Parent Directory
| 
Revision Log
Revision 1793 - (view) (download)
| 1 : | edgomez | 851 | ;/***************************************************************************** |
| 2 : | ; * | ||
| 3 : | ; * XVID MPEG-4 VIDEO CODEC | ||
| 4 : | edgomez | 1382 | ; * - Reduced-Resolution utilities - |
| 5 : | edgomez | 851 | ; * |
| 6 : | ; * Copyright(C) 2002 Pascal Massimino <skal@planet-d.net> | ||
| 7 : | ; * | ||
| 8 : | ; * XviD is free software; you can redistribute it and/or modify it | ||
| 9 : | ; * under the terms of the GNU General Public License as published by | ||
| 10 : | ; * the Free Software Foundation; either version 2 of the License, or | ||
| 11 : | ; * (at your option) any later version. | ||
| 12 : | ; * | ||
| 13 : | ; * This program 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 | ||
| 16 : | ; * GNU General Public License for more details. | ||
| 17 : | ; * | ||
| 18 : | ; * You should have received a copy of the GNU General Public License | ||
| 19 : | ; * along with this program; if not, write to the Free Software | ||
| 20 : | ; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
| 21 : | ; * | ||
| 22 : | Isibaar | 1793 | ; * $Id: reduced_mmx.asm,v 1.8 2008-11-11 20:46:24 Isibaar Exp $ |
| 23 : | edgomez | 851 | ; * |
| 24 : | ; *************************************************************************/ | ||
| 25 : | |||
| 26 : | edgomez | 1382 | BITS 32 |
| 27 : | edgomez | 851 | |
| 28 : | edgomez | 1382 | %macro cglobal 1 |
| 29 : | edgomez | 851 | %ifdef PREFIX |
| 30 : | edgomez | 1535 | %ifdef MARK_FUNCS |
| 31 : | edgomez | 1540 | global _%1:function %1.endfunc-%1 |
| 32 : | %define %1 _%1:function %1.endfunc-%1 | ||
| 33 : | Isibaar | 1793 | %define ENDFUNC .endfunc |
| 34 : | edgomez | 1535 | %else |
| 35 : | global _%1 | ||
| 36 : | %define %1 _%1 | ||
| 37 : | Isibaar | 1793 | %define ENDFUNC |
| 38 : | edgomez | 1535 | %endif |
| 39 : | edgomez | 851 | %else |
| 40 : | edgomez | 1535 | %ifdef MARK_FUNCS |
| 41 : | edgomez | 1540 | global %1:function %1.endfunc-%1 |
| 42 : | Isibaar | 1793 | %define ENDFUNC .endfunc |
| 43 : | edgomez | 1535 | %else |
| 44 : | global %1 | ||
| 45 : | Isibaar | 1793 | %define ENDFUNC |
| 46 : | edgomez | 1535 | %endif |
| 47 : | edgomez | 851 | %endif |
| 48 : | %endmacro | ||
| 49 : | |||
| 50 : | ;=========================================================================== | ||
| 51 : | |||
| 52 : | edgomez | 1382 | %ifdef FORMAT_COFF |
| 53 : | edgomez | 1519 | SECTION .rodata |
| 54 : | edgomez | 1382 | %else |
| 55 : | edgomez | 1519 | SECTION .rodata align=16 |
| 56 : | edgomez | 1382 | %endif |
| 57 : | edgomez | 851 | |
| 58 : | align 16 | ||
| 59 : | Up31 dw 3, 1, 3, 1 | ||
| 60 : | Up13 dw 1, 3, 1, 3 | ||
| 61 : | Up93 dw 9, 3, 9, 3 | ||
| 62 : | Up39 dw 3, 9, 3, 9 | ||
| 63 : | Cst0 dw 0, 0, 0, 0 | ||
| 64 : | Cst2 dw 2, 2, 2, 2 | ||
| 65 : | Cst3 dw 3, 3, 3, 3 | ||
| 66 : | Cst32 dw 32,32,32,32 | ||
| 67 : | Cst2000 dw 2, 0, 0, 0 | ||
| 68 : | Cst0002 dw 0, 0, 0, 2 | ||
| 69 : | |||
| 70 : | Mask_ff dw 0xff,0xff,0xff,0xff | ||
| 71 : | |||
| 72 : | ;=========================================================================== | ||
| 73 : | |||
| 74 : | edgomez | 1382 | SECTION .text |
| 75 : | edgomez | 851 | |
| 76 : | cglobal xvid_Copy_Upsampled_8x8_16To8_mmx | ||
| 77 : | cglobal xvid_Add_Upsampled_8x8_16To8_mmx | ||
| 78 : | cglobal xvid_Copy_Upsampled_8x8_16To8_xmm | ||
| 79 : | cglobal xvid_Add_Upsampled_8x8_16To8_xmm | ||
| 80 : | |||
| 81 : | cglobal xvid_HFilter_31_mmx | ||
| 82 : | cglobal xvid_VFilter_31_x86 | ||
| 83 : | cglobal xvid_HFilter_31_x86 | ||
| 84 : | |||
| 85 : | cglobal xvid_Filter_18x18_To_8x8_mmx | ||
| 86 : | cglobal xvid_Filter_Diff_18x18_To_8x8_mmx | ||
| 87 : | |||
| 88 : | |||
| 89 : | ;////////////////////////////////////////////////////////////////////// | ||
| 90 : | ;// 8x8 -> 16x16 upsampling (16b) | ||
| 91 : | ;////////////////////////////////////////////////////////////////////// | ||
| 92 : | |||
| 93 : | %macro MUL_PACK 4 ; %1/%2: regs %3/%4/%5: Up13/Up31 | ||
| 94 : | pmullw %1, %3 ; [Up13] | ||
| 95 : | pmullw mm4, %4 ; [Up31] | ||
| 96 : | pmullw %2, %3 ; [Up13] | ||
| 97 : | pmullw mm5, %4 ; [Up31] | ||
| 98 : | paddsw %1, [Cst2] | ||
| 99 : | paddsw %2, [Cst2] | ||
| 100 : | paddsw %1, mm4 | ||
| 101 : | paddsw %2, mm5 | ||
| 102 : | %endmacro | ||
| 103 : | |||
| 104 : | ; MMX-way of reordering columns... | ||
| 105 : | |||
| 106 : | %macro COL03 3 ;%1/%2: regs, %3: row -output: mm4/mm5 | ||
| 107 : | movq %1, [edx+%3*16+0*2] ; %1 = 0|1|2|3 | ||
| 108 : | movq %2,[edx+%3*16+1*2] ; %2 = 1|2|3|4 | ||
| 109 : | movq mm5, %1 ; mm5 = 0|1|2|3 | ||
| 110 : | movq mm4, %1 ; mm4 = 0|1|2|3 | ||
| 111 : | punpckhwd mm5,%2 ; mm5 = 2|3|3|4 | ||
