Parent Directory
| 
Revision Log
Revision 1587 - (view) (download)
| 1 : | edgomez | 1586 | ;/***************************************************************************** |
| 2 : | ; * | ||
| 3 : | ; * XVID MPEG-4 VIDEO CODEC | ||
| 4 : | ; * - Quarter-pixel interpolation - | ||
| 5 : | ; * Copyright(C) 2002 Pascal Massimino <skal@planet-d.net> | ||
| 6 : | ; * 2004 Andre Werthmann <wertmann@aei.mpg.de> | ||
| 7 : | ; * | ||
| 8 : | ; * This file is part of XviD, a free MPEG-4 video encoder/decoder | ||
| 9 : | ; * | ||
| 10 : | ; * XviD is free software; you can redistribute it and/or modify it | ||
| 11 : | ; * under the terms of the GNU General Public License as published by | ||
| 12 : | ; * the Free Software Foundation; either version 2 of the License, or | ||
| 13 : | ; * (at your option) any later version. | ||
| 14 : | ; * | ||
| 15 : | ; * This program is distributed in the hope that it will be useful, | ||
| 16 : | ; * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 17 : | ; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
| 18 : | ; * GNU General Public License for more details. | ||
| 19 : | ; * | ||
| 20 : | ; * You should have received a copy of the GNU General Public License | ||
| 21 : | ; * along with this program; if not, write to the Free Software | ||
| 22 : | ; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
| 23 : | ; * | ||
| 24 : | edgomez | 1587 | ; * $Id: qpel_mmx.asm,v 1.2 2005-01-06 23:48:08 edgomez Exp $ |
| 25 : | edgomez | 1586 | ; * |
| 26 : | ; *************************************************************************/ | ||
| 27 : | |||
| 28 : | ;/************************************************************************** | ||
| 29 : | ; * | ||
| 30 : | ; * History: | ||
| 31 : | ; * | ||
| 32 : | ; * 22.10.2002 initial coding. unoptimized 'proof of concept', | ||
| 33 : | ; * just to heft the qpel filtering. - Skal - | ||
| 34 : | ; * | ||
| 35 : | ; *************************************************************************/ | ||
| 36 : | |||
| 37 : | |||
| 38 : | %define USE_TABLES ; in order to use xvid_FIR_x_x_x_x tables | ||
| 39 : | ; instead of xvid_Expand_mmx... | ||
| 40 : | |||
| 41 : | |||
| 42 : | BITS 64 | ||
| 43 : | |||
| 44 : | |||
| 45 : | %macro cglobal 1 | ||
| 46 : | %ifdef PREFIX | ||
| 47 : | %ifdef MARK_FUNCS | ||
| 48 : | global _%1:function %1.endfunc-%1 | ||
| 49 : | %define %1 _%1:function %1.endfunc-%1 | ||
| 50 : | %else | ||
| 51 : | global _%1 | ||
| 52 : | %define %1 _%1 | ||
| 53 : | %endif | ||
| 54 : | %else | ||
| 55 : | %ifdef MARK_FUNCS | ||
| 56 : | global %1:function %1.endfunc-%1 | ||
| 57 : | %else | ||
| 58 : | global %1 | ||
| 59 : | %endif | ||
| 60 : | %endif | ||
| 61 : | %endmacro | ||
| 62 : | |||
| 63 : | ;////////////////////////////////////////////////////////////////////// | ||
| 64 : | ;// Declarations | ||
| 65 : | ;// all signatures are: | ||
| 66 : | ;// void XXX(uint8_t *dst, const uint8_t *src, | ||
| 67 : | ;// int32_t length, int32_t stride, int32_t rounding) | ||
| 68 : | ;////////////////////////////////////////////////////////////////////// | ||
| 69 : | |||
| 70 : | cglobal xvid_H_Pass_16_x86_64 | ||
| 71 : | cglobal xvid_H_Pass_Avrg_16_x86_64 | ||
| 72 : | cglobal xvid_H_Pass_Avrg_Up_16_x86_64 | ||
| 73 : | cglobal xvid_V_Pass_16_x86_64 | ||
| 74 : | cglobal xvid_V_Pass_Avrg_16_x86_64 | ||
| 75 : | cglobal xvid_V_Pass_Avrg_Up_16_x86_64 | ||
| 76 : | cglobal xvid_H_Pass_8_x86_64 | ||
| 77 : | cglobal xvid_H_Pass_Avrg_8_x86_64 | ||
| 78 : | cglobal xvid_H_Pass_Avrg_Up_8_x86_64 | ||
| 79 : | cglobal xvid_V_Pass_8_x86_64 | ||
| 80 : | cglobal xvid_V_Pass_Avrg_8_x86_64 | ||
| 81 : | cglobal xvid_V_Pass_Avrg_Up_8_x86_64 | ||
| 82 : | |||
| 83 : | cglobal xvid_H_Pass_Add_16_x86_64 | ||
| 84 : | cglobal xvid_H_Pass_Avrg_Add_16_x86_64 | ||
| 85 : | cglobal xvid_H_Pass_Avrg_Up_Add_16_x86_64 | ||
| 86 : | cglobal xvid_V_Pass_Add_16_x86_64 | ||
| 87 : | cglobal xvid_V_Pass_Avrg_Add_16_x86_64 | ||
| 88 : | cglobal xvid_V_Pass_Avrg_Up_Add_16_x86_64 | ||
| 89 : | cglobal xvid_H_Pass_8_Add_x86_64 | ||
| 90 : | cglobal xvid_H_Pass_Avrg_8_Add_x86_64 | ||
| 91 : | cglobal xvid_H_Pass_Avrg_Up_8_Add_x86_64 | ||
| 92 : | cglobal xvid_V_Pass_8_Add_x86_64 | ||
| 93 : | cglobal xvid_V_Pass_Avrg_8_Add_x86_64 | ||
| 94 : | cglobal xvid_V_Pass_Avrg_Up_8_Add_x86_64 | ||
| 95 : | |||
| 96 : | cglobal xvid_Expand_mmx | ||
| 97 : | |||
| 98 : | ;%ifdef USE_TABLES | ||
| 99 : | |||
| 100 : | cglobal xvid_FIR_1_0_0_0 | ||
| 101 : | cglobal xvid_FIR_3_1_0_0 | ||
| 102 : | cglobal xvid_FIR_6_3_1_0 | ||
| 103 : | cglobal xvid_FIR_14_3_2_1 | ||
| 104 : | cglobal xvid_FIR_20_6_3_1 | ||
| 105 : | cglobal xvid_FIR_20_20_6_3 | ||
| 106 : | cglobal xvid_FIR_23_19_6_3 | ||
| 107 : | cglobal xvid_FIR_7_20_20_6 | ||
| 108 : | cglobal xvid_FIR_6_20_20_6 | ||
| 109 : | cglobal xvid_FIR_6_20_20_7 | ||
| 110 : | cglobal xvid_FIR_3_6_20_20 | ||
| 111 : | cglobal xvid_FIR_3_6_19_23 | ||
| 112 : | cglobal xvid_FIR_1_3_6_20 | ||
| 113 : | cglobal xvid_FIR_1_2_3_14 | ||
| 114 : | cglobal xvid_FIR_0_1_3_6 | ||
| 115 : | cglobal xvid_FIR_0_0_1_3 | ||
| 116 : | cglobal xvid_FIR_0_0_0_1 | ||
| 117 : | |||
| 118 : | ;%endif | ||
| 119 : | |||
