2 * ARM NEON optimised DSP functions
3 * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
5 * This file is part of FFmpeg.
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.
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.
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
24 #include "libavcodec/avcodec.h"
25 #include "libavcodec/dsputil.h"
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);
42 void ff_avg_pixels16_neon(uint8_t *, const uint8_t *, int, int);
44 void ff_add_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
45 void ff_put_signed_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
47 void ff_put_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int);
48 void ff_put_h264_qpel16_mc10_neon(uint8_t *, uint8_t *, int);
49 void ff_put_h264_qpel16_mc20_neon(uint8_t *, uint8_t *, int);
50 void ff_put_h264_qpel16_mc30_neon(uint8_t *, uint8_t *, int);
51 void ff_put_h264_qpel16_mc01_neon(uint8_t *, uint8_t *, int);
52 void ff_put_h264_qpel16_mc11_neon(uint8_t *, uint8_t *, int);
53 void ff_put_h264_qpel16_mc21_neon(uint8_t *, uint8_t *, int);
54 void ff_put_h264_qpel16_mc31_neon(uint8_t *, uint8_t *, int);
55 void ff_put_h264_qpel16_mc02_neon(uint8_t *, uint8_t *, int);
56 void ff_put_h264_qpel16_mc12_neon(uint8_t *, uint8_t *, int);
57 void ff_put_h264_qpel16_mc22_neon(uint8_t *, uint8_t *, int);
58 void ff_put_h264_qpel16_mc32_neon(uint8_t *, uint8_t *, int);
59 void ff_put_h264_qpel16_mc03_neon(uint8_t *, uint8_t *, int);
60 void ff_put_h264_qpel16_mc13_neon(uint8_t *, uint8_t *, int);
61 void ff_put_h264_qpel16_mc23_neon(uint8_t *, uint8_t *, int);
62 void ff_put_h264_qpel16_mc33_neon(uint8_t *, uint8_t *, int);
64 void ff_put_h264_qpel8_mc00_neon(uint8_t *, uint8_t *, int);
65 void ff_put_h264_qpel8_mc10_neon(uint8_t *, uint8_t *, int);
66 void ff_put_h264_qpel8_mc20_neon(uint8_t *, uint8_t *, int);
67 void ff_put_h264_qpel8_mc30_neon(uint8_t *, uint8_t *, int);
68 void ff_put_h264_qpel8_mc01_neon(uint8_t *, uint8_t *, int);
69 void ff_put_h264_qpel8_mc11_neon(uint8_t *, uint8_t *, int);
70 void ff_put_h264_qpel8_mc21_neon(uint8_t *, uint8_t *, int);
71 void ff_put_h264_qpel8_mc31_neon(uint8_t *, uint8_t *, int);
72 void ff_put_h264_qpel8_mc02_neon(uint8_t *, uint8_t *, int);
73 void ff_put_h264_qpel8_mc12_neon(uint8_t *, uint8_t *, int);
74 void ff_put_h264_qpel8_mc22_neon(uint8_t *, uint8_t *, int);
75 void ff_put_h264_qpel8_mc32_neon(uint8_t *, uint8_t *, int);
76 void ff_put_h264_qpel8_mc03_neon(uint8_t *, uint8_t *, int);
77 void ff_put_h264_qpel8_mc13_neon(uint8_t *, uint8_t *, int);
78 void ff_put_h264_qpel8_mc23_neon(uint8_t *, uint8_t *, int);
79 void ff_put_h264_qpel8_mc33_neon(uint8_t *, uint8_t *, int);
81 void ff_avg_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int);
83 void ff_put_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
84 void ff_put_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
86 void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
87 void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
89 void ff_h264_v_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
90 int beta, int8_t *tc0);
91 void ff_h264_h_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
92 int beta, int8_t *tc0);
93 void ff_h264_v_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
94 int beta, int8_t *tc0);
95 void ff_h264_h_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
96 int beta, int8_t *tc0);
98 void ff_weight_h264_pixels_16x16_neon(uint8_t *ds, int stride, int log2_den,
99 int weight, int offset);
100 void ff_weight_h264_pixels_16x8_neon(uint8_t *ds, int stride, int log2_den,
101 int weight, int offset);
102 void ff_weight_h264_pixels_8x16_neon(uint8_t *ds, int stride, int log2_den,
103 int weight, int offset);
104 void ff_weight_h264_pixels_8x8_neon(uint8_t *ds, int stride, int log2_den,
105 int weight, int offset);
106 void ff_weight_h264_pixels_8x4_neon(uint8_t *ds, int stride, int log2_den,
107 int weight, int offset);