| 112 : | punpcklwd mm4,%2 ; mm4 = 0|1|1|2 | ||
| 113 : | punpcklwd %1,%1 ; %1 = 0|0|1|1 | ||
| 114 : | punpcklwd %2, mm5 ; %2 = 1|2|2|3 | ||
| 115 : | punpcklwd %1, mm4 ; %1 = 0|0|0|1 | ||
| 116 : | %endmacro | ||
| 117 : | |||
| 118 : | %macro COL47 3 ;%1-%2: regs, %3: row -output: mm4/mm5 | ||
| 119 : | movq mm5, [edx+%3*16+4*2] ; mm5 = 4|5|6|7 | ||
| 120 : | movq %1, [edx+%3*16+3*2] ; %1 = 3|4|5|6 | ||
| 121 : | movq %2, mm5 ; %2 = 4|5|6|7 | ||
| 122 : | movq mm4, mm5 ; mm4 = 4|5|6|7 | ||
| 123 : | punpckhwd %2, %2 ; %2 = 6|6|7|7 | ||
| 124 : | punpckhwd mm5, %2 ; mm5 = 6|7|7|7 | ||
| 125 : | movq %2, %1 ; %2 = 3|4|5|6 | ||
| 126 : | punpcklwd %1, mm4 ; %1 = 3|4|4|5 | ||
| 127 : | punpckhwd %2, mm4 ; %2 = 5|6|6|7 | ||
| 128 : | punpcklwd mm4, %2 ; mm4 = 4|5|5|6 | ||
| 129 : | %endmacro | ||
| 130 : | |||
| 131 : | %macro MIX_ROWS 4 ; %1/%2:prev %3/4:cur (preserved) mm4/mm5: output | ||
| 132 : | ; we need to perform: (%1,%3) -> (%1 = 3*%1+%3, mm4 = 3*%3+%1), %3 preserved. | ||
| 133 : | movq mm4, [Cst3] | ||
| 134 : | movq mm5, [Cst3] | ||
| 135 : | pmullw mm4, %3 | ||
| 136 : | pmullw mm5, %4 | ||
| 137 : | paddsw mm4, %1 | ||
| 138 : | paddsw mm5, %2 | ||
| 139 : | pmullw %1, [Cst3] | ||
| 140 : | pmullw %2, [Cst3] | ||
| 141 : | paddsw %1, %3 | ||
| 142 : | paddsw %2, %4 | ||
| 143 : | %endmacro | ||
| 144 : | |||
| 145 : | ;=========================================================================== | ||
| 146 : | ; | ||
| 147 : | edgomez | 1382 | ; void xvid_Copy_Upsampled_8x8_16To8_mmx(uint8_t *Dst, |
| 148 : | edgomez | 851 | ; const int16_t *Src, const int BpS); |
| 149 : | ; | ||
| 150 : | ;=========================================================================== | ||
| 151 : | |||
| 152 : | edgomez | 1382 | ; Note: we can use ">>2" instead of "/4" here, since we |
| 153 : | edgomez | 851 | ; are (supposed to be) averaging positive values |
| 154 : | |||
| 155 : | %macro STORE_1 2 | ||
| 156 : | psraw %1, 2 | ||
| 157 : | psraw %2, 2 | ||
| 158 : | edgomez | 1382 | packuswb %1,%2 |
| 159 : | edgomez | 851 | movq [ecx], %1 |
| 160 : | %endmacro | ||
| 161 : | |||
| 162 : | %macro STORE_2 2 ; pack and store (%1,%2) + (mm4,mm5) | ||
| 163 : | psraw %1, 4 | ||
| 164 : | psraw %2, 4 | ||
| 165 : | psraw mm4, 4 | ||
| 166 : | psraw mm5, 4 | ||
| 167 : | packuswb %1,%2 | ||
| 168 : | packuswb mm4, mm5 | ||
| 169 : | movq [ecx], %1 | ||
| 170 : | movq [ecx+eax], mm4 | ||
| 171 : | lea ecx, [ecx+2*eax] | ||
| 172 : | %endmacro | ||
| 173 : | |||
| 174 : | ;////////////////////////////////////////////////////////////////////// | ||
| 175 : | |||
| 176 : | align 16 | ||
| 177 : | xvid_Copy_Upsampled_8x8_16To8_mmx: ; 344c | ||
| 178 : | |||
| 179 : | mov ecx, [esp+4] ; Dst | ||
| 180 : | mov edx, [esp+8] ; Src | ||
| 181 : | mov eax, [esp+12] ; BpS | ||
| 182 : | |||
| 183 : | movq mm6, [Up13] | ||
| 184 : | movq mm7, [Up31] | ||
| 185 : | |||
| 186 : | COL03 mm0, mm1, 0 | ||
| 187 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 188 : | movq mm4, mm0 | ||
| 189 : | movq mm5, mm1 | ||
| 190 : | STORE_1 mm4, mm5 | ||
| 191 : | add ecx, eax | ||
| 192 : | |||
| 193 : | COL03 mm2, mm3, 1 | ||
| 194 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 195 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 196 : | STORE_2 mm0, mm1 | ||
| 197 : | |||
| 198 : | COL03 mm0, mm1, 2 | ||
| 199 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 200 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 201 : | STORE_2 mm2, mm3 | ||
| 202 : | |||
| 203 : | COL03 mm2, mm3, 3 | ||
| 204 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 205 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 206 : | STORE_2 mm0, mm1 | ||
| 207 : | |||
| 208 : | COL03 mm0, mm1, 4 | ||
| 209 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 210 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 211 : | STORE_2 mm2, mm3 | ||
| 212 : | |||
| 213 : | COL03 mm2, mm3, 5 | ||
| 214 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 215 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 216 : | STORE_2 mm0, mm1 | ||
| 217 : | |||
| 218 : | COL03 mm0, mm1, 6 | ||
| 219 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 220 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 221 : | STORE_2 mm2, mm3 | ||
| 222 : | |||
| 223 : | COL03 mm2, mm3, 7 | ||
| 224 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 225 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 226 : | STORE_2 mm0, mm1 | ||
| 227 : | |||
| 228 : | STORE_1 mm2, mm3 | ||
| 229 : | |||
| 230 : | mov ecx, [esp+4] | ||
| 231 : | add ecx, 8 | ||
| 232 : | |||
| 233 : | COL47 mm0, mm1, 0 | ||
| 234 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 235 : | movq mm4, mm0 | ||
| 236 : | movq mm5, mm1 | ||
