ff_put_h264_qpel16_mc10_neon;
165     c->put_h264_qpel_pixels_tab[0][ 2] = ff_put_h264_qpel16_mc20_neon;
166     c->put_h264_qpel_pixels_tab[0][ 3] = ff_put_h264_qpel16_mc30_neon;
167     c->put_h264_qpel_pixels_tab[0][ 4] = ff_put_h264_qpel16_mc01_neon;
168     c->put_h264_qpel_pixels_tab[0][ 5] = ff_put_h264_qpel16_mc11_neon;
169     c->put_h264_qpel_pixels_tab[0][ 6] = ff_put_h264_qpel16_mc21_neon;
170     c->put_h264_qpel_pixels_tab[0][ 7] = ff_put_h264_qpel16_mc31_neon;
171     c->put_h264_qpel_pixels_tab[0][ 8] = ff_put_h264_qpel16_mc02_neon;
172     c->put_h264_qpel_pixels_tab[0][ 9] = ff_put_h264_qpel16_mc12_neon;
173     c->put_h264_qpel_pixels_tab[0][10] = ff_put_h264_qpel16_mc22_neon;
174     c->put_h264_qpel_pixels_tab[0][11] = ff_put_h264_qpel16_mc32_neon;
175     c->put_h264_qpel_pixels_tab[0][12] = ff_put_h264_qpel16_mc03_neon;
176     c->put_h264_qpel_pixels_tab[0][13] = ff_put_h264_qpel16_mc13_neon;
177     c->put_h264_qpel_pixels_tab[0][14] = ff_put_h264_qpel16_mc23_neon;
178     c->put_h264_qpel_pixels_tab[0][15] = ff_put_h264_qpel16_mc33_neon;
179
180     c->put_h264_qpel_pixels_tab[1][ 0] = ff_put_h264_qpel8_mc00_neon;
181     c->put_h264_qpel_pixels_tab[1][ 1] = ff_put_h264_qpel8_mc10_neon;
182     c->put_h264_qpel_pixels_tab[1][ 2] = ff_put_h264_qpel8_mc20_neon;
183     c->put_h264_qpel_pixels_tab[1][ 3] = ff_put_h264_qpel8_mc30_neon;
184     c->put_h264_qpel_pixels_tab[1][ 4] = ff_put_h264_qpel8_mc01_neon;
185     c->put_h264_qpel_pixels_tab[1][ 5] = ff_put_h264_qpel8_mc11_neon;
186     c->put_h264_qpel_pixels_tab[1][ 6] = ff_put_h264_qpel8_mc21_neon;
187     c->put_h264_qpel_pixels_tab[1][ 7] = ff_put_h264_qpel8_mc31_neon;
188     c->put_h264_qpel_pixels_tab[1][ 8] = ff_put_h264_qpel8_mc02_neon;
189     c->put_h264_qpel_pixels_tab[1][ 9] = ff_put_h264_qpel8_mc12_neon;
190     c->put_h264_qpel_pixels_tab[1][10] = ff_put_h264_qpel8_mc22_neon;
191     c->put_h264_qpel_pixels_tab[1][11] = ff_put_h264_qpel8_mc32_neon;
192     c->put_h264_qpel_pixels_tab[1][12] = ff_put_h264_qpel8_mc03_neon;
193     c->put_h264_qpel_pixels_tab[1][13] = ff_put_h264_qpel8_mc13_neon;
194     c->put_h264_qpel_pixels_tab[1][14] = ff_put_h264_qpel8_mc23_neon;
195     c->put_h264_qpel_pixels_tab[1][15] = ff_put_h264_qpel8_mc33_neon;
196
197     c->avg_h264_qpel_pixels_tab[0][ 0] = ff_avg_h264_qpel16_mc00_neon;
198
199     c->h264_v_loop_filter_luma = ff_h264_v_loop_filter_luma_neon;
200     c->h264_h_loop_filter_luma = ff_h264_h_loop_filter_luma_neon;
201     c->h264_v_loop_filter_chroma = ff_h264_v_loop_filter_chroma_neon;
202     c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma_neon;
203
204     c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels_16x16_neon;
205     c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels_16x8_neon;
206     c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels_8x16_neon;
207     c->biweight_h264_pixels_tab[3] = ff_biweight_h264_pixels_8x8_neon;
208     c->biweight_h264_pixels_tab[4] = ff_biweight_h264_pixels_8x4_neon;
209     c->biweight_h264_pixels_tab[5] = ff_biweight_h264_pixels_4x8_neon;
210     c->biweight_h264_pixels_tab[6] = ff_biweight_h264_pixels_4x4_neon;
211     c->biweight_h264_pixels_tab[7] = ff_biweight_h264_pixels_4x2_neon;
212
213     c->h264_idct_add = ff_h264_idct_add_neon;
214     c->h264_idct_dc_add = ff_h264_idct_dc_add_neon;
215     c->h264_idct_add16      = ff_h264_idct_add16_neon;
216     c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
217     c->h264_idct_add8       = ff_h264_idct_add8_neon;
218
219     if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
220         c->float_to_int16 = ff_float_to_int16_neon;
221         c->float_to_int16_interleave = ff_float_to_int16_interleave_neon;
222     }
223 }