| 120 : | SECTION .data align=32 | ||
| 121 : | xvid_Expand_mmx: | ||
| 122 : | times 256*4 dw 0 ; uint16_t xvid_Expand_mmx[256][4] | ||
| 123 : | .endfunc | ||
| 124 : | |||
| 125 : | xvid_FIR_1_0_0_0: | ||
| 126 : | times 256*4 dw 0 | ||
| 127 : | .endfunc | ||
| 128 : | |||
| 129 : | xvid_FIR_3_1_0_0: | ||
| 130 : | times 256*4 dw 0 | ||
| 131 : | .endfunc | ||
| 132 : | |||
| 133 : | xvid_FIR_6_3_1_0: | ||
| 134 : | times 256*4 dw 0 | ||
| 135 : | .endfunc | ||
| 136 : | |||
| 137 : | xvid_FIR_14_3_2_1: | ||
| 138 : | times 256*4 dw 0 | ||
| 139 : | .endfunc | ||
| 140 : | |||
| 141 : | xvid_FIR_20_6_3_1: | ||
| 142 : | times 256*4 dw 0 | ||
| 143 : | .endfunc | ||
| 144 : | |||
| 145 : | xvid_FIR_20_20_6_3: | ||
| 146 : | times 256*4 dw 0 | ||
| 147 : | .endfunc | ||
| 148 : | |||
| 149 : | xvid_FIR_23_19_6_3: | ||
| 150 : | times 256*4 dw 0 | ||
| 151 : | .endfunc | ||
| 152 : | |||
| 153 : | xvid_FIR_7_20_20_6: | ||
| 154 : | times 256*4 dw 0 | ||
| 155 : | .endfunc | ||
| 156 : | |||
| 157 : | xvid_FIR_6_20_20_6: | ||
| 158 : | times 256*4 dw 0 | ||
| 159 : | .endfunc | ||
| 160 : | |||
| 161 : | xvid_FIR_6_20_20_7: | ||
| 162 : | times 256*4 dw 0 | ||
| 163 : | .endfunc | ||
| 164 : | |||
| 165 : | xvid_FIR_3_6_20_20: | ||
| 166 : | times 256*4 dw 0 | ||
| 167 : | .endfunc | ||
| 168 : | |||
| 169 : | xvid_FIR_3_6_19_23: | ||
| 170 : | times 256*4 dw 0 | ||
| 171 : | .endfunc | ||
| 172 : | |||
| 173 : | xvid_FIR_1_3_6_20: | ||
| 174 : | times 256*4 dw 0 | ||
| 175 : | .endfunc | ||
| 176 : | |||
| 177 : | xvid_FIR_1_2_3_14: | ||
| 178 : | times 256*4 dw 0 | ||
| 179 : | .endfunc | ||
| 180 : | |||
| 181 : | xvid_FIR_0_1_3_6: | ||
| 182 : | times 256*4 dw 0 | ||
| 183 : | .endfunc | ||
| 184 : | |||
| 185 : | xvid_FIR_0_0_1_3: | ||
| 186 : | times 256*4 dw 0 | ||
| 187 : | .endfunc | ||
| 188 : | |||
| 189 : | xvid_FIR_0_0_0_1: | ||
| 190 : | times 256*4 dw 0 | ||
| 191 : | .endfunc | ||
| 192 : | |||
| 193 : | ;////////////////////////////////////////////////////////////////////// | ||
| 194 : | |||
| 195 : | %ifdef FORMAT_COFF | ||
| 196 : | SECTION .rodata | ||
| 197 : | %else | ||
| 198 : | SECTION .rodata align=16 | ||
| 199 : | %endif | ||
| 200 : | |||
| 201 : | align 16 | ||
| 202 : | Rounder1_MMX: | ||
| 203 : | times 4 dw 1 | ||
| 204 : | Rounder0_MMX: | ||
| 205 : | times 4 dw 0 | ||
| 206 : | |||
| 207 : | align 16 | ||
| 208 : | Rounder_QP_MMX | ||
| 209 : | times 4 dw 16 | ||
| 210 : | times 4 dw 15 | ||
| 211 : | |||
| 212 : | %ifndef USE_TABLES | ||
| 213 : | |||
| 214 : | align 16 | ||
| 215 : | |||
| 216 : | ; H-Pass table shared by 16x? and 8x? filters | ||
| 217 : | |||
| 218 : | FIR_R0: dw 14, -3, 2, -1 | ||
| 219 : | align 16 | ||
| 220 : | FIR_R1: dw 23, 19, -6, 3, -1, 0, 0, 0 | ||
| 221 : | |||
| 222 : | FIR_R2: dw -7, 20, 20, -6, 3, -1, 0, 0 | ||
| 223 : | |||
| 224 : | FIR_R3: dw 3, -6, 20, 20, -6, 3, -1, 0 | ||
| 225 : | |||
| 226 : | FIR_R4: dw -1, 3, -6, 20, 20, -6, 3, -1 | ||
| 227 : | |||
| 228 : | FIR_R5: dw 0, -1, 3, -6, 20, 20, -6, 3, -1, 0, 0, 0 | ||
| 229 : | align 16 | ||
| 230 : | FIR_R6: dw 0, 0, -1, 3, -6, 20, 20, -6, 3, -1, 0, 0 | ||
| 231 : | align 16 | ||
| 232 : | FIR_R7: dw 0, 0, 0, -1, 3, -6, 20, 20, -6, 3, -1, 0 | ||
| 233 : | align 16 | ||
| 234 : | FIR_R8: dw -1, 3, -6, 20, 20, -6, 3, -1 | ||
| 235 : | |||
| 236 : | FIR_R9: dw 0, -1, 3, -6, 20, 20, -6, 3, -1, 0, 0, 0 | ||
| 237 : | align 16 | ||
| 238 : | FIR_R10: dw 0, 0, -1, 3, -6, 20, 20, -6, 3, -1, 0, 0 | ||
| 239 : | align 16 | ||
| 240 : | FIR_R11: dw 0, 0, 0, -1, 3, -6, 20, 20, -6, 3, -1, 0 | ||
| 241 : | align 16 | ||
| 242 : | FIR_R12: dw -1, 3, -6, 20, 20, -6, 3, -1 | ||
| 243 : | |||
| 244 : | FIR_R13: dw 0, -1, 3, -6, 20, 20, -6, 3 | ||
| 245 : | |||
| 246 : | FIR_R14: dw 0, 0, -1, 3, -6, 20, 20, -7 | ||
| 247 : | |||
| 248 : | FIR_R15: dw 0, 0, 0, -1, 3, -6, 19, 23 | ||
| 249 : | |||
| 250 : | FIR_R16: dw -1, 2, -3, 14 | ||
| 251 : | |||
| 252 : | %endif ; !USE_TABLES | ||
| 253 : | |||
| 254 : | ; V-Pass taps | ||
| 255 : | |||
| 256 : | align 16 | ||
| 257 : | FIR_Cm7: times 4 dw -7 | ||
| 258 : | FIR_Cm6: times 4 dw -6 | ||
| 259 : | FIR_Cm3: times 4 dw -3 | ||
| 260 : | FIR_Cm1: times 4 dw -1 | ||
| 261 : | FIR_C2: times 4 dw 2 | ||
| 262 : | FIR_C3: times 4 dw 3 | ||
| 263 : | FIR_C14: times 4 dw 14 | ||
| 264 : | FIR_C19: times 4 dw 19 | ||
| 265 : | FIR_C20: times 4 dw 20 | ||
| 266 : | FIR_C23: times 4 dw 23 | ||
| 267 : | |||
| 268 : | SECTION .text align=16 | ||
| 269 : | |||
| 270 : | ;////////////////////////////////////////////////////////////////////// | ||
| 271 : | ;// Here we go with the Q-Pel mess. | ||
| 272 : | ;// For horizontal passes, we process 4 *output* pixel in parallel | ||
| 273 : | ;// For vertical ones, we process 4 *input* pixel in parallel. | ||
| 274 : | ;////////////////////////////////////////////////////////////////////// | ||
| 275 : | |||
| 276 : | %macro PROLOG_NO_AVRG 0 | ||
| 277 : | push rbp | ||
| 278 : | mov rax, r8 ; Rnd | ||
| 279 : | and rax, 1 | ||
| 280 : | mov rbp, rcx ; BpS | ||
| 281 : | mov rcx, rdx ; Size | ||
| 282 : | ; rsi is Src | ||
| 283 : | ; rdi is Dst | ||
| 284 : | lea r9, [Rounder_QP_MMX wrt rip] | ||
| 285 : | movq mm7, [r9+rax*8] | ||
