kc3-lang/libjpeg-turbo/simd/jfmmxint.asm
-
Show log
Commit
-
Hash : b7753510
Author :
Date : 2014-05-11T09:36:25
Convert tabs to spaces in the libjpeg code and the SIMD code (TurboJPEG retains the use of tabs for historical reasons. They were annoying in the libjpeg code primarily because they were not consistently used and because they were used to format as well as indent the code. In the case of TurboJPEG, tabs are used just to indent the code, so even if the editor assumes a different tab width, the code will still be readable.) git-svn-id: svn+ssh://svn.code.sf.net/p/libjpeg-turbo/code/branches/1.3.x@1285 632fc199-4ca6-4c93-a231-07263d6284db
-
simd/jfmmxint.asm
-
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
; ; jfmmxint.asm - accurate integer FDCT (MMX) ; ; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB ; ; Based on ; x86 SIMD extension for IJG JPEG library ; Copyright (C) 1999-2006, MIYASAKA Masaru. ; For conditions of distribution and use, see copyright notice in jsimdext.inc ; ; This file should be assembled with NASM (Netwide Assembler), ; can *not* be assembled with Microsoft's MASM or any compatible ; assembler (including Borland's Turbo Assembler). ; NASM is available from http://nasm.sourceforge.net/ or ; http://sourceforge.net/project/showfiles.php?group_id=6208 ; ; This file contains a slow-but-accurate integer implementation of the ; forward DCT (Discrete Cosine Transform). The following code is based ; directly on the IJG's original jfdctint.c; see the jfdctint.c for ; more details. ; ; [TAB8] %include "jsimdext.inc" %include "jdct.inc" ; -------------------------------------------------------------------------- %define CONST_BITS 13 %define PASS1_BITS 2 %define DESCALE_P1 (CONST_BITS-PASS1_BITS) %define DESCALE_P2 (CONST_BITS+PASS1_BITS) %if CONST_BITS == 13 F_0_298 equ 2446 ; FIX(0.298631336) F_0_390 equ 3196 ; FIX(0.390180644) F_0_541 equ 4433 ; FIX(0.541196100) F_0_765 equ 6270 ; FIX(0.765366865) F_0_899 equ 7373 ; FIX(0.899976223) F_1_175 equ 9633 ; FIX(1.175875602) F_1_501 equ 12299 ; FIX(1.501321110) F_1_847 equ 15137 ; FIX(1.847759065) F_1_961 equ 16069 ; FIX(1.961570560) F_2_053 equ 16819 ; FIX(2.053119869) F_2_562 equ 20995 ; FIX(2.562915447) F_3_072 equ 25172 ; FIX(3.072711026) %else ; NASM cannot do compile-time arithmetic on floating-point constants. %define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) F_0_298 equ DESCALE( 320652955,30-CONST_BITS) ; FIX(0.298631336) F_0_390 equ DESCALE( 418953276,30-CONST_BITS) ; FIX(0.390180644) F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) F_1_175 equ DESCALE(1262586813,30-CONST_BITS) ; FIX(1.175875602) F_1_501 equ DESCALE(1612031267,30-CONST_BITS) ; FIX(1.501321110) F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) F_1_961 equ DESCALE(2106220350,30-CONST_BITS) ; FIX(1.961570560) F_2_053 equ DESCALE(2204520673,30-CONST_BITS) ; FIX(2.053119869) F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) F_3_072 equ DESCALE(3299298341,30-CONST_BITS) ; FIX(3.072711026) %endif ; -------------------------------------------------------------------------- SECTION SEG_CONST alignz 16 global EXTN(jconst_fdct_islow_mmx) EXTN(jconst_fdct_islow_mmx): PW_F130_F054 times 2 dw (F_0_541+F_0_765), F_0_541 PW_F054_MF130 times 2 dw F_0_541, (F_0_541-F_1_847) PW_MF078_F117 times 2 dw (F_1_175-F_1_961), F_1_175 PW_F117_F078 times 2 dw F_1_175, (F_1_175-F_0_390) PW_MF060_MF089 times 2 dw (F_0_298-F_0_899),-F_0_899 PW_MF089_F060 times 2 dw -F_0_899, (F_1_501-F_0_899) PW_MF050_MF256 times 2 dw (F_2_053-F_2_562),-F_2_562 PW_MF256_F050 times 2 dw -F_2_562, (F_3_072-F_2_562) PD_DESCALE_P1 times 2 dd 1 << (DESCALE_P1-1) PD_DESCALE_P2 times 2 dd 1 << (DESCALE_P2-1) PW_DESCALE_P2X times 4 dw 1 << (PASS1_BITS-1) alignz 16 ; -------------------------------------------------------------------------- SECTION SEG_TEXT BITS 32 ; ; Perform the forward DCT on one block of samples. ; ; GLOBAL(void) ; jsimd_fdct_islow_mmx (DCTELEM * data) ; %define data(b) (b)+8 ; DCTELEM * data %define original_ebp ebp+0 %define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] %define WK_NUM 2 align 16 global EXTN(jsimd_fdct_islow_mmx) EXTN(jsimd_fdct_islow_mmx): push ebp mov eax,esp ; eax = original ebp sub esp, byte 4 and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits mov [esp],eax mov ebp,esp ; ebp = aligned ebp lea esp, [wk(0)] pushpic ebx ; push ecx ; need not be preserved ; push edx ; need not be preserved ; push esi ; unused ; push edi ; unused get_GOT ebx ; get GOT address ; ---- Pass 1: process rows. mov edx, POINTER [data(eax)] ; (DCTELEM *) mov ecx, DCTSIZE/4 alignx 16,7 .rowloop: movq mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)] movq mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)] movq mm2, MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)] movq mm3, MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)] ; mm0=(20 21 22 23), mm2=(24 25 26 27) ; mm1=(30 31 32 33), mm3=(34 35 36 37) movq mm4,mm0 ; transpose coefficients(phase 1) punpcklwd mm0,mm1 ; mm0=(20 30 21 31) punpckhwd mm4,mm1 ; mm4=(22 32 23 33) movq mm5,mm2 ; transpose coefficients(phase 1) punpcklwd mm2,mm3 ; mm2=(24 34 25 35) punpckhwd mm5,mm3 ; mm5=(26 36 27 37) movq mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)] movq mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)] movq mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)] movq mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)] ; mm6=(00 01 02 03), mm1=(04 05 06 07) ; mm7=(10 11 12 13), mm3=(14 15 16 17) movq MMWORD [wk(0)], mm4 ; wk(0)=(22 32 23 33) movq MMWORD [wk(1)], mm2 ; wk(1)=(24 34 25 35) movq mm4,mm6 ; transpose coefficients(phase 1) punpcklwd mm6,mm7 ; mm6=(00 10 01 11) punpckhwd mm4,mm7 ; mm4=(02 12 03 13) movq mm2,mm1 ; transpose coefficients(phase 1) punpcklwd mm1,mm3 ; mm1=(04 14 05 15) punpckhwd mm2,mm3 ; mm2=(06 16 07 17) movq mm7,mm6 ; transpose coefficients(phase 2) punpckldq mm6,mm0 ; mm6=(00 10 20 30)=data0 punpckhdq