| 237 : | STORE_1 mm4, mm5 | ||
| 238 : | add ecx, eax | ||
| 239 : | |||
| 240 : | COL47 mm2, mm3, 1 | ||
| 241 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 242 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 243 : | STORE_2 mm0, mm1 | ||
| 244 : | |||
| 245 : | COL47 mm0, mm1, 2 | ||
| 246 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 247 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 248 : | STORE_2 mm2, mm3 | ||
| 249 : | |||
| 250 : | COL47 mm2, mm3, 3 | ||
| 251 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 252 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 253 : | STORE_2 mm0, mm1 | ||
| 254 : | |||
| 255 : | COL47 mm0, mm1, 4 | ||
| 256 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 257 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 258 : | STORE_2 mm2, mm3 | ||
| 259 : | |||
| 260 : | COL47 mm2, mm3, 5 | ||
| 261 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 262 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 263 : | STORE_2 mm0, mm1 | ||
| 264 : | |||
| 265 : | COL47 mm0, mm1, 6 | ||
| 266 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 267 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 268 : | STORE_2 mm2, mm3 | ||
| 269 : | |||
| 270 : | COL47 mm2, mm3, 7 | ||
| 271 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 272 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 273 : | STORE_2 mm0, mm1 | ||
| 274 : | |||
| 275 : | STORE_1 mm2, mm3 | ||
| 276 : | |||
| 277 : | ret | ||
| 278 : | Isibaar | 1793 | ENDFUNC |
| 279 : | edgomez | 851 | |
| 280 : | ;=========================================================================== | ||
| 281 : | ; | ||
| 282 : | edgomez | 1382 | ; void xvid_Add_Upsampled_8x8_16To8_mmx(uint8_t *Dst, |
| 283 : | edgomez | 851 | ; const int16_t *Src, const int BpS); |
| 284 : | ; | ||
| 285 : | ;=========================================================================== | ||
| 286 : | |||
| 287 : | ; Note: grrr... the 'pcmpgtw' stuff are the "/4" and "/16" operators | ||
| 288 : | edgomez | 1382 | ; implemented with ">>2" and ">>4" using: |
| 289 : | edgomez | 851 | ; x/4 = ( (x-(x<0))>>2 ) + (x<0) |
| 290 : | ; x/16 = ( (x-(x<0))>>4 ) + (x<0) | ||
| 291 : | |||
| 292 : | %macro STORE_ADD_1 2 | ||
| 293 : | ; We substract the rounder '2' for corner pixels, | ||
| 294 : | ; since when 'x' is negative, (x*4 + 2)/4 is *not* | ||
| 295 : | ; equal to 'x'. In fact, the correct relation is: | ||
| 296 : | ; (x*4 + 2)/4 = x - (x<0) | ||
| 297 : | ; So, better revert to (x*4)/4 = x. | ||
| 298 : | |||
| 299 : | psubsw %1, [Cst2000] | ||
| 300 : | psubsw %2, [Cst0002] | ||
| 301 : | pxor mm6, mm6 | ||
| 302 : | pxor mm7, mm7 | ||
| 303 : | pcmpgtw mm6, %1 | ||
| 304 : | pcmpgtw mm7, %2 | ||
| 305 : | paddsw %1, mm6 | ||
| 306 : | paddsw %2, mm7 | ||
| 307 : | psraw %1, 2 | ||
| 308 : | psraw %2, 2 | ||
| 309 : | psubsw %1, mm6 | ||
| 310 : | psubsw %2, mm7 | ||
| 311 : | |||
| 312 : | ; mix with destination [ecx] | ||
| 313 : | movq mm6, [ecx] | ||
| 314 : | movq mm7, [ecx] | ||
| 315 : | punpcklbw mm6, [Cst0] | ||
| 316 : | punpckhbw mm7, [Cst0] | ||
| 317 : | paddsw %1, mm6 | ||
| 318 : | paddsw %2, mm7 | ||
| 319 : | packuswb %1,%2 | ||
| 320 : | movq [ecx], %1 | ||
| 321 : | %endmacro | ||
| 322 : | |||
| 323 : | %macro STORE_ADD_2 2 | ||
| 324 : | pxor mm6, mm6 | ||
| 325 : | pxor mm7, mm7 | ||
| 326 : | pcmpgtw mm6, %1 | ||
| 327 : | pcmpgtw mm7, %2 | ||
| 328 : | paddsw %1, mm6 | ||
| 329 : | paddsw %2, mm7 | ||
| 330 : | psraw %1, 4 | ||
| 331 : | psraw %2, 4 | ||
| 332 : | psubsw %1, mm6 | ||
| 333 : | psubsw %2, mm7 | ||
| 334 : | |||
| 335 : | pxor mm6, mm6 | ||
| 336 : | pxor mm7, mm7 | ||
| 337 : | pcmpgtw mm6, mm4 | ||
| 338 : | pcmpgtw mm7, mm5 | ||
| 339 : | paddsw mm4, mm6 | ||
| 340 : | paddsw mm5, mm7 | ||
| 341 : | psraw mm4, 4 | ||
| 342 : | psraw mm5, 4 | ||
| 343 : | psubsw mm4, mm6 | ||
| 344 : | psubsw mm5, mm7 | ||
| 345 : | |||
| 346 : | ; mix with destination | ||
| 347 : | movq mm6, [ecx] | ||
| 348 : | movq mm7, [ecx] | ||
| 349 : | punpcklbw mm6, [Cst0] | ||
| 350 : | punpckhbw mm7, [Cst0] | ||
| 351 : | paddsw %1, mm6 | ||
| 352 : | paddsw %2, mm7 | ||
| 353 : | |||
| 354 : | movq mm6, [ecx+eax] | ||
| 355 : | movq mm7, [ecx+eax] | ||
| 356 : | |||
| 357 : | punpcklbw mm6, [Cst0] | ||
| 358 : | punpckhbw mm7, [Cst0] | ||
| 359 : | paddsw mm4, mm6 | ||
| 360 : | paddsw mm5, mm7 | ||
| 361 : | |||
| 362 : | packuswb %1,%2 | ||
| 363 : | packuswb mm4, mm5 | ||
| 364 : | |||
| 365 : | movq [ecx], %1 | ||
| 366 : | movq [ecx+eax], mm4 | ||
| 367 : | |||
| 368 : | lea ecx, [ecx+2*eax] | ||
| 369 : | %endmacro | ||
| 370 : | |||
| 371 : | ;////////////////////////////////////////////////////////////////////// | ||
| 372 : | |||