| 286 : | %endmacro | ||
| 287 : | |||
| 288 : | %macro EPILOG_NO_AVRG 0 | ||
| 289 : | pop rbp | ||
| 290 : | ret | ||
| 291 : | %endmacro | ||
| 292 : | |||
| 293 : | %macro PROLOG_AVRG 0 | ||
| 294 : | push rbx | ||
| 295 : | push rbp | ||
| 296 : | mov rax, r8 ; Rnd | ||
| 297 : | and rax, 1 | ||
| 298 : | mov rbp, rcx ; BpS | ||
| 299 : | mov rcx, rdx ; Size | ||
| 300 : | ; rsi is Src | ||
| 301 : | ; rdi is Dst | ||
| 302 : | |||
| 303 : | lea r9, [Rounder_QP_MMX wrt rip] | ||
| 304 : | movq mm7, [r9+rax*8] | ||
| 305 : | lea r9, [Rounder1_MMX wrt rip] | ||
| 306 : | lea rbx, [r9+rax*8] | ||
| 307 : | %endmacro | ||
| 308 : | |||
| 309 : | %macro EPILOG_AVRG 0 | ||
| 310 : | pop rbp | ||
| 311 : | pop rbx | ||
| 312 : | ret | ||
| 313 : | %endmacro | ||
| 314 : | |||
| 315 : | ;////////////////////////////////////////////////////////////////////// | ||
| 316 : | ;// | ||
| 317 : | ;// All horizontal passes | ||
| 318 : | ;// | ||
| 319 : | ;////////////////////////////////////////////////////////////////////// | ||
| 320 : | |||
| 321 : | ; macros for USE_TABLES | ||
| 322 : | |||
| 323 : | %macro TLOAD 2 ; %1,%2: src pixels | ||
| 324 : | movzx rax, byte [rsi+%1] | ||
| 325 : | movzx rdx, byte [rsi+%2] | ||
| 326 : | lea r9, [xvid_FIR_14_3_2_1 wrt rip] | ||
| 327 : | movq mm0, [r9 + rax*8 ] | ||
| 328 : | lea r9, [xvid_FIR_1_2_3_14 wrt rip] | ||
| 329 : | movq mm3, [r9 + rdx*8 ] | ||
| 330 : | paddw mm0, mm7 | ||
| 331 : | paddw mm3, mm7 | ||
| 332 : | %endmacro | ||
| 333 : | |||
| 334 : | %macro TACCUM2 5 ;%1:src pixel/%2-%3:Taps tables/ %4-%5:dst regs | ||
| 335 : | movzx rax, byte [rsi+%1] | ||
| 336 : | lea r9, [%2 wrt rip] | ||
| 337 : | paddw %4, [r9 + rax*8] | ||
| 338 : | lea r9, [%3 wrt rip] | ||
| 339 : | paddw %5, [r9 + rax*8] | ||
| 340 : | %endmacro | ||
| 341 : | |||
| 342 : | %macro TACCUM3 7 ;%1:src pixel/%2-%4:Taps tables/%5-%7:dst regs | ||
| 343 : | movzx rax, byte [rsi+%1] | ||
| 344 : | lea r9, [%2 wrt rip] | ||
| 345 : | paddw %5, [r9 + rax*8] | ||
| 346 : | lea r9, [%3 wrt rip] | ||
| 347 : | paddw %6, [r9 + rax*8] | ||
| 348 : | lea r9, [%4 wrt rip] | ||
| 349 : | paddw %7, [r9 + rax*8] | ||
| 350 : | %endmacro | ||
| 351 : | |||
| 352 : | ;////////////////////////////////////////////////////////////////////// | ||
| 353 : | |||
| 354 : | ; macros without USE_TABLES | ||
| 355 : | |||
| 356 : | %macro LOAD 2 ; %1,%2: src pixels | ||
| 357 : | movzx rax, byte [rsi+%1] | ||
| 358 : | movzx rdx, byte [rsi+%2] | ||
| 359 : | lea r9, [xvid_Expand_mmx wrt rip] | ||
| 360 : | movq mm0, [r9 + rax*8] | ||
| 361 : | movq mm3, [r9 + rdx*8] | ||
| 362 : | pmullw mm0, [FIR_R0 wrt rip] | ||
| 363 : | pmullw mm3, [FIR_R16 wrt rip] | ||
| 364 : | paddw mm0, mm7 | ||
| 365 : | paddw mm3, mm7 | ||
| 366 : | %endmacro | ||
| 367 : | |||
| 368 : | %macro ACCUM2 4 ;src pixel/Taps/dst regs #1-#2 | ||
| 369 : | movzx rax, byte [rsi+%1] | ||
| 370 : | lea r9, [xvid_Expand_mmx wrt rip] | ||
| 371 : | movq mm4, [r9 + rax*8] | ||
| 372 : | movq mm5, mm4 | ||
| 373 : | lea r9, [%2 wrt rip] ; data | ||
| 374 : | pmullw mm4, [r9] | ||
| 375 : | pmullw mm5, [r9+8] | ||
| 376 : | paddw %3, mm4 | ||
| 377 : | paddw %4, mm5 | ||
| 378 : | %endmacro | ||
| 379 : | |||
| 380 : | %macro ACCUM3 5 ;src pixel/Taps/dst regs #1-#2-#3 | ||
| 381 : | movzx rax, byte [rsi+%1] | ||
| 382 : | lea r9, [xvid_Expand_mmx wrt rip] | ||
| 383 : | movq mm4, [r9 + rax*8] | ||
| 384 : | movq mm5, mm4 | ||
| 385 : | movq mm6, mm5 | ||
| 386 : | lea r9, [%2 wrt rip] ; data | ||
| 387 : | pmullw mm4, [r9 ] | ||
| 388 : | pmullw mm5, [r9+ 8] | ||
| 389 : | pmullw mm6, [r9+16] | ||
| 390 : | paddw %3, mm4 | ||
| 391 : | paddw %4, mm5 | ||
| 392 : | paddw %5, mm6 | ||
| 393 : | %endmacro | ||
| 394 : | |||
| 395 : | ;////////////////////////////////////////////////////////////////////// | ||
| 396 : | |||
| 397 : | %macro MIX 3 ; %1:reg, %2:src, %3:rounder | ||
| 398 : | pxor mm6, mm6 | ||
| 399 : | movq mm4, [%2] | ||
| 400 : | movq mm1, %1 | ||
| 401 : | movq mm5, mm4 | ||
| 402 : | punpcklbw %1, mm6 | ||
| 403 : | punpcklbw mm4, mm6 | ||
| 404 : | punpckhbw mm1, mm6 | ||
| 405 : | punpckhbw mm5, mm6 | ||
| 406 : | movq mm6, [%3] ; rounder #2 | ||
| 407 : | paddusw %1, mm4 | ||
| 408 : | paddusw mm1, mm5 | ||
| 409 : | paddusw %1, mm6 | ||
| 410 : | paddusw mm1, mm6 | ||
| 411 : | psrlw %1, 1 | ||
| 412 : | psrlw mm1, 1 | ||
| 413 : | packuswb %1, mm1 | ||
| 414 : | %endmacro | ||
| 415 : | |||
| 416 : | ;////////////////////////////////////////////////////////////////////// | ||
| 417 : | |||
| 418 : | %macro H_PASS_16 2 ; %1:src-op (0=NONE,1=AVRG,2=AVRG-UP), %2:dst-op (NONE/AVRG) | ||
| 419 : | |||
| 420 : | %if (%2==0) && (%1==0) | ||
| 421 : | PROLOG_NO_AVRG | ||
| 422 : | %else | ||
| 423 : | PROLOG_AVRG | ||
| 424 : | %endif | ||
| 425 : | |||
| 426 : | .Loop | ||
| 427 : | |||
| 428 : | ; mm0..mm3 serves as a 4x4 delay line | ||
| 429 : | |||
| 430 : | %ifndef USE_TABLES | ||
| 431 : | |||
| 432 : | LOAD 0, 16 ; special case for 1rst/last pixel | ||
| 433 : | movq mm1, mm7 | ||
| 434 : | movq mm2, mm7 | ||
| 435 : | |||
| 436 : | ACCUM2 1, FIR_R1, mm0, mm1 | ||
| 437 : | ACCUM2 2, FIR_R2, mm0, mm1 | ||
