1/* 2 * SIMD-optimized halfpel functions are compiled twice for rnd/no_rnd 3 * Copyright (c) 2000, 2001 Fabrice Bellard 4 * Copyright (c) 2003-2004 Michael Niedermayer <michaelni@gmx.at> 5 * 6 * MMX optimization by Nick Kurshev <nickols_k@mail.ru> 7 * mostly rewritten by Michael Niedermayer <michaelni@gmx.at> 8 * and improved by Zdenek Kabelac <kabi@users.sf.net> 9 * 10 * This file is part of FFmpeg. 11 * 12 * FFmpeg is free software; you can redistribute it and/or 13 * modify it under the terms of the GNU Lesser General Public 14 * License as published by the Free Software Foundation; either 15 * version 2.1 of the License, or (at your option) any later version. 16 * 17 * FFmpeg is distributed in the hope that it will be useful, 18 * but WITHOUT ANY WARRANTY; without even the implied warranty of 19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 20 * Lesser General Public License for more details. 21 * 22 * You should have received a copy of the GNU Lesser General Public 23 * License along with FFmpeg; if not, write to the Free Software 24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 25 */ 26 27#include <stddef.h> 28#include <stdint.h> 29 30// put_pixels 31static void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) 32{ 33 MOVQ_BFE(mm6); 34 __asm__ volatile( 35 "lea (%3, %3), %%"REG_a" \n\t" 36 ".p2align 3 \n\t" 37 "1: \n\t" 38 "movq (%1), %%mm0 \n\t" 39 "movq 1(%1), %%mm1 \n\t" 40 "movq (%1, %3), %%mm2 \n\t" 41 "movq 1(%1, %3), %%mm3 \n\t" 42 PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) 43 "movq %%mm4, (%2) \n\t" 44 "movq %%mm5, (%2, %3) \n\t" 45 "add %%"REG_a", %1 \n\t" 46 "add %%"REG_a", %2 \n\t" 47 "movq (%1), %%mm0 \n\t" 48 "movq 1(%1), %%mm1 \n\t" 49 "movq (%1, %3), %%mm2 \n\t" 50 "movq 1(%1, %3), %%mm3 \n\t" 51 PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) 52 "movq %%mm4, (%2) \n\t" 53 "movq %%mm5, (%2, %3) \n\t" 54 "add %%"REG_a", %1 \n\t" 55 "add %%"REG_a", %2 \n\t" 56 "subl $4, %0 \n\t" 57 "jnz 1b \n\t" 58 :"+g"(h), "+S"(pixels), "+D"(block) 59 :"r"((x86_reg)line_size) 60 :REG_a, "memory"); 61} 62 63static void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) 64{ 65 MOVQ_BFE(mm6); 66 __asm__ volatile( 67 "lea (%3, %3), %%"REG_a" \n\t" 68 ".p2align 3 \n\t" 69 "1: \n\t" 70 "movq (%1), %%mm0 \n\t" 71 "movq 1(%1), %%mm1 \n\t" 72 "movq (%1, %3), %%mm2 \n\t" 73 "movq 1(%1, %3), %%mm3 \n\t" 74 PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) 75 "movq %%mm4, (%2) \n\t" 76 "movq %%mm5, (%2, %3) \n\t" 77 "movq 8(%1), %%mm0 \n\t" 78 "movq 9(%1), %%mm1 \n\t" 79 "movq 8(%1, %3), %%mm2 \n\t" 80 "movq 9(%1, %3), %%mm3 \n\t" 81 PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) 82 "movq %%mm4, 8(%2) \n\t" 83 "movq %%mm5, 8(%2, %3) \n\t" 84 "add %%"REG_a", %1 \n\t" 85 "add %%"REG_a", %2 \n\t" 86 "movq (%1), %%mm0 \n\t" 87 "movq 1(%1), %%mm1 \n\t" 88 "movq (%1, %3), %%mm2 \n\t" 89 "movq 1(%1, %3), %%mm3 \n\t" 90 PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) 91 "movq %%mm4, (%2) \n\t" 92 "movq %%mm5, (%2, %3) \n\t" 93 "movq 8(%1), %%mm0 \n\t" 94 "movq 9(%1), %%mm1 \n\t" 95 "movq 