Буду рад результатам нашего внутренного теста. На данный момент GUI отсуствует, прикрутим позже. Размер ~140 кб.

http://www.uuusoft.com/B.S.E.A.F.R.E.S. ... .E.S.H.zip

Краткое описание.

Модуль тестирования CPU&MEM для системы S.E.A.F.R.E.S.H.
Состоит из двух групп тестов.

Группа А это чистая ситетика, тест скорости выполнения базовой операции (op1 + op2) * op2.
Тип float, реализация через FPU/SSE. Варьируется размер данных, шаг и число потоков (от 1 до 4).

Группа В это реальные алгоритмы для обработки видеоданных.
Тип short, реализация через iMMX/iSSE2/iI64. Варьируется размер данных и шаг.

Время выполнения на P4-3.2 примерно 1.5 мин.
Результат Pr4-3.2/800/1mb

FPU::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 62 281 658
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 61 548 936
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 62 834 501
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 30 027 521
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 29 263 317
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 28 628 770

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 67 832 257
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 66 853 259
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 67 523 547
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 39 527 787
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 39 850 485
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 39 960 239

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 67 279 411
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 66 707 483
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 67 284 160
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 39 834 664
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 40 248 507
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 40 410 133

SSE::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 240 114 458
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 232 673 566
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 226 771 206
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 256 065 891
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 263 913 307
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 279 692 427

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 279 830 400
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 251 090 734
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 251 487 349
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 259 175 789
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 265 533 602
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 267 181 438

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 310 647 213
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 251 470 100
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 229 953 600
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 249 633 611
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 256 101 477
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 271 454 007

MMX::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 911 186 098
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 1 990 045 154
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 697 351 568
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 691 646 878

SSE2::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 2 970 669 810
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 3 109 017 160
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 695 763 526
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 690 057 539

MMX::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 335 218 343
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 329 553 848
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 316 012 690
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 313 288 659

SSE2::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 573 552 909
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 477 392 904
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 441 200 951
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 440 605 036

MMX::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 304 651 684
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 299 791 837
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 288 001 559
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 286 405 092

SSE2::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 522 205 053
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 417 278 312
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 398 621 552
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 398 474 130

MMX::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 219 135 388
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 213 321 518
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 211 655 647
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 211 808 627

SSE2::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 382 567 638
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 355 988 668
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 343 840 079
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 344 956 982

MMX::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 603 411 307
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 495 768 697
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 406 919 386
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 412 667 497

SSE2::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 029 414 019
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 645 389 148
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 446 735 021
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 455 865 142

MMX::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 2 929 646 658
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 3 148 558 316
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 685 690 807
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 682 592 650

SSE2::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 4 779 247 290
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 4 819 778 761
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 700 734 323
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 685 856 187

AMD XP2000+@1664MHz (133x12.5)
DDR333@167MHz 2.5-3-3-7

FPU::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 34 213 030
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 34 507 728
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 34 771 245
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 34 885 112
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 34 928 319
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 34 941 618

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 32 346 508
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 32 937 897
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 32 961 067
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 33 210 926
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 33 273 492
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 33 296 447

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 32 642 221
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 33 161 204
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 33 416 395
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 33 840 462
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 33 995 756
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 34 039 213

SSE::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 115 309 938
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 120 428 250
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 122 206 214
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 125 702 190
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 125 920 596
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 123 355 112

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 109 051 003
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 116 770 737
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 121 466 743
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 124 742 874
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 125 366 327
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 122 957 064

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 104 490 295
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 111 759 319
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 118 463 219
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 122 526 067
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 124 052 012
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 122 488 363

MMX::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 224 443 290
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 308 345 959
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 305 876 229
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 305 529 281

SSE2::Dp = Gradient( Const1, Const2, Const3, Const4 )

MMX::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 326 503 186
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 231 455 507
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 229 862 657
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 230 124 282

SSE2::Dp = Blure( Dp )

MMX::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 312 194 719
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 229 754 655
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 227 082 758
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 227 032 951

SSE2::Dp = Sharp( Dp )

MMX::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 229 400 390
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 184 140 386
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 182 042 868
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 180 111 690

SSE2::Dp = (SUMi[Sp - Spi])/8

MMX::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 523 797 328
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 224 932 959
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 226 416 902
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 225 501 870

SSE2::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

MMX::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 824 834 984
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 403 826 553
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 408 548 038
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 410 756 567

PS Комп на работе

seafresh это надо каждый шаг сравнивать???
хоть рейтинг бы какой нибудь.