| 438 : | ACCUM2 3, FIR_R3, mm0, mm1 | ||
| 439 : | ACCUM2 4, FIR_R4, mm0, mm1 | ||
| 440 : | |||
| 441 : | ACCUM3 5, FIR_R5, mm0, mm1, mm2 | ||
| 442 : | ACCUM3 6, FIR_R6, mm0, mm1, mm2 | ||
| 443 : | ACCUM3 7, FIR_R7, mm0, mm1, mm2 | ||
| 444 : | ACCUM2 8, FIR_R8, mm1, mm2 | ||
| 445 : | ACCUM3 9, FIR_R9, mm1, mm2, mm3 | ||
| 446 : | ACCUM3 10, FIR_R10,mm1, mm2, mm3 | ||
| 447 : | ACCUM3 11, FIR_R11,mm1, mm2, mm3 | ||
| 448 : | |||
| 449 : | ACCUM2 12, FIR_R12, mm2, mm3 | ||
| 450 : | ACCUM2 13, FIR_R13, mm2, mm3 | ||
| 451 : | ACCUM2 14, FIR_R14, mm2, mm3 | ||
| 452 : | ACCUM2 15, FIR_R15, mm2, mm3 | ||
| 453 : | |||
| 454 : | %else | ||
| 455 : | |||
| 456 : | TLOAD 0, 16 ; special case for 1rst/last pixel | ||
| 457 : | movq mm1, mm7 | ||
| 458 : | movq mm2, mm7 | ||
| 459 : | |||
| 460 : | TACCUM2 1, xvid_FIR_23_19_6_3, xvid_FIR_1_0_0_0 , mm0, mm1 | ||
| 461 : | TACCUM2 2, xvid_FIR_7_20_20_6, xvid_FIR_3_1_0_0 , mm0, mm1 | ||
| 462 : | TACCUM2 3, xvid_FIR_3_6_20_20, xvid_FIR_6_3_1_0 , mm0, mm1 | ||
| 463 : | TACCUM2 4, xvid_FIR_1_3_6_20 , xvid_FIR_20_6_3_1, mm0, mm1 | ||
| 464 : | |||
| 465 : | TACCUM3 5, xvid_FIR_0_1_3_6 , xvid_FIR_20_20_6_3, xvid_FIR_1_0_0_0 , mm0, mm1, mm2 | ||
| 466 : | TACCUM3 6, xvid_FIR_0_0_1_3 , xvid_FIR_6_20_20_6, xvid_FIR_3_1_0_0 , mm0, mm1, mm2 | ||
| 467 : | TACCUM3 7, xvid_FIR_0_0_0_1 , xvid_FIR_3_6_20_20, xvid_FIR_6_3_1_0 , mm0, mm1, mm2 | ||
| 468 : | |||
| 469 : | TACCUM2 8, xvid_FIR_1_3_6_20 , xvid_FIR_20_6_3_1 , mm1, mm2 | ||
| 470 : | |||
| 471 : | TACCUM3 9, xvid_FIR_0_1_3_6 , xvid_FIR_20_20_6_3, xvid_FIR_1_0_0_0, mm1, mm2, mm3 | ||
| 472 : | TACCUM3 10, xvid_FIR_0_0_1_3 , xvid_FIR_6_20_20_6, xvid_FIR_3_1_0_0, mm1, mm2, mm3 | ||
| 473 : | TACCUM3 11, xvid_FIR_0_0_0_1 , xvid_FIR_3_6_20_20, xvid_FIR_6_3_1_0, mm1, mm2, mm3 | ||
| 474 : | |||
| 475 : | TACCUM2 12, xvid_FIR_1_3_6_20, xvid_FIR_20_6_3_1 , mm2, mm3 | ||
| 476 : | TACCUM2 13, xvid_FIR_0_1_3_6 , xvid_FIR_20_20_6_3, mm2, mm3 | ||
| 477 : | TACCUM2 14, xvid_FIR_0_0_1_3 , xvid_FIR_6_20_20_7, mm2, mm3 | ||
| 478 : | TACCUM2 15, xvid_FIR_0_0_0_1 , xvid_FIR_3_6_19_23, mm2, mm3 | ||
| 479 : | |||
| 480 : | %endif | ||
| 481 : | |||
| 482 : | psraw mm0, 5 | ||
| 483 : | psraw mm1, 5 | ||
| 484 : | psraw mm2, 5 | ||
| 485 : | psraw mm3, 5 | ||
| 486 : | packuswb mm0, mm1 | ||
| 487 : | packuswb mm2, mm3 | ||
| 488 : | |||
| 489 : | %if (%1==1) | ||
| 490 : | MIX mm0, rsi, rbx | ||
| 491 : | %elif (%1==2) | ||
| 492 : | MIX mm0, rsi+1, rbx | ||
| 493 : | %endif | ||
| 494 : | %if (%2==1) | ||
| 495 : | lea r9, [Rounder1_MMX wrt rip] | ||
| 496 : | MIX mm0, rdi, r9 | ||
| 497 : | %endif | ||
| 498 : | |||
| 499 : | %if (%1==1) | ||
| 500 : | MIX mm2, rsi+8, rbx | ||
| 501 : | %elif (%1==2) | ||
| 502 : | MIX mm2, rsi+9, rbx | ||
| 503 : | %endif | ||
| 504 : | %if (%2==1) | ||
| 505 : | lea r9, [Rounder1_MMX wrt rip] | ||
| 506 : | MIX mm2, rdi+8, r9 | ||
| 507 : | %endif | ||
| 508 : | |||
| 509 : | lea rsi, [rsi+rbp] | ||
| 510 : | |||
| 511 : | movq [rdi+0], mm0 | ||
| 512 : | movq [rdi+8], mm2 | ||
| 513 : | |||
| 514 : | add rdi, rbp | ||
| 515 : | dec rcx | ||
| 516 : | jg .Loop | ||
| 517 : | |||
| 518 : | %if (%2==0) && (%1==0) | ||
| 519 : | EPILOG_NO_AVRG | ||
| 520 : | %else | ||
| 521 : | EPILOG_AVRG | ||
| 522 : | %endif | ||
| 523 : | |||
| 524 : | %endmacro | ||
| 525 : | |||
| 526 : | |||
| 527 : | ;////////////////////////////////////////////////////////////////////// | ||
| 528 : | |||
| 529 : | %macro H_PASS_8 2 ; %1:src-op (0=NONE,1=AVRG,2=AVRG-UP), %2:dst-op (NONE/AVRG) | ||
| 530 : | |||
| 531 : | %if (%2==0) && (%1==0) | ||
| 532 : | PROLOG_NO_AVRG | ||
| 533 : | %else | ||
| 534 : | PROLOG_AVRG | ||
| 535 : | %endif | ||
| 536 : | |||
| 537 : | .Loop | ||
| 538 : | ; mm0..mm3 serves as a 4x4 delay line | ||
| 539 : | |||
| 540 : | %ifndef USE_TABLES | ||
| 541 : | |||
| 542 : | LOAD 0, 8 ; special case for 1rst/last pixel | ||
| 543 : | ACCUM2 1, FIR_R1, mm0, mm3 | ||
| 544 : | ACCUM2 2, FIR_R2, mm0, mm3 | ||
| 545 : | ACCUM2 3, FIR_R3, mm0, mm3 | ||
| 546 : | ACCUM2 4, FIR_R4, mm0, mm3 | ||
| 547 : | |||
| 548 : | ACCUM2 5, FIR_R13, mm0, mm3 | ||
| 549 : | ACCUM2 6, FIR_R14, mm0, mm3 | ||
| 550 : | ACCUM2 7, FIR_R15, mm0, mm3 | ||
| 551 : | |||
| 552 : | %else | ||
| 553 : | |||
| 554 : | %if 1 ; test with no unrolling | ||
| 555 : | |||
| 556 : | TLOAD 0, 8 ; special case for 1rst/last pixel | ||
| 557 : | TACCUM2 1, xvid_FIR_23_19_6_3, xvid_FIR_1_0_0_0 , mm0, mm3 | ||
| 558 : | TACCUM2 2, xvid_FIR_7_20_20_6, xvid_FIR_3_1_0_0 , mm0, mm3 | ||
| 559 : | TACCUM2 3, xvid_FIR_3_6_20_20, xvid_FIR_6_3_1_0 , mm0, mm3 | ||
| 560 : | TACCUM2 4, xvid_FIR_1_3_6_20 , xvid_FIR_20_6_3_1 , mm0, mm3 | ||
| 561 : | TACCUM2 5, xvid_FIR_0_1_3_6 , xvid_FIR_20_20_6_3, mm0, mm3 | ||
| 562 : | TACCUM2 6, xvid_FIR_0_0_1_3 , xvid_FIR_6_20_20_7, mm0, mm3 | ||
| 563 : | TACCUM2 7, xvid_FIR_0_0_0_1 , xvid_FIR_3_6_19_23, mm0, mm3 | ||
| 564 : | |||
| 565 : | %else ; test with unrolling (little faster, but not much) | ||
| 566 : | |||
| 567 : | movzx rax, byte [rsi] | ||
| 568 : | movzx rdx, byte [rsi+8] | ||
| 569 : | movq mm0, [xvid_FIR_14_3_2_1 + rax*8 ] | ||
| 570 : | movzx rax, byte [rsi+1] | ||
