sad_ia64.s

wince下的xvidcore开发库,可用于MP4等视频播放开发

        shrp r66 = r67, r66, 32
        shrp r67 = r97, r67, 32
        shrp r68 = r69, r68, 32
        shrp r69 = r98, r69, 32
        shrp r70 = r71, r70, 32
        shrp r71 = r99, r71, 32
        shrp r72 = r73, r72, 32
        shrp r73 = r100, r73, 32
        shrp r74 = r75, r74, 32
        shrp r75 = r101, r75, 32
        shrp r76 = r77, r76, 32
        shrp r77 = r102, r77, 32
        shrp r78 = r79, r78, 32
        shrp r79 = r103, r79, 32
        shrp r80 = r81, r80, 32
        shrp r81 = r104, r81, 32
        shrp r82 = r83, r82, 32
        shrp r83 = r105, r83, 32
        shrp r84 = r85, r84, 32
        shrp r85 = r106, r85, 32
        shrp r86 = r87, r86, 32
        shrp r87 = r107, r87, 32
        shrp r88 = r89, r88, 32
        shrp r89 = r108, r89, 32
        shrp r90 = r91, r90, 32
        shrp r91 = r109, r91, 32
        shrp r92 = r93, r92, 32
        shrp r93 = r110, r93, 32
        shrp r94 = r95, r94, 32
        shrp r95 = r111, r95, 32
        br.cond.sptk.many .Lber
        ;;
.Lmod3:
        shrp r64 = r65, r64, 24
        shrp r65 = r96, r65, 24
        shrp r66 = r67, r66, 24
        shrp r67 = r97, r67, 24
        shrp r68 = r69, r68, 24
        shrp r69 = r98, r69, 24
        shrp r70 = r71, r70, 24
        shrp r71 = r99, r71, 24
        shrp r72 = r73, r72, 24
        shrp r73 = r100, r73, 24
        shrp r74 = r75, r74, 24
        shrp r75 = r101, r75, 24
        shrp r76 = r77, r76, 24
        shrp r77 = r102, r77, 24
        shrp r78 = r79, r78, 24
        shrp r79 = r103, r79, 24
        shrp r80 = r81, r80, 24
        shrp r81 = r104, r81, 24
        shrp r82 = r83, r82, 24
        shrp r83 = r105, r83, 24
        shrp r84 = r85, r84, 24
        shrp r85 = r106, r85, 24
        shrp r86 = r87, r86, 24
        shrp r87 = r107, r87, 24
        shrp r88 = r89, r88, 24
        shrp r89 = r108, r89, 24
        shrp r90 = r91, r90, 24
        shrp r91 = r109, r91, 24
        shrp r92 = r93, r92, 24
        shrp r93 = r110, r93, 24
        shrp r94 = r95, r94, 24
        shrp r95 = r111, r95, 24
        br.cond.sptk.many .Lber
        ;;
.Lmod2:
        shrp r64 = r65, r64, 16
        shrp r65 = r96, r65, 16
        shrp r66 = r67, r66, 16
        shrp r67 = r97, r67, 16
        shrp r68 = r69, r68, 16
        shrp r69 = r98, r69, 16
        shrp r70 = r71, r70, 16
        shrp r71 = r99, r71, 16
        shrp r72 = r73, r72, 16
        shrp r73 = r100, r73, 16
        shrp r74 = r75, r74, 16
        shrp r75 = r101, r75, 16
        shrp r76 = r77, r76, 16
        shrp r77 = r102, r77, 16
        shrp r78 = r79, r78, 16
        shrp r79 = r103, r79, 16
        shrp r80 = r81, r80, 16
        shrp r81 = r104, r81, 16
        shrp r82 = r83, r82, 16
        shrp r83 = r105, r83, 16
        shrp r84 = r85, r84, 16
        shrp r85 = r106, r85, 16
        shrp r86 = r87, r86, 16
        shrp r87 = r107, r87, 16
        shrp r88 = r89, r88, 16
        shrp r89 = r108, r89, 16
        shrp r90 = r91, r90, 16
        shrp r91 = r109, r91, 16
        shrp r92 = r93, r92, 16
        shrp r93 = r110, r93, 16
        shrp r94 = r95, r94, 16
        shrp r95 = r111, r95, 16
        br.cond.sptk.many .Lber
        ;;
.Lmod1:
        shrp r64 = r65, r64, 8
        shrp r65 = r96, r65, 8
        shrp r66 = r67, r66, 8
        shrp r67 = r97, r67, 8
        shrp r68 = r69, r68, 8
        shrp r69 = r98, r69, 8
        shrp r70 = r71, r70, 8
        shrp r71 = r99, r71, 8
        shrp r72 = r73, r72, 8
        shrp r73 = r100, r73, 8
        shrp r74 = r75, r74, 8
        shrp r75 = r101, r75, 8
        shrp r76 = r77, r76, 8
        shrp r77 = r102, r77, 8
        shrp r78 = r79, r78, 8
        shrp r79 = r103, r79, 8
        shrp r80 = r81, r80, 8
        shrp r81 = r104, r81, 8
        shrp r82 = r83, r82, 8
        shrp r83 = r105, r83, 8
        shrp r84 = r85, r84, 8
        shrp r85 = r106, r85, 8
        shrp r86 = r87, r86, 8
        shrp r87 = r107, r87, 8
        shrp r88 = r89, r88, 8
        shrp r89 = r108, r89, 8
        shrp r90 = r91, r90, 8
        shrp r91 = r109, r91, 8
        shrp r92 = r93, r92, 8
        shrp r93 = r110, r93, 8
        shrp r94 = r95, r94, 8
        shrp r95 = r111, r95, 8
.Lber:
        ;;
        psad1 r32 = r32, r64            // Here we do the calculation.