резалт

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 75 629 820
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 76 847 669
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 77 239 446
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 37 798 355
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 37 833 537
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 37 854 520

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 76 140 684
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 75 892 191
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 76 725 817
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 37 781 851
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 37 792 708
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 37 830 839

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 76 075 543
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 76 832 128
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 75 826 236
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 37 694 740
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 37 754 972
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 37 812 455

SSE::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 438 633 420
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 322 105 956
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 330 233 738
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 342 416 537
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 346 408 808
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 355 784 411

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 454 331 805
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 309 998 462
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 319 263 806
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 335 166 252
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 342 783 524
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 354 764 765

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 425 238 132
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 314 323 981
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 305 507 061
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 325 604 474
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 337 878 220
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 351 723 782

MMX::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 2 610 874 969
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 2 751 359 975
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 065 643 569
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 063 074 051

SSE2::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 3 851 607 152
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 4 241 675 578
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 076 214 936
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 068 578 485

MMX::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 417 623 913
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 415 560 454
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 405 463 293
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 405 013 898

SSE2::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 702 907 507
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 538 894 610
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 522 792 156
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 523 864 024

MMX::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 385 605 457
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 383 811 615
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 375 332 899
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 374 987 804

SSE2::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 642 652 583
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 469 991 794
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 461 977 756
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 461 826 854

MMX::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 269 308 813
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 266 953 597
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 266 338 076
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 265 936 694

SSE2::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 480 155 040
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 474 463 440
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 466 732 278
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 469 352 093

MMX::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 780 250 115
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 758 426 775
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 617 175 747
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 638 250 153

SSE2::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 354 417 517
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 1 180 700 665
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 671 256 541
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 690 594 189

MMX::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 4 434 841 524
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 4 914 912 934
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 073 609 859
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 072 456 322

SSE2::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 6 509 626 343
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 6 588 817 368
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 066 781 324
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 062 654 080

seafresh согласен с Masterovoj , нужен рейтинг

T2VOVIK интересно, процессор старый,а результат неплохой.
Забыл сказать результат пишется в файл в ту-же директорию и для запуска необходим процессор с поддержкой MMX/SSE. SSE2 не обязательно.
Насчет рейтинга, сейчас прикручу для целочисленных операций.

A64 2800+ @ 2430MHz
Тест не проходит - система виснет.

A64 X2 3800+@@2200MHz
2Gb PC4000@220MHz 3-4-3-7 1T

FPU::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 45 114 200
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 45 335 695
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 45 478 904
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 45 570 441
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 45 655 589
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 45 971 227

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 58 393 874
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 66 198 046
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 65 268 180
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 67 172 119
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 69 183 681
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 73 603 221

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 63 320 785
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 76 583 114
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 76 143 072
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 77 172 987
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 78 677 438
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 81 572 475

SSE::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 136 501 198
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 262 259 131
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 271 171 011
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 274 861 786
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 280 618 933
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 331 699 235

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 170 293 041
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 217 132 624
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 313 580 068
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 357 169 717
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 358 006 446
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 401 019 130

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 174 407 340
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 207 108 450
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 342 348 789
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 357 918 644
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 346 524 145
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 399 747 321

MMX::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 813 368 725
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 1 846 108 372
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 992 864 978
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 980 529 164

SSE2::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 2 608 742 496
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 2 805 948 406
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 113 563 872
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 111 127 537

MMX::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 436 749 471
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 444 994 917
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 425 305 074
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 424 036 837

SSE2::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 475 376 996
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 481 619 021
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 444 331 928
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 443 753 056

MMX::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 423 947 335
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 420 857 226
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 396 821 514
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 393 117 344

SSE2::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 442 228 331
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 450 791 991
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 421 949 971
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 419 789 741

MMX::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 311 422 482
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 287 851 320
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 291 943 299
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 288 703 444

SSE2::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 371 549 668
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 332 161 515
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 338 457 766
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 328 040 153

MMX::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 845 442 638
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 539 985 147
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 543 504 827
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 551 956 852

SSE2::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 049 005 102
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 505 861 120
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 503 576 947
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 508 721 995

MMX::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 2 549 971 316
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 2 685 108 342
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 090 503 059
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 079 758 099

SSE2::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 3 465 071 306
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 3 816 091 052
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 089 579 523
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 079 164 649

PS seafresh С выводом результата надо чего-то делать

Если есть возможность вышлите мне файл [xxx].log.
Я тестировал приложение на P4/P4D/AMD64, вроде все было нормально. Возможно, что вы не дождались окончания тестирования, т.к. приоритет процесса realtime, а потоков high на системе с 1 логическим процессором во время тестирования (3-4мин) система выглядит зависшей.
Добавлено спустя 3 минуты, 51 секунду


Еще раз спасибо за результаты, именно этих (AMDXP/AMDX2) у меня было.
Обновил тест, добавил вывод результатов в конец файла.