mm7,mm0 ; mm7=(01 11 21 31)=data1 movq mm3,mm2 ; transpose coefficients(phase 2) punpckldq mm2,mm5 ; mm2=(06 16 26 36)=data6 punpckhdq mm3,mm5 ; mm3=(07 17 27 37)=data7 movq mm0,mm7 movq mm5,mm6 psubw mm7,mm2 ; mm7=data1-data6=tmp6 psubw mm6,mm3 ; mm6=data0-data7=tmp7 paddw mm0,mm2 ; mm0=data1+data6=tmp1 paddw mm5,mm3 ; mm5=data0+data7=tmp0 movq mm2, MMWORD [wk(0)] ; mm2=(22 32 23 33) movq mm3, MMWORD [wk(1)] ; mm3=(24 34 25 35) movq MMWORD [wk(0)], mm7 ; wk(0)=tmp6 movq MMWORD [wk(1)], mm6 ; wk(1)=tmp7 movq mm7,mm4 ; transpose coefficients(phase 2) punpckldq mm4,mm2 ; mm4=(02 12 22 32)=data2 punpckhdq mm7,mm2 ; mm7=(03 13 23 33)=data3 movq mm6,mm1 ; transpose coefficients(phase 2) punpckldq mm1,mm3 ; mm1=(04 14 24 34)=data4 punpckhdq mm6,mm3 ; mm6=(05 15 25 35)=data5 movq mm2,mm7 movq mm3,mm4 paddw mm7,mm1 ; mm7=data3+data4=tmp3 paddw mm4,mm6 ; mm4=data2+data5=tmp2 psubw mm2,mm1 ; mm2=data3-data4=tmp4 psubw mm3,mm6 ; mm3=data2-data5=tmp5 ; -- Even part movq mm1,mm5 movq mm6,mm0 paddw mm5,mm7 ; mm5=tmp10 paddw mm0,mm4 ; mm0=tmp11 psubw mm1,mm7 ; mm1=tmp13 psubw mm6,mm4 ; mm6=tmp12 movq mm7,mm5 paddw mm5,mm0 ; mm5=tmp10+tmp11 psubw mm7,mm0 ; mm7=tmp10-tmp11 psllw mm5,PASS1_BITS ; mm5=data0 psllw mm7,PASS1_BITS ; mm7=data4 movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5 movq MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)], mm7 ; (Original) ; z1 = (tmp12 + tmp13) * 0.541196100; ; data2 = z1 + tmp13 * 0.765366865; ; data6 = z1 + tmp12 * -1.847759065; ; ; (This implementation) ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); movq mm4,mm1 ; mm1=tmp13 movq mm0,mm1 punpcklwd mm4,mm6 ; mm6=tmp12 punpckhwd mm0,mm6 movq mm1,mm4 movq mm6,mm0 pmaddwd mm4,[GOTOFF(ebx,PW_F130_F054)] ; mm4=data2L pmaddwd mm0,[GOTOFF(ebx,PW_F130_F054)] ; mm0=data2H pmaddwd mm1,[GOTOFF(ebx,PW_F054_MF130)] ; mm1=data6L pmaddwd mm6,[GOTOFF(ebx,PW_F054_MF130)] ; mm6=data6H paddd mm4,[GOTOFF(ebx,PD_DESCALE_P1)] paddd mm0,[GOTOFF(ebx,PD_DESCALE_P1)] psrad mm4,DESCALE_P1 psrad mm0,DESCALE_P1 paddd mm1,[GOTOFF(ebx,PD_DESCALE_P1)] paddd mm6,[GOTOFF(ebx,PD_DESCALE_P1)] psrad mm1,DESCALE_P1 psrad mm6,DESCALE_P1 packssdw mm4,mm0 ; mm4=data2 packssdw mm1,mm6 ; mm1=data6 movq MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4 movq MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)], mm1 ; -- Odd part movq mm5, MMWORD [wk(0)] ; mm5=tmp6 movq mm7, MMWORD [wk(1)] ; mm7=tmp7 movq mm0,mm2 ; mm2=tmp4 movq mm6,mm3 ; mm3=tmp5 paddw mm0,mm5 ; mm0=z3 paddw mm6,mm7 ; mm6=z4 ; (Original) ; z5 = (z3 + z4) * 1.175875602; ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; ; z3 += z5; z4 += z5; ; ; (This implementation) ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); movq mm4,mm0 movq mm1,mm0 punpcklwd mm4,mm6 punpckhwd mm1,mm6 movq mm0,mm4 movq mm6,mm1 pmaddwd mm4,[GOTOFF(ebx,PW_MF078_F117)] ; mm4=z3L pmaddwd mm1,[GOTOFF(ebx,PW_MF078_F117)] ; mm1=z3H pmaddwd mm0,[GOTOFF(ebx,PW_F117_F078)] ; mm0=z4L pmaddwd mm6,[GOTOFF(ebx,PW_F117_F078)] ; mm6=z4H movq MMWORD [wk(0)], mm4 ; wk(0)=z3L movq MMWORD [wk(1)], mm1 ; wk(1)=z3H ; (Original) ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; ; ; (This implementation) ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); ; data7 = tmp4 + z3; data5 = tmp5 + z4; ; data3 = tmp6 + z3; data1 = tmp7 + z4; movq mm4,mm2 movq mm1,mm2 punpcklwd mm4,mm7 punpckhwd mm1,mm7 movq mm2,mm4 movq mm7,mm1 pmaddwd mm4,[GOTOFF(ebx,PW_MF060_MF089)] ; mm4=tmp4L pmaddwd mm1,[GOTOFF(ebx,PW_MF060_MF089)] ; mm1=tmp4H pmaddwd mm2,[GOTOFF(ebx,PW_MF089_F060)] ; mm2=tmp7L pmaddwd mm7,[GOTOFF(ebx,PW_MF089_F060)] ; mm7=tmp7H paddd mm4, MMWORD [wk(0)] ; mm4=data7L paddd mm1, MMWORD [wk(1)] ; mm1=data7H paddd mm2,mm0 ; mm2=data1L paddd mm7,mm6 ; mm7=data1H paddd mm4,[GOTOFF(ebx,PD_DESCALE_P1)] paddd mm1,[GOTOFF(ebx,PD_DESCALE_P1)] psrad mm4,DESCALE_P1 psrad mm1,DESCALE_P1 paddd mm2,[GOTOFF(ebx,PD_DESCALE_P1)] paddd mm7,[GOTOFF(ebx,PD_DESCALE_P1)] psrad mm2,DESCALE_P1 psrad mm7,DESCALE_P1 packssdw mm4,mm1 ; mm4=data7 packssdw mm2,mm7 ; mm2=data1 movq MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)], mm4 movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2 movq mm1,mm3 movq mm7,mm3 punpcklwd mm1,mm5 punpckhwd mm7,mm5 movq mm3,mm1 movq mm5,mm7 pmaddwd mm1,[GOTOFF(ebx,PW_MF050_MF256)] ; mm1=tmp5L pmaddwd mm7,[GOTOFF(ebx,PW_MF050_MF256)] ; mm7=tmp5H pmaddwd mm3,[GOTOFF(ebx,PW_MF256_F050)] ; mm3=tmp6L pmaddwd mm5,[GOTOFF(ebx,PW_MF256_F050)] ; mm5=tmp6H paddd mm1,mm0 ; mm1=data5L paddd mm7,mm6 ; mm7=data5H paddd mm3, MMWORD [wk(0)] ; mm3=data3L paddd mm5, MMWORD [wk(1)] ; mm5=data3H paddd mm1,[GOTOFF(ebx,PD_DESCALE_P1)] paddd mm7,[GOTOFF(ebx,PD_DESCALE_P1)] psrad mm1,DESCALE_P1 psrad mm7,DESCALE_P1 paddd mm3,[GOTOFF(ebx,PD_DESCALE_P1)] paddd mm5,[GOTOFF(ebx,PD_DESCALE_P1)] psrad mm3,DESCALE_P1 psrad mm5,DESCALE_P1 packssdw mm1,mm7 ; mm1=data5 packssdw mm3,mm5 ; mm3=data3 movq MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)], mm1 movq MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3 add edx, byte 4*DCTSIZE*SIZEOF_DCTELEM dec ecx jnz near .rowloop ; ---- Pass 2: process columns. mov edx, POINTER [data(eax)] ; (DCTELEM *) mov ecx, DCTSIZE/4 alignx 16,7 .columnloop: movq mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)] movq mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)] movq mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)] movq mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)] ; mm0=(02 12 22 32), mm2=(42 52 62 72) ; mm1=(03 13 23 33), mm3=(43 53 63 73) movq mm4,mm0 ; transpose coefficients(phase 1) punpcklwd mm0,mm1 ; mm0=(02 03 12 13) punpckhwd mm4,mm1 ; mm4=(22 23 32 33) movq mm5,mm2 ; transpose coefficients(phase 1) punpcklwd mm2,mm3 ; mm2=(42 43 52 53) punpckhwd mm5,mm3 ; mm5=(62 63 72 73) movq mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)] movq mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)] movq mm1, MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)] movq mm3, MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)] ; mm6=(00 