ipred.h (7047B)
1 /* 2 * Copyright © 2018-2021, VideoLAN and dav1d authors 3 * Copyright © 2018, Two Orioles, LLC 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions are met: 8 * 9 * 1. Redistributions of source code must retain the above copyright notice, this 10 * list of conditions and the following disclaimer. 11 * 12 * 2. Redistributions in binary form must reproduce the above copyright notice, 13 * this list of conditions and the following disclaimer in the documentation 14 * and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 18 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 19 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR 20 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 22 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 23 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 25 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #include "src/cpu.h" 29 #include "src/ipred.h" 30 31 #define decl_fn(type, name) \ 32 decl_##type##_fn(BF(dav1d_##name, ssse3)); \ 33 decl_##type##_fn(BF(dav1d_##name, avx2)); \ 34 decl_##type##_fn(BF(dav1d_##name, avx512icl)) 35 #define init_fn(type0, type1, name, suffix) \ 36 c->type0[type1] = BF(dav1d_##name, suffix) 37 38 #define init_angular_ipred_fn(type, name, suffix) \ 39 init_fn(intra_pred, type, name, suffix) 40 #define init_cfl_pred_fn(type, name, suffix) \ 41 init_fn(cfl_pred, type, name, suffix) 42 #define init_cfl_ac_fn(type, name, suffix) \ 43 init_fn(cfl_ac, type, name, suffix) 44 45 decl_fn(angular_ipred, ipred_dc); 46 decl_fn(angular_ipred, ipred_dc_128); 47 decl_fn(angular_ipred, ipred_dc_top); 48 decl_fn(angular_ipred, ipred_dc_left); 49 decl_fn(angular_ipred, ipred_h); 50 decl_fn(angular_ipred, ipred_v); 51 decl_fn(angular_ipred, ipred_paeth); 52 decl_fn(angular_ipred, ipred_smooth); 53 decl_fn(angular_ipred, ipred_smooth_h); 54 decl_fn(angular_ipred, ipred_smooth_v); 55 decl_fn(angular_ipred, ipred_z1); 56 decl_fn(angular_ipred, ipred_z2); 57 decl_fn(angular_ipred, ipred_z3); 58 decl_fn(angular_ipred, ipred_filter); 59 60 decl_fn(cfl_pred, ipred_cfl); 61 decl_fn(cfl_pred, ipred_cfl_128); 62 decl_fn(cfl_pred, ipred_cfl_top); 63 decl_fn(cfl_pred, ipred_cfl_left); 64 65 decl_fn(cfl_ac, ipred_cfl_ac_420); 66 decl_fn(cfl_ac, ipred_cfl_ac_422); 67 decl_fn(cfl_ac, ipred_cfl_ac_444); 68 69 decl_fn(pal_pred, pal_pred); 70 71 static ALWAYS_INLINE void intra_pred_dsp_init_x86(Dav1dIntraPredDSPContext *const c) { 72 const unsigned flags = dav1d_get_cpu_flags(); 73 74 if (!(flags & DAV1D_X86_CPU_FLAG_SSSE3)) return; 75 76 init_angular_ipred_fn(DC_PRED, ipred_dc, ssse3); 77 init_angular_ipred_fn(DC_128_PRED, ipred_dc_128, ssse3); 78 init_angular_ipred_fn(TOP_DC_PRED, ipred_dc_top, ssse3); 79 init_angular_ipred_fn(LEFT_DC_PRED, ipred_dc_left, ssse3); 80 init_angular_ipred_fn(HOR_PRED, ipred_h, ssse3); 81 init_angular_ipred_fn(VERT_PRED, ipred_v, ssse3); 82 init_angular_ipred_fn(PAETH_PRED, ipred_paeth, ssse3); 83 init_angular_ipred_fn(SMOOTH_PRED, ipred_smooth, ssse3); 84 init_angular_ipred_fn(SMOOTH_H_PRED, ipred_smooth_h, ssse3); 85 init_angular_ipred_fn(SMOOTH_V_PRED, ipred_smooth_v, ssse3); 86 