Результат PresP4HT-3.2/800/1mb

Result fFPU/fSSE
fFPU [ 45 53 53 ]=152
fSSE [ 241 265 261 ]=767


Result iMMX/iSSE2/i64
iMMX [ 988 1068 435 433 ]=2925
iSSE2 [ 1636 1587 503 501 ]=4228
Добавлено спустя 3 минуты, 21 секунду


Это результат процессора из профиля?

A64 4000+@2760

Result fFPU/fSSE
fFPU [ 58 57 54 ]=170
fSSE [ 310 305 302 ]=917


Result iMMX/iSSE2/i64
iMMX [ 1329 1362 627 628 ]=3948
iSSE2 [ 1746 1845 677 679 ]=4948


лучше б на экран/в окно... а все остальное в файл...

BenchmarkCPU&MEM for S.E.A.F.R.E.H. v0.21 | g_debuge_funct1.cpp[1072][TID::684]
os :: [Win32 on Windows NT/Windows 2000./5.1/Service Pack 2]
cpu :: 1x[Intelx1995.9Mhz][granularity=4096]
memory :: [phys=314769408/535543808][pages=1126391808/1310539776]


FPU::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 22 608 556
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 9 653 938
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 9 542 953
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 9 707 597
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 9 529 960
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 9 026 790

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 22 250 451
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 9 693 243
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 9 739 150
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 9 563 329
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 9 675 850
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 8 958 894

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 21 836 987
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 9 505 825
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 9 637 101
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 9 646 124
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 9 645 241
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 8 880 889

SSE::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 176 705 246
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 180 463 508
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 181 505 097
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 183 482 295
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 188 333 440
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 139 741 459

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 163 724 574
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 172 917 977
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 177 835 847
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 180 868 766


Вот тут мне и надоело ждать выхода из спячки:(

A64 X2 3800+@2702MHz
2Gb PC4000@270MHz 3-4-3-7 1T

FPU::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 55 846 422
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 56 139 701
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 56 322 619
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 56 466 666
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 56 625 383
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 56 854 281

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 107 089 516
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 107 081 511
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 108 676 672
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 108 673 428
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 108 826 847
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 109 433 433

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 94 615 713
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 92 375 491
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 93 215 372
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 93 838 644
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 95 538 732
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 99 276 491

SSE::op1 = (op1 + op2) * op2

Step 0 Threads 1 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 276 945 051
Step 1 Threads 1 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 325 228 589
Step 2 Threads 1 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 336 847 082
Step 3 Threads 1 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 342 949 168
Step 4 Threads 1 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 350 437 053
Step 5 Threads 1 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 413 003 080

Step 0 Threads 2 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 445 128 184
Step 1 Threads 2 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 393 744 404
Step 2 Threads 2 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 424 479 670
Step 3 Threads 2 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 440 214 696
Step 4 Threads 2 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 451 419 424
Step 5 Threads 2 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 501 385 464

Step 0 Threads 4 Size[bytes] 1 048 576 Stride[bytes] 524 288 OpPerSec 440 916 063
Step 1 Threads 4 Size[bytes] 2 097 152 Stride[bytes] 1 048 576 OpPerSec 382 606 512
Step 2 Threads 4 Size[bytes] 4 194 304 Stride[bytes] 2 097 152 OpPerSec 428 673 281
Step 3 Threads 4 Size[bytes] 8 388 608 Stride[bytes] 4 194 304 OpPerSec 437 926 853
Step 4 Threads 4 Size[bytes] 16 777 216 Stride[bytes] 8 388 608 OpPerSec 451 332 669
Step 5 Threads 4 Size[bytes] 33 554 432 Stride[bytes] 16 777 216 OpPerSec 500 290 323

MMX::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 2 215 217 843
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 2 306 466 372
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 246 159 971
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 227 470 019

SSE2::Dp = Gradient( Const1, Const2, Const3, Const4 )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 3 226 712 652
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 3 481 710 924
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 386 034 897
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 397 661 088

MMX::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 532 185 922
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 531 096 959
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 500 327 818
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 497 671 697

SSE2::Dp = Blure( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 600 810 541
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 599 325 978
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 568 402 660
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 566 929 601

MMX::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 504 859 694
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 503 579 094
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 472 381 585
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 475 227 272

SSE2::Dp = Sharp( Dp )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 563 444 738
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 562 295 051
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 528 005 242
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 525 955 280

MMX::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 380 377 636
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 362 203 240
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 367 334 891
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 362 469 299

SSE2::Dp = (SUMi[Sp - Spi])/8

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 440 893 057
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 398 382 243
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 408 055 101
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 409 025 720

MMX::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 019 868 355
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 686 329 278
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 701 286 679
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 692 273 375

SSE2::Dp = Dp + ((Dp * ((Sp > Const1 ? Sp : 0) >> 1)) >> Const2)

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 1 256 249 683
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 675 502 426
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 695 748 977
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 677 871 283

MMX::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 3 137 007 298
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 3 338 964 474
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 366 735 456
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 351 238 880

SSE2::Dp[ x, y ] = unsignsat( Const2 - 2 * ( Const1 - y ) * ( Const1 - y ) )

Step 0 Threads 1 Width/Height 256 Size[bytes] 131 072 PixelPerSec 4 319 862 490
Step 1 Threads 1 Width/Height 512 Size[bytes] 524 288 PixelPerSec 4 713 681 465
Step 2 Threads 1 Width/Height 1 024 Size[bytes] 2 097 152 PixelPerSec 1 406 781 924
Step 3 Threads 1 Width/Height 2 048 Size[bytes] 8 388 608 PixelPerSec 1 392 798 897


Result fFPU/fSSE
fFPU [ 56 108 94 ]=259
fSSE [ 340 442 440 ]=1223


Result iMMX/iSSE2/i64
iMMX [ 1298 1288 775 767 ]=4129
iSSE2 [ 1734 1738 832 828 ]=5133

PS Чего-то стало очень долго...

Да. А новую версию, к сожалению, смогу протестировать только вечером.

seafresh уточните время выполнения теста, у меня ~4 минуты

A64 X2 3800+@2709MHz
2Gb PC4000@271MHz 3-4-3-7 1T

#77

Result fFPU/fSSE
fFPU [ 56 107 95 ]=259
fSSE [ 341 436 434 ]=1212


Result iMMX/iSSE2/i64
iMMX [ 1302 1287 775 771 ]=4136
iSSE2 [ 1739 1742 832 831 ]=5146

В целях уменьшения погрешности я увеличил время тестирования второй группы тестов. Вне зависимости от скорости процессора время второй группы составляет 4 минуты ровно. Время теста первой группы (FPU) зависит от скорости процессора и равно, на P4-3.2/800/1mb, 15-20 секундам.

AMD XP2000+@1664MHz (133x12.5)
DDR333@167MHz 2.5-3-3-7

Result fFPU/fSSE
fFPU [ 34 33 32 ]=101
fSSE [ 121 119 116 ]=357


Result iMMX/iSSE2/i64
iMMX [ 766 262 262 262 ]=1554
iSSE2 [ 0 0 0 0 ]=0

A64 X2 3800+@1665MHz
2Gb PC4000@166MHz 3-3-3-7 1T

Result fFPU/fSSE
fFPU [ 33 67 59 ]=160
fSSE [ 207 296 294 ]=798


Result iMMX/iSSE2/i64
iMMX [ 796 795 481 479 ]=2553
iSSE2 [ 1065 1073 520 516 ]=3177

BenchmarkCPU&MEM for S.E.A.F.R.E.H. v0.21 | g_debuge_funct1.cpp[1072][TID::4048]
os :: [Win32 on Windows NT/Windows 2000./5.1/Service Pack 2]
cpu :: 1x[Intelx1425.1Mhz][granularity=4096]
memory :: [phys=783753216/1073201152][pages=2385027072/2585575424]
display :: 1x[SM_CFULLSCREEN/SM_CSCREEN/SM_CVIRTUALSCREEN=1280x900/1280x960/1280x960] 32bitsx[SHADEBLENDCAPS/RASTERCAPS/CURVECAPS/LINECAPS/POLYGONALCAPS/TEXTCAPS/COLORMGMTCAPS=0x3/0x7e99/0x1ff/0xfe/0xff/0x7807/0x2]

Result fFPU/fSSE
fFPU [ 29 28 28 ]=86
fSSE [ 99 98 96 ]=295

Result iMMX/iSSE2/i64
iMMX [ 649 227 226 227 ]=1330
iSSE2 [ 0 0 0 0 ]=0

19.5.2006 - 18:53:38:718 : done. | BSEAFRESH.cpp [711] [TID::4048]
Все бы неплохо, но у меня Атлон 2000+ @2600+, стартовая со 114 шины и до 171, тест поймал частоту на старте, поэтому 1425мГц, но не интел же:)!