10 20 30), mm1=(40 50 60 70) ; mm7=(01 11 21 31), mm3=(41 51 61 71) movq MMWORD [wk(0)], mm4 ; wk(0)=(22 23 32 33) movq MMWORD [wk(1)], mm2 ; wk(1)=(42 43 52 53) movq mm4,mm6 ; transpose coefficients(phase 1) punpcklwd mm6,mm7 ; mm6=(00 01 10 11) punpckhwd mm4,mm7 ; mm4=(20 21 30 31) movq mm2,mm1 ; transpose coefficients(phase 1) punpcklwd mm1,mm3 ; mm1=(40 41 50 51) punpckhwd mm2,mm3 ; mm2=(60 61 70 71) movq mm7,mm6 ; transpose coefficients(phase 2) punpckldq mm6,mm0 ; mm6=(00 01 02 03)=data0 punpckhdq mm7,mm0 ; mm7=(10 11 12 13)=data1 movq mm3,mm2 ; transpose coefficients(phase 2) punpckldq mm2,mm5 ; mm2=(60 61 62 63)=data6 punpckhdq mm3,mm5 ; mm3=(70 71 72 73)=data7 movq mm0,mm7 movq mm5,mm6 psubw mm7,mm2 ; mm7=data1-data6=tmp6 psubw mm6,mm3 ; mm6=data0-data7=tmp7 paddw mm0,mm2 ; mm0=data1+data6=tmp1 paddw mm5,mm3 ; mm5=data0+data7=tmp0 movq mm2, MMWORD [wk(0)] ; mm2=(22 23 32 33) movq mm3, MMWORD [wk(1)] ; mm3=(42 43 52 53) movq MMWORD [wk(0)], mm7 ; wk(0)=tmp6 movq MMWORD [wk(1)], mm6 ; wk(1)=tmp7 movq mm7,mm4 ; transpose coefficients(phase 2) punpckldq mm4,mm2 ; mm4=(20 21 22 23)=data2 punpckhdq mm7,mm2 ; mm7=(30 31 32 33)=data3 movq mm6,mm1 ; transpose coefficients(phase 2) punpckldq mm1,mm3 ; mm1=(40 41 42 43)=data4 punpckhdq mm6,mm3 ; mm6=(50 51 52 53)=data5 movq mm2,mm7 movq mm3,mm4 paddw mm7,mm1 ; mm7=data3+data4=tmp3 paddw mm4,mm6 ; mm4=data2+data5=tmp2 psubw mm2,mm1 ; mm2=data3-data4=tmp4 psubw mm3,mm6 ; mm3=data2-data5=tmp5 ; -- Even part movq mm1,mm5 movq mm6,mm0 paddw mm5,mm7 ; mm5=tmp10 paddw mm0,mm4 ; mm0=tmp11 psubw mm1,mm7 ; mm1=tmp13 psubw mm6,mm4 ; mm6=tmp12 movq mm7,mm5 paddw mm5,mm0 ; mm5=tmp10+tmp11 psubw mm7,mm0 ; mm7=tmp10-tmp11 paddw mm5,[GOTOFF(ebx,PW_DESCALE_P2X)] paddw mm7,[GOTOFF(ebx,PW_DESCALE_P2X)] psraw mm5,PASS1_BITS ; mm5=data0 psraw mm7,PASS1_BITS ; mm7=data4 movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5 movq MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)], mm7 ; (Original) ; z1 = (tmp12 + tmp13) * 0.541196100; ; data2 = z1 + tmp13 * 0.765366865; ; data6 = z1 + tmp12 * -1.847759065; ; ; (This implementation) ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); movq mm4,mm1 ; mm1=tmp13 movq mm0,mm1 punpcklwd mm4,mm6 ; mm6=tmp12 punpckhwd mm0,mm6 movq mm1,mm4 movq mm6,mm0 pmaddwd mm4,[GOTOFF(ebx,PW_F130_F054)] ; mm4=data2L pmaddwd mm0,[GOTOFF(ebx,PW_F130_F054)] ; mm0=data2H pmaddwd mm1,[GOTOFF(ebx,PW_F054_MF130)] ; mm1=data6L pmaddwd mm6,[GOTOFF(ebx,PW_F054_MF130)] ; mm6=data6H paddd mm4,[GOTOFF(ebx,PD_DESCALE_P2)] paddd mm0,[GOTOFF(ebx,PD_DESCALE_P2)] psrad mm4,DESCALE_P2 psrad mm0,DESCALE_P2 paddd mm1,[GOTOFF(ebx,PD_DESCALE_P2)] paddd mm6,[GOTOFF(ebx,PD_DESCALE_P2)] psrad mm1,DESCALE_P2 psrad mm6,DESCALE_P2 packssdw mm4,mm0 ; mm4=data2 packssdw mm1,mm6 ; mm1=data6 movq MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4 movq MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)], mm1 ; -- Odd part movq mm5, MMWORD [wk(0)] ; mm5=tmp6 movq mm7, MMWORD [wk(1)] ; mm7=tmp7 movq mm0,mm2 ; mm2=tmp4 movq mm6,mm3 ; mm3=tmp5 paddw mm0,mm5 ; mm0=z3 paddw mm6,mm7 ; mm6=z4 ; (Original) ; z5 = (z3 + z4) * 1.175875602; ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; ; z3 += z5; z4 += z5; ; ; (This implementation) ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); movq mm4,mm0 movq mm1,mm0 punpcklwd mm4,mm6 punpckhwd mm1,mm6 movq mm0,mm4 movq mm6,mm1 pmaddwd mm4,[GOTOFF(ebx,PW_MF078_F117)] ; mm4=z3L pmaddwd mm1,[GOTOFF(ebx,PW_MF078_F117)] ; mm1=z3H pmaddwd mm0,[GOTOFF(ebx,PW_F117_F078)] ; mm0=z4L pmaddwd mm6,[GOTOFF(ebx,PW_F117_F078)] ; mm6=z4H movq MMWORD [wk(0)], mm4 ; wk(0)=z3L movq MMWORD [wk(1)], mm1 ; wk(1)=z3H ; (Original) ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; ; ; (This implementation) ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); ; data7 = tmp4 + z3; data5 = tmp5 + z4; ; data3 = tmp6 + z3; data1 = tmp7 + z4; movq mm4,mm2 movq mm1,mm2 punpcklwd mm4,mm7 punpckhwd mm1,mm7 movq mm2,mm4 movq mm7,mm1 pmaddwd mm4,[GOTOFF(ebx,PW_MF060_MF089)] ; mm4=tmp4L pmaddwd mm1,[GOTOFF(ebx,PW_MF060_MF089)] ; mm1=tmp4H pmaddwd mm2,[GOTOFF(ebx,PW_MF089_F060)] ; mm2=tmp7L pmaddwd mm7,[GOTOFF(ebx,PW_MF089_F060)] ; mm7=tmp7H paddd mm4, MMWORD [wk(0)] ; mm4=data7L paddd mm1, MMWORD [wk(1)] ; mm1=data7H paddd mm2,mm0 ; mm2=data1L paddd mm7,mm6 ; mm7=data1H paddd mm4,[GOTOFF(ebx,PD_DESCALE_P2)] paddd mm1,[GOTOFF(ebx,PD_DESCALE_P2)] psrad mm4,DESCALE_P2 psrad mm1,DESCALE_P2 paddd mm2,[GOTOFF(ebx,PD_DESCALE_P2)] paddd mm7,[GOTOFF(ebx,PD_DESCALE_P2)] psrad mm2,DESCALE_P2 psrad mm7,DESCALE_P2 packssdw mm4,mm1 ; mm4=data7 packssdw mm2,mm7 ; mm2=data1 movq MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)], mm4 movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2 movq mm1,mm3 movq mm7,mm3 punpcklwd mm1,mm5 punpckhwd mm7,mm5 movq mm3,mm1 movq mm5,mm7 pmaddwd mm1,[GOTOFF(ebx,PW_MF050_MF256)] ; mm1=tmp5L pmaddwd mm7,[GOTOFF(ebx,PW_MF050_MF256)] ; mm7=tmp5H pmaddwd mm3,[GOTOFF(ebx,PW_MF256_F050)] ; mm3=tmp6L pmaddwd mm5,[GOTOFF(ebx,PW_MF256_F050)] ; mm5=tmp6H paddd mm1,mm0 ; mm1=data5L paddd mm7,mm6 ; mm7=data5H paddd mm3, MMWORD [wk(0)] ; mm3=data3L paddd mm5, MMWORD [wk(1)] ; mm5=data3H paddd mm1,[GOTOFF(ebx,PD_DESCALE_P2)] paddd mm7,[GOTOFF(ebx,PD_DESCALE_P2)] psrad mm1,DESCALE_P2 psrad mm7,DESCALE_P2 paddd mm3,[GOTOFF(ebx,PD_DESCALE_P2)] paddd mm5,[GOTOFF(ebx,PD_DESCALE_P2)] psrad mm3,DESCALE_P2 psrad mm5,DESCALE_P2 packssdw mm1,mm7 ; mm1=data5 packssdw mm3,mm5 ; mm3=data3 movq MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)], mm1 movq MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3 add edx, byte 4*SIZEOF_DCTELEM dec ecx jnz near .columnloop emms ; empty MMX state ; pop edi ; unused ; pop esi ; unused ; pop edx ; need not be preserved ; pop ecx ; need not be preserved poppic ebx mov esp,ebp ; esp <- aligned ebp pop esp ; esp <- original ebp pop ebp ret ; For some reason, the OS X linker does not honor the request to align the ; segment unless we do this. align 16