| 571 : | movq mm3, [xvid_FIR_1_2_3_14 + rdx*8 ] | ||
| 572 : | paddw mm0, mm7 | ||
| 573 : | paddw mm3, mm7 | ||
| 574 : | |||
| 575 : | movzx rdx, byte [rsi+2] | ||
| 576 : | lea r9, [xvid_FIR_23_19_6_3 wrt rip] | ||
| 577 : | paddw mm0, [r9 + rax*8] | ||
| 578 : | lea r9, [xvid_FIR_1_0_0_0 wrt rip] | ||
| 579 : | paddw mm3, [r9 + rax*8] | ||
| 580 : | |||
| 581 : | movzx rax, byte [rsi+3] | ||
| 582 : | lea r9, [xvid_FIR_7_20_20_6 wrt rip] | ||
| 583 : | paddw mm0, [r9 + rdx*8] | ||
| 584 : | lea r9, [xvid_FIR_3_1_0_0 wrt rip] | ||
| 585 : | paddw mm3, [r9 + rdx*8] | ||
| 586 : | |||
| 587 : | movzx rdx, byte [rsi+4] | ||
| 588 : | lea r9, [xvid_FIR_3_6_20_20 wrt rip] | ||
| 589 : | paddw mm0, [r9 + rax*8] | ||
| 590 : | lea r9, [xvid_FIR_6_3_1_0 wrt rip] | ||
| 591 : | paddw mm3, [r9 + rax*8] | ||
| 592 : | |||
| 593 : | movzx rax, byte [rsi+5] | ||
| 594 : | lea r9, [xvid_FIR_1_3_6_20 wrt rip] | ||
| 595 : | paddw mm0, [r9 + rdx*8] | ||
| 596 : | lea r9, [xvid_FIR_20_6_3_1 wrt rip] | ||
| 597 : | paddw mm3, [r9 + rdx*8] | ||
| 598 : | |||
| 599 : | movzx rdx, byte [rsi+6] | ||
| 600 : | lea r9, [xvid_FIR_0_1_3_6 wrt rip] | ||
| 601 : | paddw mm0, [r9 + rax*8] | ||
| 602 : | lea r9, [xvid_FIR_20_20_6_3 wrt rip] | ||
| 603 : | paddw mm3, [r9 + rax*8] | ||
| 604 : | |||
| 605 : | movzx rax, byte [rsi+7] | ||
| 606 : | lea r9, [xvid_FIR_0_0_1_3 wrt rip] | ||
| 607 : | paddw mm0, [r9 + rdx*8] | ||
| 608 : | lea r9, [xvid_FIR_6_20_20_7 wrt rip] | ||
| 609 : | paddw mm3, [r9 + rdx*8] | ||
| 610 : | |||
| 611 : | lea r9, [xvid_FIR_0_0_0_1 wrt rip] | ||
| 612 : | paddw mm0, [r9 + rax*8] | ||
| 613 : | lea r9, [xvid_FIR_3_6_19_23 wrt rip] | ||
| 614 : | paddw mm3, [r9 + rax*8] | ||
| 615 : | |||
| 616 : | %endif | ||
| 617 : | |||
| 618 : | %endif ; !USE_TABLES | ||
| 619 : | |||
| 620 : | psraw mm0, 5 | ||
| 621 : | psraw mm3, 5 | ||
| 622 : | packuswb mm0, mm3 | ||
| 623 : | |||
| 624 : | %if (%1==1) | ||
| 625 : | MIX mm0, rsi, rbx | ||
| 626 : | %elif (%1==2) | ||
| 627 : | MIX mm0, rsi+1, rbx | ||
| 628 : | %endif | ||
| 629 : | %if (%2==1) | ||
| 630 : | lea r9, [Rounder1_MMX wrt rip] | ||
| 631 : | MIX mm0, rdi, r9 | ||
| 632 : | %endif | ||
| 633 : | |||
| 634 : | movq [rdi], mm0 | ||
| 635 : | |||
| 636 : | add rdi, rbp | ||
| 637 : | add rsi, rbp | ||
| 638 : | dec rcx | ||
| 639 : | jg .Loop | ||
| 640 : | |||
| 641 : | %if (%2==0) && (%1==0) | ||
| 642 : | EPILOG_NO_AVRG | ||
| 643 : | %else | ||
| 644 : | EPILOG_AVRG | ||
| 645 : | %endif | ||
| 646 : | |||
| 647 : | %endmacro | ||
| 648 : | |||
| 649 : | ;////////////////////////////////////////////////////////////////////// | ||
| 650 : | ;// 16x? copy Functions | ||
| 651 : | |||
| 652 : | xvid_H_Pass_16_x86_64: | ||
| 653 : | H_PASS_16 0, 0 | ||
| 654 : | .endfunc | ||
| 655 : | xvid_H_Pass_Avrg_16_x86_64: | ||
| 656 : | H_PASS_16 1, 0 | ||
| 657 : | .endfunc | ||
| 658 : | xvid_H_Pass_Avrg_Up_16_x86_64: | ||
| 659 : | H_PASS_16 2, 0 | ||
| 660 : | .endfunc | ||
| 661 : | |||
| 662 : | ;////////////////////////////////////////////////////////////////////// | ||
| 663 : | ;// 8x? copy Functions | ||
| 664 : | |||
| 665 : | xvid_H_Pass_8_x86_64: | ||
| 666 : | H_PASS_8 0, 0 | ||
| 667 : | .endfunc | ||
| 668 : | xvid_H_Pass_Avrg_8_x86_64: | ||
| 669 : | H_PASS_8 1, 0 | ||
| 670 : | .endfunc | ||
| 671 : | xvid_H_Pass_Avrg_Up_8_x86_64: | ||
| 672 : | H_PASS_8 2, 0 | ||
| 673 : | .endfunc | ||
| 674 : | |||
| 675 : | ;////////////////////////////////////////////////////////////////////// | ||
| 676 : | ;// 16x? avrg Functions | ||
| 677 : | |||
| 678 : | xvid_H_Pass_Add_16_x86_64: | ||
| 679 : | H_PASS_16 0, 1 | ||
| 680 : | .endfunc | ||
| 681 : | xvid_H_Pass_Avrg_Add_16_x86_64: | ||
| 682 : | H_PASS_16 1, 1 | ||
| 683 : | .endfunc | ||
| 684 : | xvid_H_Pass_Avrg_Up_Add_16_x86_64: | ||
| 685 : | H_PASS_16 2, 1 | ||
| 686 : | .endfunc | ||
| 687 : | |||
| 688 : | ;////////////////////////////////////////////////////////////////////// | ||
| 689 : | ;// 8x? avrg Functions | ||
| 690 : | |||
| 691 : | xvid_H_Pass_8_Add_x86_64: | ||
| 692 : | H_PASS_8 0, 1 | ||
| 693 : | .endfunc | ||
| 694 : | xvid_H_Pass_Avrg_8_Add_x86_64: | ||
| 695 : | H_PASS_8 1, 1 | ||
| 696 : | .endfunc | ||
| 697 : | xvid_H_Pass_Avrg_Up_8_Add_x86_64: | ||
| 698 : | H_PASS_8 2, 1 | ||
| 699 : | .endfunc | ||
| 700 : | |||
| 701 : | |||
| 702 : | ;////////////////////////////////////////////////////////////////////// | ||
| 703 : | ;// | ||
| 704 : | ;// All vertical passes | ||
| 705 : | ;// | ||
| 706 : | ;////////////////////////////////////////////////////////////////////// | ||
| 707 : | |||
| 708 : | %macro V_LOAD 1 ; %1=Last? | ||
| 709 : | |||
| 710 : | movd mm4, [rdx] | ||
| 711 : | pxor mm6, mm6 | ||
| 712 : | %if (%1==0) | ||
| 713 : | add rdx, rbp | ||
| 714 : | %endif | ||
| 715 : | punpcklbw mm4, mm6 | ||
| 716 : | |||
| 717 : | %endmacro | ||
| 718 : | |||
| 719 : | %macro V_ACC1 2 ; %1:reg; 2:tap | ||
| 720 : | pmullw mm4, [%2] | ||
| 721 : | paddw %1, mm4 | ||
| 722 : | %endmacro | ||
| 723 : | |||
| 724 : | %macro V_ACC2 4 ; %1-%2: regs, %3-%4: taps | ||
| 725 : | movq mm5, mm4 | ||
| 726 : | movq mm6, mm4 | ||
| 727 : | pmullw mm5, [%3] | ||
| 728 : | pmullw mm6, [%4] | ||
| 729 : | paddw %1, mm5 | ||
| 730 : | paddw %2, mm6 | ||
| 731 : | %endmacro | ||
| 732 : | |||
| 733 : | %macro V_ACC2l 4 ; %1-%2: regs, %3-%4: taps | ||
| 734 : | movq mm5, mm4 | ||
| 735 : | pmullw mm5, [%3] | ||
| 736 : | pmullw mm4, [%4] | ||
| 737 : | paddw %1, mm5 | ||
| 738 : | paddw %2, mm4 | ||
| 739 : | %endmacro | ||
| 740 : | |||
| 741 : | %macro V_ACC4 8 ; %1-%4: regs, %5-%8: taps | ||
| 742 : | V_ACC2 %1,%2, %5,%6 | ||
| 743 : | V_ACC2l %3,%4, %7,%8 | ||
| 744 : | %endmacro | ||
| 745 : | |||
| 746 : | |||
| 747 : | %macro V_MIX 3 ; %1:dst-reg, %2:src, %3: rounder | ||
| 748 : | pxor mm6, mm6 | ||
| 749 : | movq mm4, [%2] | ||
| 750 : | punpcklbw %1, mm6 | ||
| 751 : | punpcklbw mm4, mm6 | ||
| 752 : | paddusw %1, mm4 | ||
| 753 : | paddusw %1, [%3] | ||
| 754 : | psrlw %1, 1 | ||
| 755 : | packuswb %1, %1 | ||
| 756 : | %endmacro | ||
| 757 : | |||
| 758 : | %macro V_STORE 4 ; %1-%2: mix ops, %3: reg, %4:last? | ||
| 759 : | |||
| 760 : | psraw %3, 5 | ||
| 761 : | packuswb %3, %3 | ||
| 762 : | |||
| 763 : | %if (%1==1) | ||
| 764 : | V_MIX %3, rsi, rbx | ||
| 765 : | add rsi, rbp | ||
| 766 : | %elif (%1==2) | ||
| 767 : | add rsi, rbp | ||
| 768 : | V_MIX %3, rsi, rbx | ||
| 769 : | %endif | ||
| 770 : | %if (%2==1) | ||
| 771 : | lea r9, [Rounder1_MMX wrt rip] | ||
| 772 : | V_MIX %3, rdi, r9 | ||
| 773 : | %endif | ||
| 774 : | |||
| 775 : | edgomez | 1587 | movd eax, %3 |
| 776 : | mov [rdi], eax | ||
| 777 : | edgomez | 1586 | |
| 778 : | %if (%4==0) | ||
| 779 : | add rdi, rbp | ||
| 780 : | %endif | ||
| 781 : | |||
| 782 : | %endmacro | ||
| 783 : | |||
| 784 : | ;////////////////////////////////////////////////////////////////////// | ||
| 785 : | |||
| 786 : | %macro V_PASS_16 2 ; %1:src-op (0=NONE,1=AVRG,2=AVRG-UP), %2:dst-op (NONE/AVRG) | ||
| 787 : | |||
| 788 : | %if (%2==0) && (%1==0) | ||
| 789 : | PROLOG_NO_AVRG | ||
| 790 : | %else | ||
| 791 : | PROLOG_AVRG | ||
| 792 : | %endif | ||
| 793 : | |||
| 794 : | ; we process one stripe of 4x16 pixel each time. | ||
| 795 : | ; the size (3rd argument) is meant to be a multiple of 4 | ||
| 796 : | ; mm0..mm3 serves as a 4x4 delay line | ||
| 797 : | |||
| 798 : | .Loop | ||
| 799 : | |||
| 800 : | push rdi | ||
| 801 : | push rsi ; esi is preserved for src-mixing | ||
| 802 : | mov rdx, rsi ; rsi is Src | ||
| 803 : | |||
| 804 : | ; ouput rows [0..3], from input rows [0..8] | ||
| 805 : | |||
| 806 : | movq mm0, mm7 | ||
| 807 : | movq mm1, mm7 | ||
| 808 : | movq mm2, mm7 | ||
| 809 : | movq mm3, mm7 | ||
| 810 : | |||
| 811 : | V_LOAD 0 ; add rdx, rbp, rbp is Bps | ||
| 812 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C14 wrt rip, FIR_Cm3 wrt rip, FIR_C2 wrt rip, FIR_Cm1 wrt rip | ||
| 813 : | V_LOAD 0 | ||
| 814 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C23 wrt rip, FIR_C19 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 815 : | V_LOAD 0 | ||
| 816 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm7 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip | ||
| 817 : | V_LOAD 0 | ||
| 818 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip | ||
| 819 : | V_LOAD 0 | ||
| 820 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip | ||
| 821 : | V_STORE %1, %2, mm0, 0 | ||
| 822 : | |||
| 823 : | V_LOAD 0 | ||
| 824 : | V_ACC2 mm1, mm2, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 825 : | V_ACC1 mm3, FIR_Cm6 wrt rip | ||
| 826 : | V_STORE %1, %2, mm1, 0 | ||
| 827 : | |||
| 828 : | V_LOAD 0 | ||
| 829 : | V_ACC2l mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 830 : | V_STORE %1, %2, mm2, 0 | ||
| 831 : | |||
| 832 : | V_LOAD 1 | ||
| 833 : | V_ACC1 mm3, FIR_Cm1 wrt rip | ||
| 834 : | V_STORE %1, %2, mm3, 0 | ||
| 835 : | |||
| 836 : | ; ouput rows [4..7], from input rows [1..11] (!!) | ||
| 837 : | |||
| 838 : | ; mov esi, [esp] | ||
| 839 : | mov rsi, [rsp] ; rsi on stack... | ||
| 840 : | lea rdx, [rsi+rbp] | ||
| 841 : | |||
| 842 : | lea rsi, [rsi+4*rbp] ; for src-mixing | ||
| 843 : | push rsi ; this will be the new value for next round | ||
| 844 : | |||
| 845 : | movq mm0, mm7 | ||
| 846 : | movq mm1, mm7 | ||
| 847 : | movq mm2, mm7 | ||
| 848 : | movq mm3, mm7 | ||
| 849 : | |||
| 850 : | V_LOAD 0 | ||
| 851 : | V_ACC1 mm0, FIR_Cm1 wrt rip | ||
| 852 : | |||
| 853 : | V_LOAD 0 | ||
| 854 : | V_ACC2l mm0, mm1, FIR_C3 wrt rip, FIR_Cm1 wrt rip | ||
| 855 : | |||
| 856 : | V_LOAD 0 | ||
| 857 : | V_ACC2 mm0, mm1, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 858 : | V_ACC1 mm2, FIR_Cm1 wrt rip | ||
| 859 : | |||
| 860 : | V_LOAD 0 | ||
| 861 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C20 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip, FIR_Cm1 wrt rip | ||
| 862 : | V_LOAD 0 | ||
| 863 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 864 : | V_LOAD 0 | ||
| 865 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip | ||
| 866 : | V_LOAD 0 | ||
| 867 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip | ||
| 868 : | V_LOAD 0 | ||
| 869 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip | ||
| 870 : | V_STORE %1, %2, mm0, 0 | ||
| 871 : | |||
| 872 : | V_LOAD 0 | ||
| 873 : | V_ACC2 mm1, mm2, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 874 : | V_ACC1 mm3, FIR_Cm6 wrt rip | ||
