struct Vector4

{

    float x, y, z, w;

};

struct Matrix

{

    float _M[][];

};

void SSE_VectorMultiplyMatrix(const Vector4& v,const Matrix& m1,Vector4& ret)

{

    Vector4 va,vb,vc,vd;

    Vector4 *pva,*pvb,*pvc,*pvd;

    const Vector4 *pv;

    //取出矩阵每一列

    va.x = m1._M[][];

    va.y = m1._M[][];

    va.z = m1._M[][];

    va.w = m1._M[][];

    vb.x = m1._M[][];

    vb.y = m1._M[][];

    vb.z = m1._M[][];

    vb.w = m1._M[][];

    vc.x = m1._M[][];

    vc.y = m1._M[][];

    vc.z = m1._M[][];

    vc.w = m1._M[][];

    vd.x = m1._M[][];

    vd.y = m1._M[][];

    vd.z = m1._M[][];

    vd.w = m1._M[][];

    pva = &va;

    pvb = &vb;

    pvc = &vc;

    pvd = &vd;

    pv = &v;

    __asm

    {

        //矩阵四列放入mmx0-mmx3

        MOV EAX, pva                           // Load pointer into CPU reg

        MOVUPS XMM0, [EAX]

        MOV EAX, pvb                           // Load pointer into CPU reg

        MOVUPS XMM1, [EAX]

        MOV EAX, pvc                           // Load pointer into CPU reg

        MOVUPS XMM2, [EAX]

        MOV EAX, pvd                           // Load pointer into CPU reg

        MOVUPS XMM3, [EAX]

        //向量放入 mmx4

        MOV EAX, pv

        MOVUPS XMM4, [EAX]

        //向量点乘矩阵每列

        MULPS XMM0,XMM4

        MULPS XMM1,XMM4

        MULPS XMM2,XMM4

        MULPS XMM3,XMM4 

        //输出四个分量

        MOVUPS [va], XMM0

        MOVUPS [vb], XMM1

        MOVUPS [vc], XMM2

        MOVUPS [vd], XMM3

    }

    //四个分量求和得变换后向量

    ret.x = va.w + va.x + va.y + va.z;

    ret.y = vb.w + vb.x + vb.y + vb.z;

    ret.z = vc.w + vc.x + vc.y + vc.z;

    ret.w = vd.w + vd.x + vd.y + vd.z;

}