108 void ff_weight_h264_pixels_4x8_neon(uint8_t *ds, int stride, int log2_den,
109 int weight, int offset);
110 void ff_weight_h264_pixels_4x4_neon(uint8_t *ds, int stride, int log2_den,
111 int weight, int offset);
112 void ff_weight_h264_pixels_4x2_neon(uint8_t *ds, int stride, int log2_den,
113 int weight, int offset);
115 void ff_biweight_h264_pixels_16x16_neon(uint8_t *dst, uint8_t *src, int stride,
116 int log2_den, int weightd, int weights,
118 void ff_biweight_h264_pixels_16x8_neon(uint8_t *dst, uint8_t *src, int stride,
119 int log2_den, int weightd, int weights,
121 void ff_biweight_h264_pixels_8x16_neon(uint8_t *dst, uint8_t *src, int stride,
122 int log2_den, int weightd, int weights,
124 void ff_biweight_h264_pixels_8x8_neon(uint8_t *dst, uint8_t *src, int stride,
125 int log2_den, int weightd, int weights,
127 void ff_biweight_h264_pixels_8x4_neon(uint8_t *dst, uint8_t *src, int stride,
128 int log2_den, int weightd, int weights,
130 void ff_biweight_h264_pixels_4x8_neon(uint8_t *dst, uint8_t *src, int stride,
131 int log2_den, int weightd, int weights,
133 void ff_biweight_h264_pixels_4x4_neon(uint8_t *dst, uint8_t *src, int stride,
134 int log2_den, int weightd, int weights,
136 void ff_biweight_h264_pixels_4x2_neon(uint8_t *dst, uint8_t *src, int stride,
137 int log2_den, int weightd, int weights,
140 void ff_h264_idct_add_neon(uint8_t *dst, DCTELEM *block, int stride);
141 void ff_h264_idct_dc_add_neon(uint8_t *dst, DCTELEM *block, int stride);
142 void ff_h264_idct_add16_neon(uint8_t *dst, const int *block_offset,
143 DCTELEM *block, int stride,
144 const uint8_t nnzc[6*8]);
145 void ff_h264_idct_add16intra_neon(uint8_t *dst, const int *block_offset,
146 DCTELEM *block, int stride,
147 const uint8_t nnzc[6*8]);
148 void ff_h264_idct_add8_neon(uint8_t **dest, const int *block_offset,
149 DCTELEM *block, int stride,
150 const uint8_t nnzc[6*8]);
152 void ff_vector_fmul_neon(float *dst, const float *src, int len);
153 void ff_vector_fmul_window_neon(float *dst, const float *src0,
154 const float *src1, const float *win,
155 float add_bias, int len);
157 void ff_float_to_int16_neon(int16_t *, const float *, long);
158 void ff_float_to_int16_interleave_neon(int16_t *, const float **, long, int);
160 void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
162 c->put_pixels_tab[0][0] = ff_put_pixels16_neon;
163 c->put_pixels_tab[0][1] = ff_put_pixels16_x2_neon;
164 c->put_pixels_tab[0][2] = ff_put_pixels16_y2_neon;
165 c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_neon;
166 c->put_pixels_tab[1][0] = ff_put_pixels8_neon;
167 c->put_pixels_tab[1][1] = ff_put_pixels8_x2_neon;
168 c->put_pixels_tab[1][2] = ff_put_pixels8_y2_neon;
169 c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_neon;
171 c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_neon;
172 c->put_no_rnd_pixels_tab[0][1] = ff_put_pixels16_x2_no_rnd_neon;
173 c->put_no_rnd_pixels_tab[0][2] = ff_put_pixels16_y2_no_rnd_neon;
174 c->put_no_rnd_pixels_tab[0][3] = ff_put_pixels16_xy2_no_rnd_neon;
175 c->put_no_rnd_pixels_tab[1][0] = ff_put_pixels8_neon;
176 c->put_no_rnd_pixels_tab[1][1] = ff_put_pixels8_x2_no_rnd_neon;
177 c->put_no_rnd_pixels_tab[1][2] = ff_put_pixels8_y2_no_rnd_neon;
178 c->put_no_rnd_pixels_tab[1][3] = ff_put_pixels8_xy2_no_rnd_neon;
180 c->avg_pixels_tab[0][0] = ff_avg_pixels16_neon;
182 c->add_pixels_clamped = ff_add_pixels_clamped_neon;
183 c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_neon;
185 c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon;
186 c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon;
188 c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_neon;
189 c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_neon;
191 c->put_h264_qpel_pixels_tab[0][ 0] = ff_put_h264_qpel16_mc00_neon;
192 c->put_h264_qpel_pixels_tab[0][ 1] = ff_put_h264_qpel16_mc10_neon;
193 c->put_h264_qpel_pixels_tab[0][ 2] = ff_put_h264_qpel16_mc20_neon;
194 c->put_h264_qpel_pixels_tab[0][ 3] = ff_put_h264_qpel16_mc30_neon;
