vp8dsp_init_aarch64.c (6088B)
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 21 #include "libavutil/attributes.h" 22 #include "libavutil/aarch64/cpu.h" 23 #include "libavcodec/vp8dsp.h" 24 #include "vp8dsp.h" 25 26 void ff_vp8_luma_dc_wht_neon(int16_t block[4][4][16], int16_t dc[16]); 27 28 void ff_vp8_idct_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride); 29 void ff_vp8_idct_dc_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride); 30 void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride); 31 void ff_vp8_idct_dc_add4uv_neon(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride); 32 33 VP8_LF(neon); 34 35 VP8_EPEL(16, neon); 36 VP8_EPEL(8, neon); 37 VP8_EPEL(4, neon); 38 39 VP8_BILIN(16, neon); 40 VP8_BILIN(8, neon); 41 VP8_BILIN(4, neon); 42 43 av_cold void ff_vp78dsp_init_aarch64(VP8DSPContext *dsp) 44 { 45 if (!have_neon(av_get_cpu_flags())) 46 return; 47 dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon; 48 dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon; 49 dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon; 50 dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_neon; 51 52 dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon; 53 dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_neon; 54 dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_neon; 55 dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_neon; 56 dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_neon; 57 dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_neon; 58 dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_neon; 59 dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_neon; 60 dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_neon; 61 62 dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_neon; 63 dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_neon; 64 dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_neon; 65 dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_neon; 66 dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_neon; 67 dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_neon; 68 dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_neon; 69 dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_neon; 70 71 dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon; 72 dsp->put_vp8_bilinear_pixels_tab[0][0][1] = ff_put_vp8_bilin16_h_neon; 73 dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilin16_h_neon; 74 dsp->put_vp8_bilinear_pixels_tab[0][1][0] = ff_put_vp8_bilin16_v_neon; 75 dsp->put_vp8_bilinear_pixels_tab[0][1][1] = ff_put_vp8_bilin16_hv_neon; 76 dsp->put_vp8_bilinear_pixels_tab[0][1][2] = ff_put_vp8_bilin16_hv_neon; 77 dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilin16_v_neon; 78 dsp->put_vp8_bilinear_pixels_tab[0][2][1] = ff_put_vp8_bilin16_hv_neon; 79 dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilin16_hv_neon; 80 81 dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon; 82 dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilin8_h_neon; 83 dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilin8_h_neon; 84 dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilin8_v_neon; 85 dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilin8_hv_neon; 86 dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilin8_hv_neon; 87 dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilin8_v_neon; 88 dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilin8_hv_neon; 89 dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilin8_hv_neon; 90 91 dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilin4_h_neon; 92 dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilin4_h_neon; 93 dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilin4_v_neon; 94 dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilin4_hv_neon; 95 dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilin4_hv_neon; 96 dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilin4_v_neon; 97 dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_neon; 98 dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_neon; 99 } 100 101 av_cold void ff_vp8dsp_init_aarch64(VP8DSPContext *dsp) 102 { 103 if (!have_neon(av_get_cpu_flags())) 104 return; 105 dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_neon; 106 107 dsp->vp8_idct_add = ff_vp8_idct_add_neon; 108 dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_neon; 109 dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_neon; 110 dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon; 111 112 dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon; 113 dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon; 114 dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon; 115 dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon; 116 117 dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon; 118 dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon; 119 dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon; 120 dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon; 121 122 dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon; 123 dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon; 124 }