init_angular_ipred_fn(Z1_PRED, ipred_z1, ssse3); 87 init_angular_ipred_fn(Z2_PRED, ipred_z2, ssse3); 88 init_angular_ipred_fn(Z3_PRED, ipred_z3, ssse3); 89 init_angular_ipred_fn(FILTER_PRED, ipred_filter, ssse3); 90 91 init_cfl_pred_fn(DC_PRED, ipred_cfl, ssse3); 92 init_cfl_pred_fn(DC_128_PRED, ipred_cfl_128, ssse3); 93 init_cfl_pred_fn(TOP_DC_PRED, ipred_cfl_top, ssse3); 94 init_cfl_pred_fn(LEFT_DC_PRED, ipred_cfl_left, ssse3); 95 96 init_cfl_ac_fn(DAV1D_PIXEL_LAYOUT_I420 - 1, ipred_cfl_ac_420, ssse3); 97 init_cfl_ac_fn(DAV1D_PIXEL_LAYOUT_I422 - 1, ipred_cfl_ac_422, ssse3); 98 init_cfl_ac_fn(DAV1D_PIXEL_LAYOUT_I444 - 1, ipred_cfl_ac_444, ssse3); 99 100 c->pal_pred = BF(dav1d_pal_pred, ssse3); 101 102 #if ARCH_X86_64 103 if (!(flags & DAV1D_X86_CPU_FLAG_AVX2)) return; 104 105 init_angular_ipred_fn(DC_PRED, ipred_dc, avx2); 106 init_angular_ipred_fn(DC_128_PRED, ipred_dc_128, avx2); 107 init_angular_ipred_fn(TOP_DC_PRED, ipred_dc_top, avx2); 108 init_angular_ipred_fn(LEFT_DC_PRED, ipred_dc_left, avx2); 109 init_angular_ipred_fn(HOR_PRED, ipred_h, avx2); 110 init_angular_ipred_fn(VERT_PRED, ipred_v, avx2); 111 init_angular_ipred_fn(PAETH_PRED, ipred_paeth, avx2); 112 init_angular_ipred_fn(SMOOTH_PRED, ipred_smooth, avx2); 113 init_angular_ipred_fn(SMOOTH_H_PRED, ipred_smooth_h, avx2); 114 init_angular_ipred_fn(SMOOTH_V_PRED, ipred_smooth_v, avx2); 115 init_angular_ipred_fn(Z1_PRED, ipred_z1, avx2); 116 init_angular_ipred_fn(Z2_PRED, ipred_z2, avx2); 117 init_angular_ipred_fn(Z3_PRED, ipred_z3, avx2); 118 init_angular_ipred_fn(FILTER_PRED, ipred_filter, avx2); 119 120 init_cfl_pred_fn(DC_PRED, ipred_cfl, avx2); 121 init_cfl_pred_fn(DC_128_PRED, ipred_cfl_128, avx2); 122 init_cfl_pred_fn(TOP_DC_PRED, ipred_cfl_top, avx2); 123 init_cfl_pred_fn(LEFT_DC_PRED, ipred_cfl_left, avx2); 124 125 init_cfl_ac_fn(DAV1D_PIXEL_LAYOUT_I420 - 1, ipred_cfl_ac_420, avx2); 126 init_cfl_ac_fn(DAV1D_PIXEL_LAYOUT_I422 - 1, ipred_cfl_ac_422, avx2); 127 init_cfl_ac_fn(DAV1D_PIXEL_LAYOUT_I444 - 1, ipred_cfl_ac_444, avx2); 128 129 c->pal_pred = BF(dav1d_pal_pred, avx2); 130 131 if (!(flags & DAV1D_X86_CPU_FLAG_AVX512ICL)) return; 132 133 #if BITDEPTH == 8 134 init_angular_ipred_fn(DC_PRED, ipred_dc, avx512icl); 135 init_angular_ipred_fn(DC_128_PRED, ipred_dc_128, avx512icl); 136 init_angular_ipred_fn(TOP_DC_PRED, ipred_dc_top, avx512icl); 137 init_angular_ipred_fn(LEFT_DC_PRED, ipred_dc_left, avx512icl); 138 init_angular_ipred_fn(HOR_PRED, ipred_h, avx512icl); 139 init_angular_ipred_fn(VERT_PRED, ipred_v, avx512icl); 140 init_angular_ipred_fn(Z2_PRED, ipred_z2, avx512icl); 141 #endif 142 init_angular_ipred_fn(PAETH_PRED, ipred_paeth, avx512icl); 143 init_angular_ipred_fn(SMOOTH_PRED, ipred_smooth, avx512icl); 144 init_angular_ipred_fn(SMOOTH_H_PRED, ipred_smooth_h, avx512icl); 145 init_angular_ipred_fn(SMOOTH_V_PRED, ipred_smooth_v, avx512icl); 146 init_angular_ipred_fn(Z1_PRED, ipred_z1, avx512icl); 147 init_angular_ipred_fn(Z2_PRED, ipred_z2, avx512icl); 148 init_angular_ipred_fn(Z3_PRED, ipred_z3, avx512icl); 149 init_angular_ipred_fn(FILTER_PRED, ipred_filter, avx512icl); 150 151 c->pal_pred = BF(dav1d_pal_pred, avx512icl); 152 #endif 153 }