195 c->put_h264_qpel_pixels_tab[0][ 4] = ff_put_h264_qpel16_mc01_neon;
196 c->put_h264_qpel_pixels_tab[0][ 5] = ff_put_h264_qpel16_mc11_neon;
197 c->put_h264_qpel_pixels_tab[0][ 6] = ff_put_h264_qpel16_mc21_neon;
198 c->put_h264_qpel_pixels_tab[0][ 7] = ff_put_h264_qpel16_mc31_neon;
199 c->put_h264_qpel_pixels_tab[0][ 8] = ff_put_h264_qpel16_mc02_neon;
200 c->put_h264_qpel_pixels_tab[0][ 9] = ff_put_h264_qpel16_mc12_neon;
201 c->put_h264_qpel_pixels_tab[0][10] = ff_put_h264_qpel16_mc22_neon;
202 c->put_h264_qpel_pixels_tab[0][11] = ff_put_h264_qpel16_mc32_neon;
203 c->put_h264_qpel_pixels_tab[0][12] = ff_put_h264_qpel16_mc03_neon;
204 c->put_h264_qpel_pixels_tab[0][13] = ff_put_h264_qpel16_mc13_neon;
205 c->put_h264_qpel_pixels_tab[0][14] = ff_put_h264_qpel16_mc23_neon;
206 c->put_h264_qpel_pixels_tab[0][15] = ff_put_h264_qpel16_mc33_neon;
208 c->put_h264_qpel_pixels_tab[1][ 0] = ff_put_h264_qpel8_mc00_neon;
209 c->put_h264_qpel_pixels_tab[1][ 1] = ff_put_h264_qpel8_mc10_neon;
210 c->put_h264_qpel_pixels_tab[1][ 2] = ff_put_h264_qpel8_mc20_neon;
211 c->put_h264_qpel_pixels_tab[1][ 3] = ff_put_h264_qpel8_mc30_neon;
212 c->put_h264_qpel_pixels_tab[1][ 4] = ff_put_h264_qpel8_mc01_neon;
213 c->put_h264_qpel_pixels_tab[1][ 5] = ff_put_h264_qpel8_mc11_neon;
214 c->put_h264_qpel_pixels_tab[1][ 6] = ff_put_h264_qpel8_mc21_neon;
215 c->put_h264_qpel_pixels_tab[1][ 7] = ff_put_h264_qpel8_mc31_neon;
216 c->put_h264_qpel_pixels_tab[1][ 8] = ff_put_h264_qpel8_mc02_neon;
217 c->put_h264_qpel_pixels_tab[1][ 9] = ff_put_h264_qpel8_mc12_neon;
218 c->put_h264_qpel_pixels_tab[1][10] = ff_put_h264_qpel8_mc22_neon;
219 c->put_h264_qpel_pixels_tab[1][11] = ff_put_h264_qpel8_mc32_neon;
220 c->put_h264_qpel_pixels_tab[1][12] = ff_put_h264_qpel8_mc03_neon;
221 c->put_h264_qpel_pixels_tab[1][13] = ff_put_h264_qpel8_mc13_neon;
222 c->put_h264_qpel_pixels_tab[1][14] = ff_put_h264_qpel8_mc23_neon;
223 c->put_h264_qpel_pixels_tab[1][15] = ff_put_h264_qpel8_mc33_neon;
225 c->avg_h264_qpel_pixels_tab[0][ 0] = ff_avg_h264_qpel16_mc00_neon;
227 c->h264_v_loop_filter_luma = ff_h264_v_loop_filter_luma_neon;
228 c->h264_h_loop_filter_luma = ff_h264_h_loop_filter_luma_neon;
229 c->h264_v_loop_filter_chroma = ff_h264_v_loop_filter_chroma_neon;
230 c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma_neon;
232 c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels_16x16_neon;
233 c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels_16x8_neon;
234 c->weight_h264_pixels_tab[2] = ff_weight_h264_pixels_8x16_neon;
235 c->weight_h264_pixels_tab[3] = ff_weight_h264_pixels_8x8_neon;
236 c->weight_h264_pixels_tab[4] = ff_weight_h264_pixels_8x4_neon;
237 c->weight_h264_pixels_tab[5] = ff_weight_h264_pixels_4x8_neon;
238 c->weight_h264_pixels_tab[6] = ff_weight_h264_pixels_4x4_neon;
239 c->weight_h264_pixels_tab[7] = ff_weight_h264_pixels_4x2_neon;
241 c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels_16x16_neon;
242 c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels_16x8_neon;
243 c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels_8x16_neon;
244 c->biweight_h264_pixels_tab[3] = ff_biweight_h264_pixels_8x8_neon;
245 c->biweight_h264_pixels_tab[4] = ff_biweight_h264_pixels_8x4_neon;
246 c->biweight_h264_pixels_tab[5] = ff_biweight_h264_pixels_4x8_neon;
247 c->biweight_h264_pixels_tab[6] = ff_biweight_h264_pixels_4x4_neon;
248 c->biweight_h264_pixels_tab[7] = ff_biweight_h264_pixels_4x2_neon;
250 c->h264_idct_add = ff_h264_idct_add_neon;
251 c->h264_idct_dc_add = ff_h264_idct_dc_add_neon;
252 c->h264_idct_add16 = ff_h264_idct_add16_neon;
253 c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
254 c->h264_idct_add8 = ff_h264_idct_add8_neon;
256 c->vector_fmul = ff_vector_fmul_neon;
257 c->vector_fmul_window = ff_vector_fmul_window_neon;
259 if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
260 c->float_to_int16 = ff_float_to_int16_neon;
261 c->float_to_int16_interleave = ff_float_to_int16_interleave_neon;