| 875 : | V_STORE %1, %2, mm1, 0 | ||
| 876 : | |||
| 877 : | V_LOAD 0 | ||
| 878 : | V_ACC2l mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 879 : | V_STORE %1, %2, mm2, 0 | ||
| 880 : | |||
| 881 : | V_LOAD 1 | ||
| 882 : | V_ACC1 mm3, FIR_Cm1 wrt rip | ||
| 883 : | V_STORE %1, %2, mm3, 0 | ||
| 884 : | |||
| 885 : | ; ouput rows [8..11], from input rows [5..15] | ||
| 886 : | |||
| 887 : | pop rsi | ||
| 888 : | lea rdx, [rsi+rbp] | ||
| 889 : | |||
| 890 : | lea rsi, [rsi+4*rbp] ; for src-mixing | ||
| 891 : | push rsi ; this will be the new value for next round | ||
| 892 : | |||
| 893 : | movq mm0, mm7 | ||
| 894 : | movq mm1, mm7 | ||
| 895 : | movq mm2, mm7 | ||
| 896 : | movq mm3, mm7 | ||
| 897 : | |||
| 898 : | V_LOAD 0 | ||
| 899 : | V_ACC1 mm0, FIR_Cm1 wrt rip | ||
| 900 : | |||
| 901 : | V_LOAD 0 | ||
| 902 : | V_ACC2l mm0, mm1, FIR_C3 wrt rip, FIR_Cm1 wrt rip | ||
| 903 : | |||
| 904 : | V_LOAD 0 | ||
| 905 : | V_ACC2 mm0, mm1, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 906 : | V_ACC1 mm2, FIR_Cm1 wrt rip | ||
| 907 : | |||
| 908 : | V_LOAD 0 | ||
| 909 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C20 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip, FIR_Cm1 wrt rip | ||
| 910 : | V_LOAD 0 | ||
| 911 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 912 : | V_LOAD 0 | ||
| 913 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip | ||
| 914 : | V_LOAD 0 | ||
| 915 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip | ||
| 916 : | V_LOAD 0 | ||
| 917 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip | ||
| 918 : | |||
| 919 : | V_STORE %1, %2, mm0, 0 | ||
| 920 : | |||
| 921 : | V_LOAD 0 | ||
| 922 : | V_ACC2 mm1, mm2, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 923 : | V_ACC1 mm3, FIR_Cm6 wrt rip | ||
| 924 : | V_STORE %1, %2, mm1, 0 | ||
| 925 : | |||
| 926 : | V_LOAD 0 | ||
| 927 : | V_ACC2l mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 928 : | V_STORE %1, %2, mm2, 0 | ||
| 929 : | |||
| 930 : | V_LOAD 1 | ||
| 931 : | V_ACC1 mm3, FIR_Cm1 wrt rip | ||
| 932 : | V_STORE %1, %2, mm3, 0 | ||
| 933 : | |||
| 934 : | |||
| 935 : | ; ouput rows [12..15], from input rows [9.16] | ||
| 936 : | |||
| 937 : | pop rsi | ||
| 938 : | lea rdx, [rsi+rbp] | ||
| 939 : | |||
| 940 : | %if (%1!=0) | ||
| 941 : | lea rsi, [rsi+4*rbp] ; for src-mixing | ||
| 942 : | %endif | ||
| 943 : | |||
| 944 : | movq mm0, mm7 | ||
| 945 : | movq mm1, mm7 | ||
| 946 : | movq mm2, mm7 | ||
| 947 : | movq mm3, mm7 | ||
| 948 : | |||
| 949 : | V_LOAD 0 | ||
| 950 : | V_ACC1 mm3, FIR_Cm1 wrt rip | ||
| 951 : | |||
| 952 : | V_LOAD 0 | ||
| 953 : | V_ACC2l mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 954 : | |||
| 955 : | V_LOAD 0 | ||
| 956 : | V_ACC2 mm1, mm2, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 957 : | V_ACC1 mm3, FIR_Cm6 wrt rip | ||
| 958 : | |||
| 959 : | V_LOAD 0 | ||
| 960 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip | ||
| 961 : | V_LOAD 0 | ||
| 962 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip | ||
| 963 : | V_LOAD 0 | ||
| 964 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm7 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip | ||
| 965 : | V_LOAD 0 | ||
| 966 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C23 wrt rip, FIR_C19 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 967 : | V_LOAD 1 | ||
| 968 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C14 wrt rip, FIR_Cm3 wrt rip, FIR_C2 wrt rip, FIR_Cm1 wrt rip | ||
| 969 : | |||
| 970 : | V_STORE %1, %2, mm3, 0 | ||
| 971 : | V_STORE %1, %2, mm2, 0 | ||
| 972 : | V_STORE %1, %2, mm1, 0 | ||
| 973 : | V_STORE %1, %2, mm0, 1 | ||
| 974 : | |||
| 975 : | ; ... next 4 columns | ||
| 976 : | |||
| 977 : | pop rsi | ||
| 978 : | pop rdi | ||
| 979 : | add rsi, 4 | ||
| 980 : | add rdi, 4 | ||
| 981 : | sub rcx, 4 | ||
| 982 : | jg .Loop | ||
| 983 : | |||
| 984 : | %if (%2==0) && (%1==0) | ||
| 985 : | EPILOG_NO_AVRG | ||
| 986 : | %else | ||
| 987 : | EPILOG_AVRG | ||
| 988 : | %endif | ||
| 989 : | |||
| 990 : | %endmacro | ||
| 991 : | |||
| 992 : | ;////////////////////////////////////////////////////////////////////// | ||
| 993 : | |||
| 994 : | %macro V_PASS_8 2 ; %1:src-op (0=NONE,1=AVRG,2=AVRG-UP), %2:dst-op (NONE/AVRG) | ||
| 995 : | |||
| 996 : | %if (%2==0) && (%1==0) | ||
| 997 : | PROLOG_NO_AVRG | ||
| 998 : | %else | ||
| 999 : | PROLOG_AVRG | ||
| 1000 : | %endif | ||
| 1001 : | |||
| 1002 : | ; we process one stripe of 4x8 pixel each time | ||
| 1003 : | ; the size (3rd argument) is meant to be a multiple of 4 | ||
| 1004 : | ; mm0..mm3 serves as a 4x4 delay line | ||
| 1005 : | .Loop | ||
| 1006 : | |||
| 1007 : | push rdi | ||
| 1008 : | push rsi ; esi is preserved for src-mixing | ||
| 1009 : | mov rdx, rsi | ||
| 1010 : | |||
| 1011 : | ; ouput rows [0..3], from input rows [0..8] | ||
| 1012 : | |||
| 1013 : | movq mm0, mm7 | ||
| 1014 : | movq mm1, mm7 | ||
| 1015 : | movq mm2, mm7 | ||
| 1016 : | movq mm3, mm7 | ||
| 1017 : | |||
| 1018 : | V_LOAD 0 | ||
| 1019 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C14 wrt rip, FIR_Cm3 wrt rip, FIR_C2 wrt rip, FIR_Cm1 wrt rip | ||