| 373 : | align 16 | ||
| 374 : | xvid_Add_Upsampled_8x8_16To8_mmx: ; 579c | ||
| 375 : | |||
| 376 : | mov ecx, [esp+4] ; Dst | ||
| 377 : | mov edx, [esp+8] ; Src | ||
| 378 : | mov eax, [esp+12] ; BpS | ||
| 379 : | |||
| 380 : | COL03 mm0, mm1, 0 | ||
| 381 : | edgomez | 1382 | MUL_PACK mm0,mm1, [Up13], [Up31] |
| 382 : | edgomez | 851 | movq mm4, mm0 |
| 383 : | movq mm5, mm1 | ||
| 384 : | STORE_ADD_1 mm4, mm5 | ||
| 385 : | add ecx, eax | ||
| 386 : | |||
| 387 : | COL03 mm2, mm3, 1 | ||
| 388 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 389 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 390 : | STORE_ADD_2 mm0, mm1 | ||
| 391 : | |||
| 392 : | COL03 mm0, mm1, 2 | ||
| 393 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 394 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 395 : | STORE_ADD_2 mm2, mm3 | ||
| 396 : | |||
| 397 : | COL03 mm2, mm3, 3 | ||
| 398 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 399 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 400 : | STORE_ADD_2 mm0, mm1 | ||
| 401 : | |||
| 402 : | COL03 mm0, mm1, 4 | ||
| 403 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 404 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 405 : | STORE_ADD_2 mm2, mm3 | ||
| 406 : | |||
| 407 : | COL03 mm2, mm3, 5 | ||
| 408 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 409 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 410 : | STORE_ADD_2 mm0, mm1 | ||
| 411 : | |||
| 412 : | COL03 mm0, mm1, 6 | ||
| 413 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 414 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 415 : | STORE_ADD_2 mm2, mm3 | ||
| 416 : | |||
| 417 : | COL03 mm2, mm3, 7 | ||
| 418 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 419 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 420 : | STORE_ADD_2 mm0, mm1 | ||
| 421 : | |||
| 422 : | STORE_ADD_1 mm2, mm3 | ||
| 423 : | |||
| 424 : | |||
| 425 : | mov ecx, [esp+4] | ||
| 426 : | add ecx, 8 | ||
| 427 : | |||
| 428 : | COL47 mm0, mm1, 0 | ||
| 429 : | edgomez | 1382 | MUL_PACK mm0,mm1, [Up13], [Up31] |
| 430 : | edgomez | 851 | movq mm4, mm0 |
| 431 : | edgomez | 1382 | movq mm5, mm1 |
| 432 : | edgomez | 851 | STORE_ADD_1 mm4, mm5 |
| 433 : | add ecx, eax | ||
| 434 : | |||
| 435 : | COL47 mm2, mm3, 1 | ||
| 436 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 437 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 438 : | STORE_ADD_2 mm0, mm1 | ||
| 439 : | |||
| 440 : | COL47 mm0, mm1, 2 | ||
| 441 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 442 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 443 : | STORE_ADD_2 mm2, mm3 | ||
| 444 : | |||
| 445 : | COL47 mm2, mm3, 3 | ||
| 446 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 447 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 448 : | STORE_ADD_2 mm0, mm1 | ||
| 449 : | |||
| 450 : | COL47 mm0, mm1, 4 | ||
| 451 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 452 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 453 : | STORE_ADD_2 mm2, mm3 | ||
| 454 : | |||
| 455 : | COL47 mm2, mm3, 5 | ||
| 456 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 457 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 458 : | STORE_ADD_2 mm0, mm1 | ||
| 459 : | |||
| 460 : | COL47 mm0, mm1, 6 | ||
| 461 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 462 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 463 : | STORE_ADD_2 mm2, mm3 | ||
| 464 : | |||
| 465 : | COL47 mm2, mm3, 7 | ||
| 466 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 467 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 468 : | STORE_ADD_2 mm0, mm1 | ||
| 469 : | |||
| 470 : | STORE_ADD_1 mm2, mm3 | ||
| 471 : | |||
| 472 : | ret | ||
| 473 : | Isibaar | 1793 | ENDFUNC |
| 474 : | edgomez | 851 | |
| 475 : | ;=========================================================================== | ||
| 476 : | ; | ||
| 477 : | edgomez | 1382 | ; void xvid_Copy_Upsampled_8x8_16To8_xmm(uint8_t *Dst, |
| 478 : | edgomez | 851 | ; const int16_t *Src, const int BpS); |
| 479 : | ; | ||
| 480 : | ;=========================================================================== | ||
| 481 : | |||
| 482 : | ; xmm version can take (little) advantage of 'pshufw' | ||
| 483 : | |||
| 484 : | %macro COL03_SSE 3 ;%1/%2: regs, %3: row -trashes mm4/mm5 | ||
| 485 : | movq %2, [edx+%3*16+0*2] ; <- 0|1|2|3 | ||
| 486 : | pshufw %1, %2, (0+0*4+0*16+1*64) ; %1 = 0|0|0|1 | ||
| 487 : | pshufw mm4, %2, (0+1*4+1*16+2*64) ; mm4= 0|1|1|2 | ||
| 488 : | pshufw %2, %2, (1+2*4+2*16+3*64) ; %2 = 1|2|2|3 | ||
| 489 : | pshufw mm5, [edx+%3*16+2*2], (0+1*4+1*16+2*64) ; mm5 = 2|3|3|4 | ||
| 490 : | %endmacro | ||
| 491 : | |||
| 492 : | %macro COL47_SSE 3 ;%1-%2: regs, %3: row -trashes mm4/mm5 | ||