8(%1, %3), %%mm2 \n\t" 96 "movq 9(%1, %3), %%mm3 \n\t" 97 PAVGBP(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) 98 "movq %%mm4, 8(%2) \n\t" 99 "movq %%mm5, 8(%2, %3) \n\t" 100 "add %%"REG_a", %1 \n\t" 101 "add %%"REG_a", %2 \n\t" 102 "subl $4, %0 \n\t" 103 "jnz 1b \n\t" 104 :"+g"(h), "+S"(pixels), "+D"(block) 105 :"r"((x86_reg)line_size) 106 :REG_a, "memory"); 107} 108 109static void DEF(put, pixels8_y2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) 110{ 111 MOVQ_BFE(mm6); 112 __asm__ volatile( 113 "lea (%3, %3), %%"REG_a" \n\t" 114 "movq (%1), %%mm0 \n\t" 115 ".p2align 3 \n\t" 116 "1: \n\t" 117 "movq (%1, %3), %%mm1 \n\t" 118 "movq (%1, %%"REG_a"),%%mm2 \n\t" 119 PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) 120 "movq %%mm4, (%2) \n\t" 121 "movq %%mm5, (%2, %3) \n\t" 122 "add %%"REG_a", %1 \n\t" 123 "add %%"REG_a", %2 \n\t" 124 "movq (%1, %3), %%mm1 \n\t" 125 "movq (%1, %%"REG_a"),%%mm0 \n\t" 126 PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) 127 "movq %%mm4, (%2) \n\t" 128 "movq %%mm5, (%2, %3) \n\t" 129 "add %%"REG_a", %1 \n\t" 130 "add %%"REG_a", %2 \n\t" 131 "subl $4, %0 \n\t" 132 "jnz 1b \n\t" 133 :"+g"(h), "+S"(pixels), "+D"(block) 134 :"r"((x86_reg)line_size) 135 :REG_a, "memory"); 136} 137 138static void DEF(avg, pixels16_x2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) 139{ 140 MOVQ_BFE(mm6); 141 JUMPALIGN(); 142 do { 143 __asm__ volatile( 144 "movq %1, %%mm0 \n\t" 145 "movq 1%1, %%mm1 \n\t" 146 "movq %0, %%mm3 \n\t" 147 PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) 148 PAVGB_MMX(%%mm3, %%mm2, %%mm0, %%mm6) 149 "movq %%mm0, %0 \n\t" 150 "movq 8%1, %%mm0 \n\t" 151 "movq 9%1, %%mm1 \n\t" 152 "movq 8%0, %%mm3 \n\t" 153 PAVGB(%%mm0, %%mm1, %%mm2, %%mm6) 154 PAVGB_MMX(%%mm3, %%mm2, %%mm0, %%mm6) 155 "movq %%mm0, 8%0 \n\t" 156 :"+m"(*block) 157 :"m"(*pixels) 158 :"memory"); 159 pixels += line_size; 160 block += line_size; 161 } while (--h); 162} 163 164static void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) 165{ 166 MOVQ_BFE(mm6); 167 __asm__ volatile( 168 "lea (%3, %3), %%"REG_a" \n\t" 169 "movq (%1), %%mm0 \n\t" 170 ".p2align 3 \n\t" 171 "1: \n\t" 172 "movq (%1, %3), %%mm1 \n\t" 173 "movq (%1, %%"REG_a"), %%mm2 \n\t" 174 PAVGBP(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) 175 "movq (%2), %%mm3 \n\t" 176 PAVGB_MMX(%%mm3, %%mm4, %%mm0, %%mm6) 177 "movq (%2, %3), %%mm3 \n\t" 178 PAVGB_MMX(%%mm3, %%mm5, %%mm1, %%mm6) 179 "movq %%mm0, (%2) \n\t" 180 "movq %%mm1, (%2, %3) \n\t" 181 "add %%"REG_a", %1 \n\t" 182 "add %%"REG_a", %2 \n\t" 183 184 "movq (%1, %3), %%mm1 \n\t" 185 "movq (%1, %%"REG_a"), %%mm0 \n\t" 186 PAVGBP(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) 187 "movq (%2), %%mm3 \n\t" 188 PAVGB_MMX(%%mm3, %%mm4, %%mm2, %%mm6) 189 "movq (%2, %3), %%mm3 \n\t" 190 PAVGB_MMX(%%mm3, %%mm5, %%mm1, %%mm6) 191 "movq %%mm2, (%2) \n\t" 192 "movq %%mm1, (%2, %3) \n\t" 193 "add %%"REG_a", %1 \n\t" 194 "add %%"REG_a", %2 \n\t" 195 196 "subl $4, %0 \n\t" 197 "jnz 1b \n\t" 198 :"+g"(h), "+S"(pixels), "+D"(block) 199 :"r"((x86_reg)line_size) 200 :REG_a, "memory"); 201} 202