        psad1 r33 = r33, r65            // The machine is providing a fast method
        psad1 r34 = r34, r66            // for calculating sad, so we use it
        psad1 r35 = r35, r67
        psad1 r36 = r36, r68
        psad1 r37 = r37, r69
        psad1 r38 = r38, r70
        psad1 r39 = r39, r71
        psad1 r40 = r40, r72
        psad1 r41 = r41, r73
        psad1 r42 = r42, r74
        psad1 r43 = r43, r75
        psad1 r44 = r44, r76
        psad1 r45 = r45, r77
        psad1 r46 = r46, r78
        psad1 r47 = r47, r79
        psad1 r48 = r48, r80
        psad1 r49 = r49, r81
        psad1 r50 = r50, r82
        psad1 r51 = r51, r83
        psad1 r52 = r52, r84
        psad1 r53 = r53, r85
        psad1 r54 = r54, r86
        psad1 r55 = r55, r87
        psad1 r56 = r56, r88
        psad1 r57 = r57, r89
        psad1 r58 = r58, r90
        psad1 r59 = r59, r91
        psad1 r60 = r60, r92
        psad1 r61 = r61, r93
        psad1 r62 = r62, r94
        psad1 r63 = r63, r95
        ;;
        add r32 = r32, r63              // at last, we have to sum up
        add r33 = r33, r62              // in 5 stages
        add r34 = r34, r61
        add r35 = r35, r60
        add r36 = r36, r59
        add r37 = r37, r58
        add r38 = r38, r57
        add r39 = r39, r56
        add r40 = r40, r55
        add r41 = r41, r54
        add r42 = r42, r53
        add r43 = r43, r52
        add r44 = r44, r51
        add r45 = r45, r50
        add r46 = r46, r49
        add r47 = r47, r48
        ;;
        add r32 = r32, r47
        add r33 = r33, r46
        add r34 = r34, r45
        add r35 = r35, r44
        add r36 = r36, r43
        add r37 = r37, r42
        add r38 = r38, r41
        add r39 = r39, r40
        ;;
        add r32 = r32, r39
        add r33 = r33, r38
        add r34 = r34, r37
        add r35 = r35, r36
        ;;
        add r32 = r32, r35
        add r33 = r33, r34
        ;;
        add r8 = r32, r33               // and store the result in r8
        mov pr = r2, -1
        mov ar.pfs = r1
        br.ret.sptk.many b0
        .endp sad16_ia64#



        
        .align 16
        .global sad8_ia64#
        .proc sad8_ia64#
sad8_ia64:
        alloc r1 = ar.pfs, 3, 21, 0, 0
        mov r2 = pr
         dep r14 = r0, r33, 0, 3 // calculate aligned version of ref
         dep.z r31 = r33, 0, 3           // calculate misalignment of ref
         ;;
         mov r40 = r34           //(1) calculate multiples of stride
         shl r41 = r34, 1                //(2)
         shladd r42 = r34, 1, r34        //(3)
         shl r43 = r34, 2                //(4)
         shladd r44 = r34, 2, r34        //(5)
         ;;
         cmp.eq p16, p17 = 0, r31        // set predicates according to the misalignment of ref
         cmp.eq p18, p19 = 2, r31
         shl r45 = r42, 1                //(6)
         cmp.eq p20, p21 = 4, r31
         cmp.eq p22, p23 = 6, r31
         shladd r46 = r42, 1, r34        //(7)
         cmp.eq p24, p25 = 1, r31
         cmp.eq p26, p27 = 3, r31
         cmp.eq p28, p29 = 5, r31
         ;;
         mov r48 = r14           // calculate memory adresses of data
         add r33 = r32, r40
         add r49 = r14, r40
         add r34 = r32, r41
         add r50 = r14, r41
         add r35 = r32, r42
         add r51 = r14, r42
         add r36 = r32, r43
         add r52 = r14, r43
         add r37 = r32, r44
         add r53 = r14, r44
         add r38 = r32, r45
         add r54 = r14, r45
         add r39 = r32, r46
         add r55 = r14, r46
         ;;
         ld8 r32 = [r32]         // load everythingund alles wird geladen
         ld8 r33 = [r33]         // cur is located in r32 - r39
         ld8 r34 = [r34]         // ref in r40 - r47
         ld8 r35 = [r35]
         ld8 r36 = [r36]
         ld8 r37 = [r37]
         ld8 r38 = [r38]
         ld8 r39 = [r39]
         ld8 r40 = [r48] ,8
         ld8 r41 = [r49] ,8