| 493 : | pshufw %1, [edx+%3*16+2*2], (1+2*4+2*16+3*64) ; 3|4|4|5 | ||
| 494 : | movq mm5, [edx+%3*16+2*4] ; <- 4|5|6|7 | ||
| 495 : | pshufw mm4, mm5, (0+1*4+1*16+2*64) ; 4|5|5|6 | ||
| 496 : | pshufw %2, mm5, (1+2*4+2*16+3*64) ; 5|6|6|7 | ||
| 497 : | pshufw mm5, mm5, (2+3*4+3*16+3*64) ; 6|7|7|7 | ||
| 498 : | %endmacro | ||
| 499 : | |||
| 500 : | |||
| 501 : | ;////////////////////////////////////////////////////////////////////// | ||
| 502 : | |||
| 503 : | align 16 | ||
| 504 : | xvid_Copy_Upsampled_8x8_16To8_xmm: ; 315c | ||
| 505 : | |||
| 506 : | mov ecx, [esp+4] ; Dst | ||
| 507 : | mov edx, [esp+8] ; Src | ||
| 508 : | mov eax, [esp+12] ; BpS | ||
| 509 : | |||
| 510 : | movq mm6, [Up13] | ||
| 511 : | movq mm7, [Up31] | ||
| 512 : | |||
| 513 : | COL03_SSE mm0, mm1, 0 | ||
| 514 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 515 : | movq mm4, mm0 | ||
| 516 : | movq mm5, mm1 | ||
| 517 : | STORE_1 mm4, mm5 | ||
| 518 : | add ecx, eax | ||
| 519 : | |||
| 520 : | COL03_SSE mm2, mm3, 1 | ||
| 521 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 522 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 523 : | STORE_2 mm0, mm1 | ||
| 524 : | |||
| 525 : | COL03_SSE mm0, mm1, 2 | ||
| 526 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 527 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 528 : | STORE_2 mm2, mm3 | ||
| 529 : | |||
| 530 : | COL03_SSE mm2, mm3, 3 | ||
| 531 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 532 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 533 : | STORE_2 mm0, mm1 | ||
| 534 : | |||
| 535 : | COL03_SSE mm0, mm1, 4 | ||
| 536 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 537 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 538 : | STORE_2 mm2, mm3 | ||
| 539 : | |||
| 540 : | COL03_SSE mm2, mm3, 5 | ||
| 541 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 542 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 543 : | STORE_2 mm0, mm1 | ||
| 544 : | |||
| 545 : | COL03_SSE mm0, mm1, 6 | ||
| 546 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 547 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 548 : | STORE_2 mm2, mm3 | ||
| 549 : | |||
| 550 : | COL03_SSE mm2, mm3, 7 | ||
| 551 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 552 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 553 : | STORE_2 mm0, mm1 | ||
| 554 : | |||
| 555 : | STORE_1 mm2, mm3 | ||
| 556 : | |||
| 557 : | mov ecx, [esp+4] | ||
| 558 : | add ecx, 8 | ||
| 559 : | |||
| 560 : | COL47_SSE mm0, mm1, 0 | ||
| 561 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 562 : | movq mm4, mm0 | ||
| 563 : | movq mm5, mm1 | ||
| 564 : | STORE_1 mm4, mm5 | ||
| 565 : | add ecx, eax | ||
| 566 : | |||
| 567 : | COL47_SSE mm2, mm3, 1 | ||
| 568 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 569 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 570 : | STORE_2 mm0, mm1 | ||
| 571 : | |||
| 572 : | COL47_SSE mm0, mm1, 2 | ||
| 573 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 574 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 575 : | STORE_2 mm2, mm3 | ||
| 576 : | |||
| 577 : | COL47_SSE mm2, mm3, 3 | ||
| 578 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 579 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 580 : | STORE_2 mm0, mm1 | ||
| 581 : | |||
| 582 : | COL47_SSE mm0, mm1, 4 | ||
| 583 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 584 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 585 : | STORE_2 mm2, mm3 | ||
| 586 : | |||
| 587 : | COL47_SSE mm2, mm3, 5 | ||
| 588 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 589 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 590 : | STORE_2 mm0, mm1 | ||
| 591 : | |||
| 592 : | COL47_SSE mm0, mm1, 6 | ||
| 593 : | MUL_PACK mm0,mm1, mm6, mm7 | ||
| 594 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 595 : | STORE_2 mm2, mm3 | ||
| 596 : | |||
| 597 : | COL47_SSE mm2, mm3, 7 | ||
| 598 : | MUL_PACK mm2,mm3, mm6, mm7 | ||
| 599 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 600 : | STORE_2 mm0, mm1 | ||
| 601 : | |||
| 602 : | STORE_1 mm2, mm3 | ||
| 603 : | |||
| 604 : | ret | ||
| 605 : | Isibaar | 1793 | ENDFUNC |
| 606 : | edgomez | 851 | |
| 607 : | ;=========================================================================== | ||
| 608 : | ; | ||
| 609 : | edgomez | 1382 | ; void xvid_Add_Upsampled_8x8_16To8_xmm(uint8_t *Dst, |
| 610 : | edgomez | 851 | ; const int16_t *Src, const int BpS); |
| 611 : | ; | ||
| 612 : | ;=========================================================================== | ||
| 613 : | |||
| 614 : | align 16 | ||
| 615 : | xvid_Add_Upsampled_8x8_16To8_xmm: ; 549c | ||
| 616 : | |||
| 617 : | mov ecx, [esp+4] ; Dst | ||
| 618 : | mov edx, [esp+8] ; Src | ||
| 619 : | mov eax, [esp+12] ; BpS | ||
| 620 : | |||
| 621 : | COL03_SSE mm0, mm1, 0 | ||
| 622 : | edgomez | 1382 | MUL_PACK mm0,mm1, [Up13], [Up31] |
| 623 : | edgomez | 851 | movq mm4, mm0 |
| 624 : | movq mm5, mm1 | ||
| 625 : | STORE_ADD_1 mm4, mm5 | ||
| 626 : | add ecx, eax | ||
| 627 : | |||
| 628 : | COL03_SSE mm2, mm3, 1 | ||
| 629 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 630 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 631 : | STORE_ADD_2 mm0, mm1 | ||
| 632 : | |||
| 633 : | COL03_SSE mm0, mm1, 2 | ||
| 634 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 635 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 636 : | STORE_ADD_2 mm2, mm3 | ||
| 637 : | |||
| 638 : | COL03_SSE mm2, mm3, 3 | ||
| 639 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 640 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 641 : | STORE_ADD_2 mm0, mm1 | ||
| 642 : | |||
| 643 : | COL03_SSE mm0, mm1, 4 | ||
| 644 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 645 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 646 : | STORE_ADD_2 mm2, mm3 | ||
| 647 : | |||
| 648 : | COL03_SSE mm2, mm3, 5 | ||
| 649 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 650 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 651 : | STORE_ADD_2 mm0, mm1 | ||
| 652 : | |||
| 653 : | COL03_SSE mm0, mm1, 6 | ||
| 654 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 655 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 656 : | STORE_ADD_2 mm2, mm3 | ||
| 657 : | |||
| 658 : | COL03_SSE mm2, mm3, 7 | ||
| 659 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 660 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 661 : | STORE_ADD_2 mm0, mm1 | ||
| 662 : | |||
| 663 : | STORE_ADD_1 mm2, mm3 | ||
| 664 : | |||
| 665 : | |||
| 666 : | mov ecx, [esp+4] | ||
| 667 : | add ecx, 8 | ||
| 668 : | |||
| 669 : | COL47_SSE mm0, mm1, 0 | ||
| 670 : | edgomez | 1382 | MUL_PACK mm0,mm1, [Up13], [Up31] |
| 671 : | edgomez | 851 | movq mm4, mm0 |
| 672 : | edgomez | 1382 | movq mm5, mm1 |
| 673 : | edgomez | 851 | STORE_ADD_1 mm4, mm5 |
| 674 : | add ecx, eax | ||
| 675 : | |||
| 676 : | COL47_SSE mm2, mm3, 1 | ||
| 677 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 678 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 679 : | STORE_ADD_2 mm0, mm1 | ||
| 680 : | |||
| 681 : | COL47_SSE mm0, mm1, 2 | ||
| 682 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 683 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 684 : | STORE_ADD_2 mm2, mm3 | ||
| 685 : | |||
| 686 : | COL47_SSE mm2, mm3, 3 | ||
| 687 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 688 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 689 : | STORE_ADD_2 mm0, mm1 | ||
| 690 : | |||
| 691 : | COL47_SSE mm0, mm1, 4 | ||
| 692 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 693 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 694 : | STORE_ADD_2 mm2, mm3 | ||
| 695 : | |||
| 696 : | COL47_SSE mm2, mm3, 5 | ||
| 697 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 698 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 699 : | STORE_ADD_2 mm0, mm1 | ||
| 700 : | |||
| 701 : | COL47_SSE mm0, mm1, 6 | ||
| 702 : | MUL_PACK mm0,mm1, [Up13], [Up31] | ||
| 703 : | MIX_ROWS mm2, mm3, mm0, mm1 | ||
| 704 : | STORE_ADD_2 mm2, mm3 | ||
| 705 : | |||
| 706 : | COL47_SSE mm2, mm3, 7 | ||
| 707 : | MUL_PACK mm2,mm3, [Up13], [Up31] | ||
| 708 : | MIX_ROWS mm0, mm1, mm2, mm3 | ||
| 709 : | STORE_ADD_2 mm0, mm1 | ||
| 710 : | |||
| 711 : | STORE_ADD_1 mm2, mm3 | ||
| 712 : | |||
| 713 : | ret | ||
| 714 : | Isibaar | 1793 | ENDFUNC |
| 715 : | edgomez | 851 | |
| 716 : | |||
| 717 : | ;=========================================================================== | ||
| 718 : | ; | ||
| 719 : | ; void xvid_HFilter_31_mmx(uint8_t *Src1, uint8_t *Src2, int Nb_Blks); | ||
| 720 : | ; void xvid_VFilter_31_x86(uint8_t *Src1, uint8_t *Src2, const int BpS, int Nb_Blks); | ||
| 721 : | ; void xvid_HFilter_31_x86(uint8_t *Src1, uint8_t *Src2, int Nb_Blks); | ||
| 722 : | ; | ||
| 723 : | ;=========================================================================== | ||
| 724 : | |||
| 725 : | ;////////////////////////////////////////////////////////////////////// | ||
| 726 : | ;// horizontal/vertical filtering: [x,y] -> [ (3x+y+2)>>2, (x+3y+2)>>2 ] | ||
| 727 : | ;// | ||
| 728 : | ;// We use the trick: tmp = (x+y+2) -> [x = (tmp+2x)>>2, y = (tmp+2y)>>2] | ||
| 729 : | ;////////////////////////////////////////////////////////////////////// | ||
| 730 : | |||
| 731 : | align 16 | ||
| 732 : | xvid_HFilter_31_mmx: | ||
| 733 : | push esi | ||
| 734 : | push edi | ||
| 735 : | mov esi, [esp+4 +8] ; Src1 | ||
| 736 : | mov edi, [esp+8 +8] ; Src2 | ||