Учел рекомендации T2VOVIK. Уменьшил время тестирования, на P4-3.2 примерно 2 минуты, добавил GUI, изменил набор функций во второй группе, в силу этого результаты не соответствуют предыдущей версии.
Размер файла увеличился до 300 кб.

Краткое описание.
Модуль тестирования CPU&Mem для системы S.E.A.F.R.E.S.H.
Состоит из двух групп тестов.

Группа А тип данных float. Время выполнения зависит от скорости процессора [P4HT-3.2/800/1mb - 15-20 секунд].
Включает в себя
1. Cинтетику, тест скорости выполнения базовой операции (op1 + op2) * op2.
Реализация через FPU/SSE. Варьируется размер данных, шаг и число потоков (от 1 до 4).

Группа В тип данных short. Время выполнения не зависит от скорости процессора и примерно равно 2 минутам.
Включает в себя
1. Полусинтетику, алгоритмы обработки данных, чистый asmx86[iMMX/iSSE2/iI64]. Варьируется размер входных данных.
Dp[ x, y ] = Gradient( Const1, Const2, Const3, Const4, x, y )
Dp[ x, y ] = Blure( Dp[ x, y ] )
Dp[ x, y ] = Sharp( Dp[ x, y ] )
Dp[ x, y ] = (SUMi[ SUMj[Sp[ x, y ] - Sp[ x+i, y+j ]] ]) >> 3
Dp[ x, y ] = Dp[ x, y ] + ((Dp[ x, y ] * ((Sp[ x, y ] > Const1 ? Sp[ x, y ] : 0) >> 1)) >> Const2)
Dp[ x, y ] = ( abs( Const1 - x ) + abs( Const2 - y ) ) >> Const3

2. Алгоритм векторизации контуров c рассчетом площадей и векторов скорости. Смесь C++ и asmx86[iMMX/iSSE2/iI64]. Варьируется размер входных данных.
Используется в детекторе движения.

Для работы необходим CPU с поддержкой, как минимум, MMX/SSE.
Подробный отчет пишется в файл.

Результат P4HT-3.2/800/1mb/2Gb
[img]http://www.uuusoft.com/B.S.E.A.F.R.E.S.H./BSEAFRESH[P4-3.2-800-1mb-2Gb].jpg[/img]

FFPU [ 36 50 49 ]=139
FSSE [ 159 183 169 ]=512
IMMX [ 583 524 310 310 ]=1728
ISSE2 [ 934 803 366 365 ]=2469

Результат P4-3.2/800/1mb/2Gb
[img]http://www.uuusoft.com/B.S.E.A.F.R.E.S.H./BSEAFRESH[P4xHT-3.2-800-1mb-2Gb].JPG[/img]

FFPU [ 47 46 46 ]=140
FSSE [ 216 212 207 ]=636
IMMX [ 716 724 429 414 ]=2284
ISSE2 [ 1114 1063 478 467 ]=3124

Гы, рисовал сам, т.к. художники по выходным не работают. Старался как мог.

cpu :: 1x[Intelx1607.0Mhz][granularity=4096]
memory :: [phys=280039424/536317952][pages=776351744/1029353472]
display :: 1x[SM_CFULLSCREEN/SM_CSCREEN/SM_CVIRTUALSCREEN=1280x977/1280x1024/1280x1024] 32bitsx[SHADEBLENDCAPS/RASTERCAPS/CURVECAPS/LINECAPS/POLYGONALCAPS/TEXTCAPS/COLORMGMTCAPS=0x3/0x7e99/0x1ff/0xfe/0xff/0x7807/0x2]


Result fFPU/fSSE
FFPU [ 33 33 33 ]=99
FSSE [ 184 181 177 ]=543


Result iMMX/iSSE2/i64
IMMX [ 511 341 339 339 ]=1531
ISSE2 [ 582 349 352 350 ]=1635
Добавлено спустя 1 минуту, 8 секунд
cpu sempron 2600 po default'y

гы, исправим, спасибо.