         ld8 r42 = [r50] ,8
         ld8 r43 = [r51] ,8
         ld8 r44 = [r52] ,8
         ld8 r45 = [r53] ,8
         ld8 r46 = [r54] ,8
         ld8 r47 = [r55] ,8
         (p16) br.cond.dptk.many .Lber2  // if ref is aligned, we can start the calculation
         ;;
         ld8 r48 = [r48]         // if not, we have to load some more
         ld8 r49 = [r49]         // because of the alignment of ld8
         ld8 r50 = [r50]
         ld8 r51 = [r51]
         ld8 r52 = [r52]
         ld8 r53 = [r53]
         ld8 r54 = [r54]
         ld8 r55 = [r55]
         (p24) br.cond.dptk.many .Lmode1
         (p18) br.cond.dpnt.many .Lmode2
         (p26) br.cond.dpnt.many .Lmode3
         (p20) br.cond.dpnt.many .Lmode4
         (p28) br.cond.dpnt.many .Lmode5
         (p22) br.cond.dpnt.many .Lmode6
         ;;
 .Lmode7:                                // this jump piont is not needed, it is for better understandment
         shrp r40 = r48, r40, 56 // here we do some preprocessing on the data
         shrp r41 = r49, r41, 56 // this is because of the alignment problem of ref
         shrp r42 = r50, r42, 56
         shrp r43 = r51, r43, 56
         shrp r44 = r52, r44, 56
         shrp r45 = r53, r45, 56
         shrp r46 = r54, r46, 56
         shrp r47 = r55, r47, 56
         br.cond.sptk.many .Lber2
         ;;
 .Lmode6:
         shrp r40 = r48, r40, 48
         shrp r41 = r49, r41, 48
         shrp r42 = r50, r42, 48
         shrp r43 = r51, r43, 48
         shrp r44 = r52, r44, 48
         shrp r45 = r53, r45, 48
         shrp r46 = r54, r46, 48
         shrp r47 = r55, r47, 48
         br.cond.sptk.many .Lber2
         ;;
 .Lmode5:
         shrp r40 = r48, r40, 40
         shrp r41 = r49, r41, 40
         shrp r42 = r50, r42, 40
         shrp r43 = r51, r43, 40
         shrp r44 = r52, r44, 40
         shrp r45 = r53, r45, 40
         shrp r46 = r54, r46, 40
         shrp r47 = r55, r47, 40
         br.cond.sptk.many .Lber2
         ;;
 .Lmode4:
         shrp r40 = r48, r40, 32
         shrp r41 = r49, r41, 32
         shrp r42 = r50, r42, 32
         shrp r43 = r51, r43, 32
         shrp r44 = r52, r44, 32
         shrp r45 = r53, r45, 32
         shrp r46 = r54, r46, 32
         shrp r47 = r55, r47, 32
         br.cond.sptk.many .Lber2
         ;;
 .Lmode3:
         shrp r40 = r48, r40, 24
         shrp r41 = r49, r41, 24
         shrp r42 = r50, r42, 24
         shrp r43 = r51, r43, 24
         shrp r44 = r52, r44, 24
         shrp r45 = r53, r45, 24
         shrp r46 = r54, r46, 24
         shrp r47 = r55, r47, 24
         br.cond.sptk.many .Lber2
         ;;
 .Lmode2:
         shrp r40 = r48, r40, 16
         shrp r41 = r49, r41, 16
         shrp r42 = r50, r42, 16
         shrp r43 = r51, r43, 16
         shrp r44 = r52, r44, 16
         shrp r45 = r53, r45, 16
         shrp r46 = r54, r46, 16
         shrp r47 = r55, r47, 16
         br.cond.sptk.many .Lber2
         ;;
 .Lmode1:
         shrp r40 = r48, r40, 8
         shrp r41 = r49, r41, 8
         shrp r42 = r50, r42, 8
         shrp r43 = r51, r43, 8
         shrp r44 = r52, r44, 8
         shrp r45 = r53, r45, 8
         shrp r46 = r54, r46, 8
         shrp r47 = r55, r47, 8
 .Lber2:
         ;;
         psad1 r32 = r32, r40    // we start calculating sad
         psad1 r33 = r33, r41    // using th psad1 command of IA64
         psad1 r34 = r34, r42
         psad1 r35 = r35, r43
         psad1 r36 = r36, r44
         psad1 r37 = r37, r45
         psad1 r38 = r38, r46
         psad1 r39 = r39, r47
         ;;
         add r32 = r32, r33              // then we sum up everything
         add r33 = r34, r35
         add r34 = r36, r37
         add r35 = r38, r39
         ;;
         add r32 = r32, r33
         add r33 = r34, r35
         ;;
         add r8 = r32, r33               // and store the result un r8
         mov pr = r2, -1
         mov ar.pfs = r1
         br.ret.sptk.many b0
         .endp sad8_ia64#