| 1020 : | V_LOAD 0 | ||
| 1021 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C23 wrt rip, FIR_C19 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 1022 : | V_LOAD 0 | ||
| 1023 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm7 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip | ||
| 1024 : | V_LOAD 0 | ||
| 1025 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip | ||
| 1026 : | V_LOAD 0 | ||
| 1027 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip | ||
| 1028 : | V_STORE %1, %2, mm0, 0 | ||
| 1029 : | |||
| 1030 : | V_LOAD 0 | ||
| 1031 : | V_ACC2 mm1, mm2, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 1032 : | V_ACC1 mm3, FIR_Cm6 wrt rip | ||
| 1033 : | |||
| 1034 : | V_STORE %1, %2, mm1, 0 | ||
| 1035 : | |||
| 1036 : | V_LOAD 0 | ||
| 1037 : | V_ACC2l mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 1038 : | V_STORE %1, %2, mm2, 0 | ||
| 1039 : | |||
| 1040 : | V_LOAD 1 | ||
| 1041 : | V_ACC1 mm3, FIR_Cm1 wrt rip | ||
| 1042 : | V_STORE %1, %2, mm3, 0 | ||
| 1043 : | |||
| 1044 : | ; ouput rows [4..7], from input rows [1..9] | ||
| 1045 : | |||
| 1046 : | mov rsi, [rsp] ; rsi on stack... | ||
| 1047 : | lea rdx, [rsi+rbp] | ||
| 1048 : | |||
| 1049 : | %if (%1!=0) | ||
| 1050 : | lea rsi, [rsi+4*rbp] ; for src-mixing | ||
| 1051 : | %endif | ||
| 1052 : | |||
| 1053 : | movq mm0, mm7 | ||
| 1054 : | movq mm1, mm7 | ||
| 1055 : | movq mm2, mm7 | ||
| 1056 : | movq mm3, mm7 | ||
| 1057 : | |||
| 1058 : | V_LOAD 0 | ||
| 1059 : | V_ACC1 mm3, FIR_Cm1 wrt rip | ||
| 1060 : | |||
| 1061 : | V_LOAD 0 | ||
| 1062 : | V_ACC2l mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 1063 : | |||
| 1064 : | V_LOAD 0 | ||
| 1065 : | V_ACC2 mm1, mm2, FIR_Cm1 wrt rip, FIR_C3 wrt rip | ||
| 1066 : | V_ACC1 mm3, FIR_Cm6 wrt rip | ||
| 1067 : | |||
| 1068 : | V_LOAD 0 | ||
| 1069 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm1 wrt rip, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip | ||
| 1070 : | V_LOAD 0 | ||
| 1071 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C3 wrt rip, FIR_Cm6 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip | ||
| 1072 : | V_LOAD 0 | ||
| 1073 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_Cm7 wrt rip, FIR_C20 wrt rip, FIR_C20 wrt rip, FIR_Cm6 wrt rip | ||
| 1074 : | V_LOAD 0 | ||
| 1075 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C23 wrt rip, FIR_C19 wrt rip, FIR_Cm6 wrt rip, FIR_C3 wrt rip | ||
| 1076 : | V_LOAD 1 | ||
| 1077 : | V_ACC4 mm0, mm1, mm2, mm3, FIR_C14 wrt rip, FIR_Cm3 wrt rip, FIR_C2 wrt rip, FIR_Cm1 wrt rip | ||
| 1078 : | |||
| 1079 : | V_STORE %1, %2, mm3, 0 | ||
| 1080 : | V_STORE %1, %2, mm2, 0 | ||
| 1081 : | V_STORE %1, %2, mm1, 0 | ||
| 1082 : | V_STORE %1, %2, mm0, 1 | ||
| 1083 : | |||
| 1084 : | ; ... next 4 columns | ||
| 1085 : | |||
| 1086 : | pop rsi | ||
| 1087 : | pop rdi | ||
| 1088 : | add rsi, 4 | ||
| 1089 : | add rdi, 4 | ||
| 1090 : | sub rcx, 4 | ||
| 1091 : | jg .Loop | ||
| 1092 : | |||
| 1093 : | %if (%2==0) && (%1==0) | ||
| 1094 : | EPILOG_NO_AVRG | ||
| 1095 : | %else | ||
| 1096 : | EPILOG_AVRG | ||
| 1097 : | %endif | ||
| 1098 : | |||
| 1099 : | %endmacro | ||
| 1100 : | |||
| 1101 : | |||
| 1102 : | ;////////////////////////////////////////////////////////////////////// | ||
| 1103 : | ;// 16x? copy Functions | ||
| 1104 : | |||
| 1105 : | xvid_V_Pass_16_x86_64: | ||
| 1106 : | V_PASS_16 0, 0 | ||
| 1107 : | .endfunc | ||
| 1108 : | xvid_V_Pass_Avrg_16_x86_64: | ||
| 1109 : | V_PASS_16 1, 0 | ||
| 1110 : | .endfunc | ||
| 1111 : | xvid_V_Pass_Avrg_Up_16_x86_64: | ||
| 1112 : | V_PASS_16 2, 0 | ||
| 1113 : | .endfunc | ||
| 1114 : | |||
| 1115 : | ;////////////////////////////////////////////////////////////////////// | ||
| 1116 : | ;// 8x? copy Functions | ||
| 1117 : | |||
| 1118 : | xvid_V_Pass_8_x86_64: | ||
| 1119 : | V_PASS_8 0, 0 | ||
| 1120 : | .endfunc | ||
| 1121 : | xvid_V_Pass_Avrg_8_x86_64: | ||
| 1122 : | V_PASS_8 1, 0 | ||
| 1123 : | .endfunc | ||
| 1124 : | xvid_V_Pass_Avrg_Up_8_x86_64: | ||
| 1125 : | V_PASS_8 2, 0 | ||
| 1126 : | .endfunc | ||
| 1127 : | |||
| 1128 : | ;////////////////////////////////////////////////////////////////////// | ||
| 1129 : | ;// 16x? avrg Functions | ||
| 1130 : | |||
| 1131 : | xvid_V_Pass_Add_16_x86_64: | ||
| 1132 : | V_PASS_16 0, 1 | ||
| 1133 : | .endfunc | ||
| 1134 : | xvid_V_Pass_Avrg_Add_16_x86_64: | ||
| 1135 : | V_PASS_16 1, 1 | ||
| 1136 : | .endfunc | ||
| 1137 : | xvid_V_Pass_Avrg_Up_Add_16_x86_64: | ||
| 1138 : | V_PASS_16 2, 1 | ||
| 1139 : | .endfunc | ||
| 1140 : | |||
| 1141 : | ;////////////////////////////////////////////////////////////////////// | ||
| 1142 : | ;// 8x? avrg Functions | ||
| 1143 : | |||
| 1144 : | xvid_V_Pass_8_Add_x86_64: | ||
| 1145 : | V_PASS_8 0, 1 | ||
| 1146 : | .endfunc | ||
| 1147 : | xvid_V_Pass_Avrg_8_Add_x86_64: | ||
| 1148 : | V_PASS_8 1, 1 | ||
| 1149 : | .endfunc | ||
| 1150 : | xvid_V_Pass_Avrg_Up_8_Add_x86_64: | ||
| 1151 : | V_PASS_8 2, 1 | ||
| 1152 : | .endfunc | ||
| 1153 : | |||
| 1154 : | ;////////////////////////////////////////////////////////////////////// |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |