| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * SIMD-optimized halfpel functions | ||
| 3 | * Copyright (c) 2000, 2001 Fabrice Bellard | ||
| 4 | * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at> | ||
| 5 | * | ||
| 6 | * This file is part of FFmpeg. | ||
| 7 | * | ||
| 8 | * FFmpeg is free software; you can redistribute it and/or | ||
| 9 | * modify it under the terms of the GNU Lesser General Public | ||
| 10 | * License as published by the Free Software Foundation; either | ||
| 11 | * version 2.1 of the License, or (at your option) any later version. | ||
| 12 | * | ||
| 13 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 16 | * Lesser General Public License for more details. | ||
| 17 | * | ||
| 18 | * You should have received a copy of the GNU Lesser General Public | ||
| 19 | * License along with FFmpeg; if not, write to the Free Software | ||
| 20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 21 | * | ||
| 22 | * MMX optimization by Nick Kurshev <nickols_k@mail.ru> | ||
| 23 | */ | ||
| 24 | |||
| 25 | #include <stddef.h> | ||
| 26 | #include <stdint.h> | ||
| 27 | |||
| 28 | #include "libavutil/attributes.h" | ||
| 29 | #include "libavutil/cpu.h" | ||
| 30 | #include "libavutil/x86/cpu.h" | ||
| 31 | #include "libavcodec/avcodec.h" | ||
| 32 | #include "libavcodec/hpeldsp.h" | ||
| 33 | #include "libavcodec/pixels.h" | ||
| 34 | #include "fpel.h" | ||
| 35 | #include "hpeldsp.h" | ||
| 36 | |||
| 37 | void ff_put_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 38 | ptrdiff_t line_size, int h); | ||
| 39 | void ff_put_pixels16_x2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 40 | ptrdiff_t line_size, int h); | ||
| 41 | void ff_put_pixels16_x2_sse2(uint8_t *block, const uint8_t *pixels, | ||
| 42 | ptrdiff_t line_size, int h); | ||
| 43 | void ff_avg_pixels16_x2_sse2(uint8_t *block, const uint8_t *pixels, | ||
| 44 | ptrdiff_t line_size, int h); | ||
| 45 | void ff_put_pixels16_y2_sse2(uint8_t *block, const uint8_t *pixels, | ||
| 46 | ptrdiff_t line_size, int h); | ||
| 47 | void ff_avg_pixels16_y2_sse2(uint8_t *block, const uint8_t *pixels, | ||
| 48 | ptrdiff_t line_size, int h); | ||
| 49 | void ff_put_no_rnd_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 50 | ptrdiff_t line_size, int h); | ||
| 51 | void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block, | ||
| 52 | const uint8_t *pixels, | ||
| 53 | ptrdiff_t line_size, int h); | ||
| 54 | void ff_put_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 55 | ptrdiff_t line_size, int h); | ||
| 56 | void ff_put_no_rnd_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 57 | ptrdiff_t line_size, int h); | ||
| 58 | void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block, | ||
| 59 | const uint8_t *pixels, | ||
| 60 | ptrdiff_t line_size, int h); | ||
| 61 | void ff_avg_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 62 | ptrdiff_t line_size, int h); | ||
| 63 | void ff_avg_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 64 | ptrdiff_t line_size, int h); | ||
| 65 | void ff_avg_approx_pixels8_xy2_mmxext(uint8_t *block, const uint8_t *pixels, | ||
| 66 | ptrdiff_t line_size, int h); | ||
| 67 | |||
| 68 | #define put_pixels8_mmx ff_put_pixels8_mmx | ||
| 69 | #define put_pixels16_mmx ff_put_pixels16_mmx | ||
| 70 | #define put_pixels8_xy2_mmx ff_put_pixels8_xy2_mmx | ||
| 71 | #define put_no_rnd_pixels8_mmx ff_put_pixels8_mmx | ||
| 72 | #define put_no_rnd_pixels16_mmx ff_put_pixels16_mmx | ||
| 73 | |||
| 74 | #if HAVE_INLINE_ASM | ||
| 75 | |||
| 76 | /***********************************/ | ||
| 77 | /* MMX no rounding */ | ||
| 78 | #define DEF(x, y) x ## _no_rnd_ ## y ## _mmx | ||
| 79 | #define SET_RND MOVQ_WONE | ||
| 80 | #define STATIC static | ||
| 81 | |||
| 82 | #include "rnd_template.c" | ||
| 83 | |||
| 84 | #undef DEF | ||
| 85 | #undef SET_RND | ||
| 86 | #undef STATIC | ||
| 87 | |||
| 88 | // this routine is 'slightly' suboptimal but mostly unused | ||
| 89 | ✗ | static void avg_no_rnd_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels, | |
| 90 | ptrdiff_t line_size, int h) | ||
| 91 | { | ||
| 92 | ✗ | MOVQ_ZERO(mm7); | |
| 93 | ✗ | MOVQ_WONE(mm6); // =2 for rnd and =1 for no_rnd version | |
| 94 | ✗ | __asm__ volatile( | |
| 95 | "movq (%1), %%mm0 \n\t" | ||
| 96 | "movq 1(%1), %%mm4 \n\t" | ||
| 97 | "movq %%mm0, %%mm1 \n\t" | ||
| 98 | "movq %%mm4, %%mm5 \n\t" | ||
| 99 | "punpcklbw %%mm7, %%mm0 \n\t" | ||
| 100 | "punpcklbw %%mm7, %%mm4 \n\t" | ||
| 101 | "punpckhbw %%mm7, %%mm1 \n\t" | ||
| 102 | "punpckhbw %%mm7, %%mm5 \n\t" | ||
| 103 | "paddusw %%mm0, %%mm4 \n\t" | ||
| 104 | "paddusw %%mm1, %%mm5 \n\t" | ||
| 105 | "xor %%"FF_REG_a", %%"FF_REG_a" \n\t" | ||
| 106 | "add %3, %1 \n\t" | ||
| 107 | ".p2align 3 \n\t" | ||
| 108 | "1: \n\t" | ||
| 109 | "movq (%1, %%"FF_REG_a"), %%mm0 \n\t" | ||
| 110 | "movq 1(%1, %%"FF_REG_a"), %%mm2 \n\t" | ||
| 111 | "movq %%mm0, %%mm1 \n\t" | ||
| 112 | "movq %%mm2, %%mm3 \n\t" | ||
| 113 | "punpcklbw %%mm7, %%mm0 \n\t" | ||
| 114 | "punpcklbw %%mm7, %%mm2 \n\t" | ||
| 115 | "punpckhbw %%mm7, %%mm1 \n\t" | ||
| 116 | "punpckhbw %%mm7, %%mm3 \n\t" | ||
| 117 | "paddusw %%mm2, %%mm0 \n\t" | ||
| 118 | "paddusw %%mm3, %%mm1 \n\t" | ||
| 119 | "paddusw %%mm6, %%mm4 \n\t" | ||
| 120 | "paddusw %%mm6, %%mm5 \n\t" | ||
| 121 | "paddusw %%mm0, %%mm4 \n\t" | ||
| 122 | "paddusw %%mm1, %%mm5 \n\t" | ||
| 123 | "psrlw $2, %%mm4 \n\t" | ||
| 124 | "psrlw $2, %%mm5 \n\t" | ||
| 125 | "movq (%2, %%"FF_REG_a"), %%mm3 \n\t" | ||
| 126 | "packuswb %%mm5, %%mm4 \n\t" | ||
| 127 | "pcmpeqd %%mm2, %%mm2 \n\t" | ||
| 128 | "paddb %%mm2, %%mm2 \n\t" | ||
| 129 | PAVGB_MMX(%%mm3, %%mm4, %%mm5, %%mm2) | ||
| 130 | "movq %%mm5, (%2, %%"FF_REG_a") \n\t" | ||
| 131 | "add %3, %%"FF_REG_a" \n\t" | ||
| 132 | |||
| 133 | "movq (%1, %%"FF_REG_a"), %%mm2 \n\t" // 0 <-> 2 1 <-> 3 | ||
| 134 | "movq 1(%1, %%"FF_REG_a"), %%mm4 \n\t" | ||
| 135 | "movq %%mm2, %%mm3 \n\t" | ||
| 136 | "movq %%mm4, %%mm5 \n\t" | ||
| 137 | "punpcklbw %%mm7, %%mm2 \n\t" | ||
| 138 | "punpcklbw %%mm7, %%mm4 \n\t" | ||
| 139 | "punpckhbw %%mm7, %%mm3 \n\t" | ||
| 140 | "punpckhbw %%mm7, %%mm5 \n\t" | ||
| 141 | "paddusw %%mm2, %%mm4 \n\t" | ||
| 142 | "paddusw %%mm3, %%mm5 \n\t" | ||
| 143 | "paddusw %%mm6, %%mm0 \n\t" | ||
| 144 | "paddusw %%mm6, %%mm1 \n\t" | ||
| 145 | "paddusw %%mm4, %%mm0 \n\t" | ||
| 146 | "paddusw %%mm5, %%mm1 \n\t" | ||
| 147 | "psrlw $2, %%mm0 \n\t" | ||
| 148 | "psrlw $2, %%mm1 \n\t" | ||
| 149 | "movq (%2, %%"FF_REG_a"), %%mm3 \n\t" | ||
| 150 | "packuswb %%mm1, %%mm0 \n\t" | ||
| 151 | "pcmpeqd %%mm2, %%mm2 \n\t" | ||
| 152 | "paddb %%mm2, %%mm2 \n\t" | ||
| 153 | PAVGB_MMX(%%mm3, %%mm0, %%mm1, %%mm2) | ||
| 154 | "movq %%mm1, (%2, %%"FF_REG_a") \n\t" | ||
| 155 | "add %3, %%"FF_REG_a" \n\t" | ||
| 156 | |||
| 157 | "subl $2, %0 \n\t" | ||
| 158 | "jnz 1b \n\t" | ||
| 159 | :"+g"(h), "+S"(pixels) | ||
| 160 | :"D"(block), "r"((x86_reg)line_size) | ||
| 161 | :FF_REG_a, "memory"); | ||
| 162 | ✗ | } | |
| 163 | |||
| 164 | ✗ | static void put_no_rnd_pixels8_x2_mmx(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), %%"FF_REG_a" \n\t" | ||
| 169 | ".p2align 3 \n\t" | ||
| 170 | "1: \n\t" | ||
| 171 | "movq (%1), %%mm0 \n\t" | ||
| 172 | "movq 1(%1), %%mm1 \n\t" | ||
| 173 | "movq (%1, %3), %%mm2 \n\t" | ||
| 174 | "movq 1(%1, %3), %%mm3 \n\t" | ||
| 175 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) | ||
| 176 | "movq %%mm4, (%2) \n\t" | ||
| 177 | "movq %%mm5, (%2, %3) \n\t" | ||
| 178 | "add %%"FF_REG_a", %1 \n\t" | ||
| 179 | "add %%"FF_REG_a", %2 \n\t" | ||
| 180 | "movq (%1), %%mm0 \n\t" | ||
| 181 | "movq 1(%1), %%mm1 \n\t" | ||
| 182 | "movq (%1, %3), %%mm2 \n\t" | ||
| 183 | "movq 1(%1, %3), %%mm3 \n\t" | ||
| 184 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) | ||
| 185 | "movq %%mm4, (%2) \n\t" | ||
| 186 | "movq %%mm5, (%2, %3) \n\t" | ||
| 187 | "add %%"FF_REG_a", %1 \n\t" | ||
| 188 | "add %%"FF_REG_a", %2 \n\t" | ||
| 189 | "subl $4, %0 \n\t" | ||
| 190 | "jnz 1b \n\t" | ||
| 191 | :"+g"(h), "+S"(pixels), "+D"(block) | ||
| 192 | :"r"((x86_reg)line_size) | ||
| 193 | :FF_REG_a, "memory"); | ||
| 194 | ✗ | } | |
| 195 | |||
| 196 | 1096 | static void put_no_rnd_pixels16_x2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) | |
| 197 | { | ||
| 198 | 1096 | MOVQ_BFE(mm6); | |
| 199 | 1096 | __asm__ volatile( | |
| 200 | "lea (%3, %3), %%"FF_REG_a" \n\t" | ||
| 201 | ".p2align 3 \n\t" | ||
| 202 | "1: \n\t" | ||
| 203 | "movq (%1), %%mm0 \n\t" | ||
| 204 | "movq 1(%1), %%mm1 \n\t" | ||
| 205 | "movq (%1, %3), %%mm2 \n\t" | ||
| 206 | "movq 1(%1, %3), %%mm3 \n\t" | ||
| 207 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) | ||
| 208 | "movq %%mm4, (%2) \n\t" | ||
| 209 | "movq %%mm5, (%2, %3) \n\t" | ||
| 210 | "movq 8(%1), %%mm0 \n\t" | ||
| 211 | "movq 9(%1), %%mm1 \n\t" | ||
| 212 | "movq 8(%1, %3), %%mm2 \n\t" | ||
| 213 | "movq 9(%1, %3), %%mm3 \n\t" | ||
| 214 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) | ||
| 215 | "movq %%mm4, 8(%2) \n\t" | ||
| 216 | "movq %%mm5, 8(%2, %3) \n\t" | ||
| 217 | "add %%"FF_REG_a", %1 \n\t" | ||
| 218 | "add %%"FF_REG_a", %2 \n\t" | ||
| 219 | "movq (%1), %%mm0 \n\t" | ||
| 220 | "movq 1(%1), %%mm1 \n\t" | ||
| 221 | "movq (%1, %3), %%mm2 \n\t" | ||
| 222 | "movq 1(%1, %3), %%mm3 \n\t" | ||
| 223 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) | ||
| 224 | "movq %%mm4, (%2) \n\t" | ||
| 225 | "movq %%mm5, (%2, %3) \n\t" | ||
| 226 | "movq 8(%1), %%mm0 \n\t" | ||
| 227 | "movq 9(%1), %%mm1 \n\t" | ||
| 228 | "movq 8(%1, %3), %%mm2 \n\t" | ||
| 229 | "movq 9(%1, %3), %%mm3 \n\t" | ||
| 230 | PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) | ||
| 231 | "movq %%mm4, 8(%2) \n\t" | ||
| 232 | "movq %%mm5, 8(%2, %3) \n\t" | ||
| 233 | "add %%"FF_REG_a", %1 \n\t" | ||
| 234 | "add %%"FF_REG_a", %2 \n\t" | ||
| 235 | "subl $4, %0 \n\t" | ||
| 236 | "jnz 1b \n\t" | ||
| 237 | :"+g"(h), "+S"(pixels), "+D"(block) | ||
| 238 | :"r"((x86_reg)line_size) | ||
| 239 | :FF_REG_a, "memory"); | ||
| 240 | 1096 | } | |
| 241 | |||
| 242 | 2954 | static void put_no_rnd_pixels8_y2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) | |
| 243 | { | ||
| 244 | 2954 | MOVQ_BFE(mm6); | |
| 245 | 2954 | __asm__ volatile( | |
| 246 | "lea (%3, %3), %%"FF_REG_a" \n\t" | ||
| 247 | "movq (%1), %%mm0 \n\t" | ||
| 248 | ".p2align 3 \n\t" | ||
| 249 | "1: \n\t" | ||
| 250 | "movq (%1, %3), %%mm1 \n\t" | ||
| 251 | "movq (%1, %%"FF_REG_a"),%%mm2\n\t" | ||
| 252 | PAVGBP_MMX_NO_RND(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) | ||
| 253 | "movq %%mm4, (%2) \n\t" | ||
| 254 | "movq %%mm5, (%2, %3) \n\t" | ||
| 255 | "add %%"FF_REG_a", %1 \n\t" | ||
| 256 | "add %%"FF_REG_a", %2 \n\t" | ||
| 257 | "movq (%1, %3), %%mm1 \n\t" | ||
| 258 | "movq (%1, %%"FF_REG_a"),%%mm0\n\t" | ||
| 259 | PAVGBP_MMX_NO_RND(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) | ||
| 260 | "movq %%mm4, (%2) \n\t" | ||
| 261 | "movq %%mm5, (%2, %3) \n\t" | ||
| 262 | "add %%"FF_REG_a", %1 \n\t" | ||
| 263 | "add %%"FF_REG_a", %2 \n\t" | ||
| 264 | "subl $4, %0 \n\t" | ||
| 265 | "jnz 1b \n\t" | ||
| 266 | :"+g"(h), "+S"(pixels), "+D"(block) | ||
| 267 | :"r"((x86_reg)line_size) | ||
| 268 | :FF_REG_a, "memory"); | ||
| 269 | 2954 | } | |
| 270 | |||
| 271 | ✗ | static void avg_no_rnd_pixels16_x2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) | |
| 272 | { | ||
| 273 | ✗ | MOVQ_BFE(mm6); | |
| 274 | ✗ | __asm__ volatile( | |
| 275 | ".p2align 3 \n\t" | ||
| 276 | "1: \n\t" | ||
| 277 | "movq (%1), %%mm0 \n\t" | ||
| 278 | "movq 1(%1), %%mm1 \n\t" | ||
| 279 | "movq (%2), %%mm3 \n\t" | ||
| 280 | PAVGB_MMX_NO_RND(%%mm0, %%mm1, %%mm2, %%mm6) | ||
| 281 | PAVGB_MMX(%%mm3, %%mm2, %%mm0, %%mm6) | ||
| 282 | "movq %%mm0, (%2) \n\t" | ||
| 283 | "movq 8(%1), %%mm0 \n\t" | ||
| 284 | "movq 9(%1), %%mm1 \n\t" | ||
| 285 | "movq 8(%2), %%mm3 \n\t" | ||
| 286 | PAVGB_MMX_NO_RND(%%mm0, %%mm1, %%mm2, %%mm6) | ||
| 287 | PAVGB_MMX(%%mm3, %%mm2, %%mm0, %%mm6) | ||
| 288 | "movq %%mm0, 8(%2) \n\t" | ||
| 289 | "add %3, %1 \n\t" | ||
| 290 | "add %3, %2 \n\t" | ||
| 291 | "subl $1, %0 \n\t" | ||
| 292 | "jnz 1b \n\t" | ||
| 293 | :"+g"(h), "+S"(pixels), "+D"(block) | ||
| 294 | :"r"((x86_reg)line_size) | ||
| 295 | :"memory"); | ||
| 296 | ✗ | } | |
| 297 | |||
| 298 | ✗ | static void avg_no_rnd_pixels8_y2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) | |
| 299 | { | ||
| 300 | ✗ | MOVQ_BFE(mm6); | |
| 301 | ✗ | __asm__ volatile( | |
| 302 | "lea (%3, %3), %%"FF_REG_a" \n\t" | ||
| 303 | "movq (%1), %%mm0 \n\t" | ||
| 304 | ".p2align 3 \n\t" | ||
| 305 | "1: \n\t" | ||
| 306 | "movq (%1, %3), %%mm1 \n\t" | ||
| 307 | "movq (%1, %%"FF_REG_a"), %%mm2 \n\t" | ||
| 308 | PAVGBP_MMX_NO_RND(%%mm1, %%mm0, %%mm4, %%mm2, %%mm1, %%mm5) | ||
| 309 | "movq (%2), %%mm3 \n\t" | ||
| 310 | PAVGB_MMX(%%mm3, %%mm4, %%mm0, %%mm6) | ||
| 311 | "movq (%2, %3), %%mm3 \n\t" | ||
| 312 | PAVGB_MMX(%%mm3, %%mm5, %%mm1, %%mm6) | ||
| 313 | "movq %%mm0, (%2) \n\t" | ||
| 314 | "movq %%mm1, (%2, %3) \n\t" | ||
| 315 | "add %%"FF_REG_a", %1 \n\t" | ||
| 316 | "add %%"FF_REG_a", %2 \n\t" | ||
| 317 | |||
| 318 | "movq (%1, %3), %%mm1 \n\t" | ||
| 319 | "movq (%1, %%"FF_REG_a"), %%mm0 \n\t" | ||
| 320 | PAVGBP_MMX_NO_RND(%%mm1, %%mm2, %%mm4, %%mm0, %%mm1, %%mm5) | ||
| 321 | "movq (%2), %%mm3 \n\t" | ||
| 322 | PAVGB_MMX(%%mm3, %%mm4, %%mm2, %%mm6) | ||
| 323 | "movq (%2, %3), %%mm3 \n\t" | ||
| 324 | PAVGB_MMX(%%mm3, %%mm5, %%mm1, %%mm6) | ||
| 325 | "movq %%mm2, (%2) \n\t" | ||
| 326 | "movq %%mm1, (%2, %3) \n\t" | ||
| 327 | "add %%"FF_REG_a", %1 \n\t" | ||
| 328 | "add %%"FF_REG_a", %2 \n\t" | ||
| 329 | |||
| 330 | "subl $4, %0 \n\t" | ||
| 331 | "jnz 1b \n\t" | ||
| 332 | :"+g"(h), "+S"(pixels), "+D"(block) | ||
| 333 | :"r"((x86_reg)line_size) | ||
| 334 | :FF_REG_a, "memory"); | ||
| 335 | ✗ | } | |
| 336 | |||
| 337 | #if HAVE_MMX | ||
| 338 | ✗ | CALL_2X_PIXELS(avg_no_rnd_pixels16_y2_mmx, avg_no_rnd_pixels8_y2_mmx, 8) | |
| 339 | 1477 | CALL_2X_PIXELS(put_no_rnd_pixels16_y2_mmx, put_no_rnd_pixels8_y2_mmx, 8) | |
| 340 | |||
| 341 | ✗ | CALL_2X_PIXELS(avg_no_rnd_pixels16_xy2_mmx, avg_no_rnd_pixels8_xy2_mmx, 8) | |
| 342 | 10118 | CALL_2X_PIXELS(put_no_rnd_pixels16_xy2_mmx, put_no_rnd_pixels8_xy2_mmx, 8) | |
| 343 | #endif | ||
| 344 | |||
| 345 | /***********************************/ | ||
| 346 | /* MMX rounding */ | ||
| 347 | |||
| 348 | #define SET_RND MOVQ_WTWO | ||
| 349 | #define DEF(x, y) ff_ ## x ## _ ## y ## _mmx | ||
| 350 | #define STATIC | ||
| 351 | |||
| 352 | #include "rnd_template.c" | ||
| 353 | |||
| 354 | #undef NO_AVG | ||
| 355 | #undef DEF | ||
| 356 | #undef SET_RND | ||
| 357 | |||
| 358 | #if HAVE_MMX | ||
| 359 | ✗ | CALL_2X_PIXELS(put_pixels16_xy2_mmx, ff_put_pixels8_xy2_mmx, 8) | |
| 360 | #endif | ||
| 361 | |||
| 362 | #endif /* HAVE_INLINE_ASM */ | ||
| 363 | |||
| 364 | |||
| 365 | #if HAVE_X86ASM | ||
| 366 | |||
| 367 | #define HPELDSP_AVG_PIXELS16(CPUEXT) \ | ||
| 368 | CALL_2X_PIXELS(put_no_rnd_pixels16_x2 ## CPUEXT, ff_put_no_rnd_pixels8_x2 ## CPUEXT, 8) \ | ||
| 369 | CALL_2X_PIXELS(put_pixels16_y2 ## CPUEXT, ff_put_pixels8_y2 ## CPUEXT, 8) \ | ||
| 370 | CALL_2X_PIXELS(put_no_rnd_pixels16_y2 ## CPUEXT, ff_put_no_rnd_pixels8_y2 ## CPUEXT, 8) \ | ||
| 371 | CALL_2X_PIXELS(avg_pixels16_x2 ## CPUEXT, ff_avg_pixels8_x2 ## CPUEXT, 8) \ | ||
| 372 | CALL_2X_PIXELS(avg_pixels16_y2 ## CPUEXT, ff_avg_pixels8_y2 ## CPUEXT, 8) \ | ||
| 373 | CALL_2X_PIXELS(avg_pixels16_xy2 ## CPUEXT, ff_avg_pixels8_xy2 ## CPUEXT, 8) \ | ||
| 374 | CALL_2X_PIXELS(avg_approx_pixels16_xy2## CPUEXT, ff_avg_approx_pixels8_xy2## CPUEXT, 8) | ||
| 375 | |||
| 376 | 17364 | HPELDSP_AVG_PIXELS16(_mmxext) | |
| 377 | |||
| 378 | #endif /* HAVE_X86ASM */ | ||
| 379 | |||
| 380 | #define SET_HPEL_FUNCS_EXT(PFX, IDX, SIZE, CPU) \ | ||
| 381 | if (HAVE_MMX_EXTERNAL) \ | ||
| 382 | c->PFX ## _pixels_tab IDX [0] = PFX ## _pixels ## SIZE ## _ ## CPU | ||
| 383 | |||
| 384 | #define SET_HPEL_FUNCS03(PFX, IDX, SIZE, CPU) \ | ||
| 385 | do { \ | ||
| 386 | SET_HPEL_FUNCS_EXT(PFX, IDX, SIZE, CPU); \ | ||
| 387 | c->PFX ## _pixels_tab IDX [3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \ | ||
| 388 | } while (0) | ||
| 389 | #define SET_HPEL_FUNCS12(PFX, IDX, SIZE, CPU) \ | ||
| 390 | do { \ | ||
| 391 | c->PFX ## _pixels_tab IDX [1] = PFX ## _pixels ## SIZE ## _x2_ ## CPU; \ | ||
| 392 | c->PFX ## _pixels_tab IDX [2] = PFX ## _pixels ## SIZE ## _y2_ ## CPU; \ | ||
| 393 | } while (0) | ||
| 394 | #define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU) \ | ||
| 395 | do { \ | ||
| 396 | SET_HPEL_FUNCS03(PFX, IDX, SIZE, CPU); \ | ||
| 397 | SET_HPEL_FUNCS12(PFX, IDX, SIZE, CPU); \ | ||
| 398 | } while (0) | ||
| 399 | |||
| 400 | 190 | static void hpeldsp_init_mmx(HpelDSPContext *c, int flags) | |
| 401 | { | ||
| 402 | #if HAVE_MMX_INLINE | ||
| 403 | 190 | SET_HPEL_FUNCS03(put, [0], 16, mmx); | |
| 404 | 190 | SET_HPEL_FUNCS(put_no_rnd, [0], 16, mmx); | |
| 405 | 190 | SET_HPEL_FUNCS12(avg_no_rnd, , 16, mmx); | |
| 406 | 190 | c->avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels16_xy2_mmx; | |
| 407 | 190 | SET_HPEL_FUNCS03(put, [1], 8, mmx); | |
| 408 | 190 | SET_HPEL_FUNCS(put_no_rnd, [1], 8, mmx); | |
| 409 | #endif | ||
| 410 | 190 | } | |
| 411 | |||
| 412 | 190 | static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags) | |
| 413 | { | ||
| 414 | #if HAVE_MMXEXT_EXTERNAL | ||
| 415 | 190 | c->put_pixels_tab[0][1] = ff_put_pixels16_x2_mmxext; | |
| 416 | 190 | c->put_pixels_tab[0][2] = put_pixels16_y2_mmxext; | |
| 417 | |||
| 418 | 190 | c->avg_pixels_tab[0][0] = ff_avg_pixels16_mmxext; | |
| 419 | 190 | c->avg_pixels_tab[0][1] = avg_pixels16_x2_mmxext; | |
| 420 | 190 | c->avg_pixels_tab[0][2] = avg_pixels16_y2_mmxext; | |
| 421 | 190 | c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmxext; | |
| 422 | |||
| 423 | 190 | c->put_pixels_tab[1][1] = ff_put_pixels8_x2_mmxext; | |
| 424 | 190 | c->put_pixels_tab[1][2] = ff_put_pixels8_y2_mmxext; | |
| 425 | |||
| 426 | 190 | c->avg_pixels_tab[1][0] = ff_avg_pixels8_mmxext; | |
| 427 | 190 | c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_mmxext; | |
| 428 | 190 | c->avg_pixels_tab[1][2] = ff_avg_pixels8_y2_mmxext; | |
| 429 | 190 | c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_mmxext; | |
| 430 | |||
| 431 | 190 | c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; | |
| 432 | 190 | c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; | |
| 433 | |||
| 434 | 190 | c->avg_no_rnd_pixels_tab[0] = ff_avg_pixels16_mmxext; | |
| 435 | |||
| 436 |
2/2✓ Branch 0 taken 173 times.
✓ Branch 1 taken 17 times.
|
190 | if (!(flags & AV_CODEC_FLAG_BITEXACT)) { |
| 437 | 173 | c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmxext; | |
| 438 | 173 | c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmxext; | |
| 439 | 173 | c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_mmxext; | |
| 440 | 173 | c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_mmxext; | |
| 441 | |||
| 442 | 173 | c->avg_pixels_tab[0][3] = avg_approx_pixels16_xy2_mmxext; | |
| 443 | 173 | c->avg_pixels_tab[1][3] = ff_avg_approx_pixels8_xy2_mmxext; | |
| 444 | } | ||
| 445 | #endif /* HAVE_MMXEXT_EXTERNAL */ | ||
| 446 | 190 | } | |
| 447 | |||
| 448 | 190 | static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags) | |
| 449 | { | ||
| 450 | #if HAVE_SSE2_EXTERNAL | ||
| 451 | 190 | c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; | |
| 452 | 190 | c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2; | |
| 453 | 190 | c->put_pixels_tab[0][1] = ff_put_pixels16_x2_sse2; | |
| 454 | 190 | c->put_pixels_tab[0][2] = ff_put_pixels16_y2_sse2; | |
| 455 | 190 | c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_sse2; | |
| 456 | 190 | c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2; | |
| 457 | 190 | c->avg_pixels_tab[0][1] = ff_avg_pixels16_x2_sse2; | |
| 458 | 190 | c->avg_pixels_tab[0][2] = ff_avg_pixels16_y2_sse2; | |
| 459 | 190 | c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_sse2; | |
| 460 | 190 | c->avg_no_rnd_pixels_tab[0] = ff_avg_pixels16_sse2; | |
| 461 | #endif /* HAVE_SSE2_EXTERNAL */ | ||
| 462 | 190 | } | |
| 463 | |||
| 464 | 190 | static void hpeldsp_init_ssse3(HpelDSPContext *c, int flags) | |
| 465 | { | ||
| 466 | #if HAVE_SSSE3_EXTERNAL | ||
| 467 | 190 | c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_ssse3; | |
| 468 | 190 | c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_ssse3; | |
| 469 | 190 | c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_ssse3; | |
| 470 | 190 | c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_ssse3; | |
| 471 | #endif | ||
| 472 | 190 | } | |
| 473 | |||
| 474 | 1134 | av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) | |
| 475 | { | ||
| 476 | 1134 | int cpu_flags = av_get_cpu_flags(); | |
| 477 | |||
| 478 |
2/2✓ Branch 0 taken 190 times.
✓ Branch 1 taken 944 times.
|
1134 | if (INLINE_MMX(cpu_flags)) |
| 479 | 190 | hpeldsp_init_mmx(c, flags); | |
| 480 | |||
| 481 |
2/2✓ Branch 0 taken 190 times.
✓ Branch 1 taken 944 times.
|
1134 | if (EXTERNAL_MMXEXT(cpu_flags)) |
| 482 | 190 | hpeldsp_init_mmxext(c, flags); | |
| 483 | |||
| 484 |
3/4✓ Branch 0 taken 190 times.
✓ Branch 1 taken 944 times.
✓ Branch 2 taken 190 times.
✗ Branch 3 not taken.
|
1134 | if (EXTERNAL_SSE2_FAST(cpu_flags)) |
| 485 | 190 | hpeldsp_init_sse2_fast(c, flags); | |
| 486 | |||
| 487 |
2/2✓ Branch 0 taken 190 times.
✓ Branch 1 taken 944 times.
|
1134 | if (EXTERNAL_SSSE3(cpu_flags)) |
| 488 | 190 | hpeldsp_init_ssse3(c, flags); | |
| 489 | 1134 | } | |
| 490 |