| 737 : | mov eax, [esp+12 +8] ; Nb_Blks | ||
| 738 : | lea eax,[eax*2] | ||
| 739 : | movq mm5, [Cst2] | ||
| 740 : | pxor mm7, mm7 | ||
| 741 : | |||
| 742 : | lea esi, [esi+eax*4] | ||
| 743 : | lea edi, [edi+eax*4] | ||
| 744 : | |||
| 745 : | neg eax | ||
| 746 : | |||
| 747 : | .Loop: ;12c | ||
| 748 : | edgomez | 1382 | movd mm0, [esi+eax*4] |
| 749 : | edgomez | 851 | movd mm1, [edi+eax*4] |
| 750 : | edgomez | 1382 | movq mm2, mm5 |
| 751 : | edgomez | 851 | punpcklbw mm0, mm7 |
| 752 : | punpcklbw mm1, mm7 | ||
| 753 : | paddsw mm2, mm0 | ||
| 754 : | paddsw mm0, mm0 | ||
| 755 : | paddsw mm2, mm1 | ||
| 756 : | paddsw mm1, mm1 | ||
| 757 : | paddsw mm0, mm2 | ||
| 758 : | paddsw mm1, mm2 | ||
| 759 : | psraw mm0, 2 | ||
| 760 : | psraw mm1, 2 | ||
| 761 : | packuswb mm0, mm7 | ||
| 762 : | packuswb mm1, mm7 | ||
| 763 : | movd [esi+eax*4], mm0 | ||
| 764 : | movd [edi+eax*4], mm1 | ||
| 765 : | add eax,1 | ||
| 766 : | jl .Loop | ||
| 767 : | |||
| 768 : | pop edi | ||
| 769 : | pop esi | ||
| 770 : | ret | ||
| 771 : | Isibaar | 1793 | ENDFUNC |
| 772 : | edgomez | 851 | |
| 773 : | ; mmx is of no use here. Better use plain ASM. Moreover, | ||
| 774 : | ; this is for the fun of ASM coding, coz' every modern compiler can | ||
| 775 : | ; end up with a code that looks very much like this one... | ||
| 776 : | |||
| 777 : | align 16 | ||
| 778 : | xvid_VFilter_31_x86: | ||
| 779 : | push esi | ||
| 780 : | push edi | ||
| 781 : | push ebx | ||
| 782 : | push ebp | ||
| 783 : | mov esi, [esp+4 +16] ; Src1 | ||
| 784 : | mov edi, [esp+8 +16] ; Src2 | ||
| 785 : | mov ebp, [esp+12 +16] ; BpS | ||
| 786 : | mov eax, [esp+16 +16] ; Nb_Blks | ||
| 787 : | lea eax,[eax*8] | ||
| 788 : | |||
| 789 : | .Loop: ;7c | ||
| 790 : | movzx ecx, byte [esi] | ||
| 791 : | movzx edx, byte [edi] | ||
| 792 : | |||
| 793 : | lea ebx, [ecx+edx+2] | ||
| 794 : | lea ecx,[ebx+2*ecx] | ||
| 795 : | lea edx,[ebx+2*edx] | ||
| 796 : | |||
| 797 : | shr ecx,2 | ||
| 798 : | shr edx,2 | ||
| 799 : | mov [esi], cl | ||
| 800 : | mov [edi], dl | ||
| 801 : | lea esi, [esi+ebp] | ||
| 802 : | lea edi, [edi+ebp] | ||
| 803 : | dec eax | ||
| 804 : | jg .Loop | ||
| 805 : | |||
| 806 : | pop ebp | ||
| 807 : | pop ebx | ||
| 808 : | pop edi | ||
| 809 : | pop esi | ||
| 810 : | ret | ||
| 811 : | Isibaar | 1793 | ENDFUNC |
| 812 : | edgomez | 851 | |
| 813 : | ; this one's just a little faster than gcc's code. Very little. | ||
| 814 : | |||
| 815 : | align 16 | ||
| 816 : | xvid_HFilter_31_x86: | ||
| 817 : | push esi | ||
| 818 : | push edi | ||
| 819 : | push ebx | ||
| 820 : | mov esi, [esp+4 +12] ; Src1 | ||
| 821 : | mov edi, [esp+8 +12] ; Src2 | ||
| 822 : | mov eax, [esp+12 +12] ; Nb_Blks | ||
| 823 : | |||
| 824 : | lea eax,[eax*8] | ||
| 825 : | lea esi, [esi+eax] | ||
| 826 : | lea edi, [esi+eax] | ||
| 827 : | neg eax | ||
| 828 : | |||
| 829 : | .Loop: ; 6c | ||
| 830 : | movzx ecx, byte [esi+eax] | ||
| 831 : | movzx edx, byte [edi+eax] | ||
| 832 : | |||
| 833 : | lea ebx, [ecx+edx+2] | ||
| 834 : | lea ecx,[ebx+2*ecx] | ||
| 835 : | lea edx,[ebx+2*edx] | ||
| 836 : | shr ecx,2 | ||
| 837 : | shr edx,2 | ||
| 838 : | mov [esi+eax], cl | ||
| 839 : | mov [edi+eax], dl | ||
| 840 : | inc eax | ||
| 841 : | |||
| 842 : | jl .Loop | ||
| 843 : | |||
| 844 : | pop ebx | ||
| 845 : | pop edi | ||
| 846 : | pop esi | ||
| 847 : | ret | ||
| 848 : | Isibaar | 1793 | ENDFUNC |
| 849 : | edgomez | 851 | |
| 850 : | ;////////////////////////////////////////////////////////////////////// | ||
| 851 : | ;// 16b downsampling 16x16 -> 8x8 | ||
| 852 : | ;////////////////////////////////////////////////////////////////////// | ||
| 853 : | |||
| 854 : | %macro HFILTER_1331 2 ;%1:src %2:dst reg. -trashes mm0/mm1/mm2 | ||
| 855 : | movq mm2, [Mask_ff] | ||
| 856 : | movq %2, [%1-1] ;-10123456 | ||
| 857 : | movq mm0, [%1] ; 01234567 | ||
| 858 : | movq mm1, [%1+1] ; 12345678 | ||
| 859 : | pand %2, mm2 ;-1|1|3|5 | ||
| 860 : | pand mm0, mm2 ; 0|2|4|6 | ||
| 861 : | pand mm1, mm2 ; 1|3|5|7 | ||
| 862 : | pand mm2, [%1+2] ; 2|4|6|8 | ||
| 863 : | paddusw mm0, mm1 | ||
| 864 : | paddusw %2, mm2 | ||
| 865 : | pmullw mm0, mm7 | ||
| 866 : | paddusw %2, mm0 | ||
| 867 : | %endmacro | ||
| 868 : | |||
| 869 : | %macro VFILTER_1331 4 ; %1-4: regs %1-%2: trashed | ||
| 870 : | paddsw %1, [Cst32] | ||
| 871 : | edgomez | 1382 | paddsw %2, %3 |
| 872 : | edgomez | 851 | pmullw %2, mm7 |
| 873 : | paddsw %1,%4 | ||
| 874 : | paddsw %1, %2 | ||
| 875 : | psraw %1, 6 | ||
| 876 : | %endmacro | ||
| 877 : | |||
| 878 : | ;=========================================================================== | ||
| 879 : | ; | ||
| 880 : | ; void xvid_Filter_18x18_To_8x8_mmx(int16_t *Dst, | ||
| 881 : | ; const uint8_t *Src, const int BpS); | ||
| 882 : | ; | ||
| 883 : | ;=========================================================================== | ||
| 884 : | |||
| 885 : | %macro COPY_TWO_LINES_1331 1 ; %1: dst | ||
| 886 : | HFILTER_1331 edx , mm5 | ||
| 887 : | HFILTER_1331 edx+eax, mm6 | ||
| 888 : | lea edx, [edx+2*eax] | ||
| 889 : | VFILTER_1331 mm3,mm4,mm5, mm6 | ||
| 890 : | movq [%1], mm3 | ||
| 891 : | |||
| 892 : | HFILTER_1331 edx , mm3 | ||
| 893 : | HFILTER_1331 edx+eax, mm4 | ||
| 894 : | lea edx, [edx+2*eax] | ||
| 895 : | VFILTER_1331 mm5,mm6,mm3,mm4 | ||
| 896 : | movq [%1+16], mm5 | ||
| 897 : | %endmacro | ||
| 898 : | |||
| 899 : | align 16 | ||
| 900 : | xvid_Filter_18x18_To_8x8_mmx: ; 283c (~4.4c per output pixel) | ||
| 901 : | |||
| 902 : | mov ecx, [esp+4] ; Dst | ||
| 903 : | mov edx, [esp+8] ; Src | ||
| 904 : | mov eax, [esp+12] ; BpS | ||
| 905 : | |||
| 906 : | movq mm7, [Cst3] | ||
| 907 : | sub edx, eax | ||
| 908 : | |||
| 909 : | ; mm3/mm4/mm5/mm6 is used as a 4-samples delay line. | ||
| 910 : | |||
| 911 : | ; process columns 0-3 | ||
| 912 : | |||
| 913 : | HFILTER_1331 edx , mm3 ; pre-load mm3/mm4 | ||
| 914 : | HFILTER_1331 edx+eax, mm4 | ||
| 915 : | lea edx, [edx+2*eax] | ||
| 916 : | |||
| 917 : | COPY_TWO_LINES_1331 ecx + 0*16 | ||
| 918 : | COPY_TWO_LINES_1331 ecx + 2*16 | ||
| 919 : | COPY_TWO_LINES_1331 ecx + 4*16 | ||
| 920 : | COPY_TWO_LINES_1331 ecx + 6*16 | ||
| 921 : | |||
| 922 : | ; process columns 4-7 | ||
| 923 : | |||
| 924 : | mov edx, [esp+8] | ||
| 925 : | sub edx, eax | ||
| 926 : | add edx, 8 | ||
| 927 : | |||
| 928 : | HFILTER_1331 edx , mm3 ; pre-load mm3/mm4 | ||
| 929 : | HFILTER_1331 edx+eax, mm4 | ||
| 930 : | lea edx, [edx+2*eax] | ||
| 931 : | |||
| 932 : | COPY_TWO_LINES_1331 ecx + 0*16 +8 | ||
| 933 : | COPY_TWO_LINES_1331 ecx + 2*16 +8 | ||
| 934 : | COPY_TWO_LINES_1331 ecx + 4*16 +8 | ||
| 935 : | COPY_TWO_LINES_1331 ecx + 6*16 +8 | ||
| 936 : | |||
| 937 : | ret | ||
| 938 : | Isibaar | 1793 | ENDFUNC |
| 939 : | edgomez | 851 | |
| 940 : | ;=========================================================================== | ||
| 941 : | ; | ||
| 942 : | ; void xvid_Filter_Diff_18x18_To_8x8_mmx(int16_t *Dst, | ||
| 943 : | ; const uint8_t *Src, const int BpS); | ||
| 944 : | ; | ||
| 945 : | ;=========================================================================== | ||
| 946 : | |||
| 947 : | %macro DIFF_TWO_LINES_1331 1 ; %1: dst | ||
| 948 : | HFILTER_1331 edx , mm5 | ||
| 949 : | HFILTER_1331 edx+eax, mm6 | ||
| 950 : | lea edx, [edx+2*eax] | ||
| 951 : | movq mm2, [%1] | ||
| 952 : | VFILTER_1331 mm3,mm4,mm5, mm6 | ||
| 953 : | psubsw mm2, mm3 | ||
| 954 : | movq [%1], mm2 | ||
| 955 : | |||
| 956 : | HFILTER_1331 edx , mm3 | ||
| 957 : | HFILTER_1331 edx+eax, mm4 | ||
| 958 : | lea edx, [edx+2*eax] | ||
| 959 : | movq mm2, [%1+16] | ||
| 960 : | VFILTER_1331 mm5,mm6,mm3,mm4 | ||
| 961 : | psubsw mm2, mm5 | ||
| 962 : | movq [%1+16], mm2 | ||
| 963 : | %endmacro | ||
| 964 : | |||
| 965 : | align 16 | ||
| 966 : | xvid_Filter_Diff_18x18_To_8x8_mmx: ; 302c | ||
| 967 : | |||
| 968 : | mov ecx, [esp+4] ; Dst | ||
| 969 : | mov edx, [esp+8] ; Src | ||
| 970 : | mov eax, [esp+12] ; BpS | ||
| 971 : | |||
| 972 : | movq mm7, [Cst3] | ||
| 973 : | sub edx, eax | ||
| 974 : | |||
| 975 : | ; mm3/mm4/mm5/mm6 is used as a 4-samples delay line. | ||
| 976 : | |||
| 977 : | ; process columns 0-3 | ||
| 978 : | |||
| 979 : | HFILTER_1331 edx , mm3 ; pre-load mm3/mm4 | ||
| 980 : | HFILTER_1331 edx+eax, mm4 | ||
| 981 : | lea edx, [edx+2*eax] | ||
| 982 : | |||
| 983 : | DIFF_TWO_LINES_1331 ecx + 0*16 | ||
| 984 : | DIFF_TWO_LINES_1331 ecx + 2*16 | ||
| 985 : | DIFF_TWO_LINES_1331 ecx + 4*16 | ||
| 986 : | DIFF_TWO_LINES_1331 ecx + 6*16 | ||
| 987 : | |||
| 988 : | ; process columns 4-7 | ||
| 989 : | mov edx, [esp+8] | ||
| 990 : | sub edx, eax | ||
| 991 : | add edx, 8 | ||
| 992 : | |||
| 993 : | HFILTER_1331 edx , mm3 ; pre-load mm3/mm4 | ||
| 994 : | HFILTER_1331 edx+eax, mm4 | ||
| 995 : | lea edx, [edx+2*eax] | ||
| 996 : | |||
| 997 : | DIFF_TWO_LINES_1331 ecx + 0*16 +8 | ||
| 998 : | DIFF_TWO_LINES_1331 ecx + 2*16 +8 | ||
| 999 : | DIFF_TWO_LINES_1331 ecx + 4*16 +8 | ||
| 1000 : | DIFF_TWO_LINES_1331 ecx + 6*16 +8 | ||
| 1001 : | |||
| 1002 : | ret | ||
| 1003 : | Isibaar | 1793 | ENDFUNC |
| 1004 : | edgomez | 851 | |
| 1005 : | ;////////////////////////////////////////////////////////////////////// | ||
| 1006 : | |||
| 1007 : | ; pfeewwww... Never Do That On Stage Again. :) | ||
| 1008 : | |||
| 1009 : | Isibaar | 1790 | |
| 1010 : | %ifidn __OUTPUT_FORMAT__,elf | ||
| 1011 : | section ".note.GNU-stack" noalloc noexec nowrite progbits | ||
| 1012 : | %endif | ||
| 1013 : |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |