乒乓--平滑,翻转

// 一行行平滑

#define MAX_W 2592
static unsigned char inBuf[MAX_W*4];  // in L2 SRAM
static unsigned char outBuf[MAX_W*2]; // in L2 SRAM

static void smooth_one_line( const unsigned char* restrict in1, const unsigned char* restrict in2, const unsigned char* restrict in3, unsigned char* outLine, short outLen )
{
    short j;
    static char mask[16] = { 1,2,1, 2,4,2, 1,2,1,  0,0,0,0,0,0,0 };

    unsigned int  mask1_0, mask2_0, mask3_0;
    unsigned int  mask1_1, mask2_1, mask3_1;

    double r1_76543210, r2_76543210, r3_76543210;
    unsigned int r1_7654, r1_3210;
    unsigned int r2_7654, r2_3210;
    unsigned int r3_7654, r3_3210;
    unsigned int r1_5432, r2_5432, r3_5432;

    int sum0, sum1, sum2, sum3;
    unsigned int  sum32, sum10, sum3210;

    // get 4 bytes, shift left 8 bits, shift right 8 bits 
    mask1_0 = _extu( _mem4_const(&mask[0]), 8, 8);
    mask2_0 = _extu( _mem4_const(&mask[3]), 8, 8);
    mask3_0 = _extu( _mem4_const(&mask[6]), 8, 8);

    mask1_1 = mask1_0 << 8;
    mask2_1 = mask2_0 << 8;
    mask3_1 = mask3_0 << 8;

    // the first pixel is *IN2
    sum3 = *in2;
    for( j=0; j<outLen-7; j+=4 )
    {
        r1_76543210 = _memd8_const(&in1[j]);
        r2_76543210 = _memd8_const(&in2[j]);
        r3_76543210 = _memd8_const(&in3[j]);

        r1_3210 = _lo(r1_76543210);
        r2_3210 = _lo(r2_76543210);
        r3_3210 = _lo(r3_76543210);
        r1_7654 = _hi(r1_76543210);
        r2_7654 = _hi(r2_76543210);
        r3_7654 = _hi(r3_76543210);

        sum0 = (_dotpsu4(mask1_0, r1_3210) + _dotpsu4(mask2_0, r2_3210)
                + _dotpsu4(mask3_0, r3_3210) + 8 ) >> 4;

        sum1 = (_dotpsu4(mask1_1, r1_3210) + _dotpsu4(mask2_1, r2_3210)
                + _dotpsu4(mask3_1, r3_3210) + 8 ) >> 4;

//        sum10 = _spack2(sum1, sum0);
        sum10 = _spack2(sum0, sum3);

        r1_5432 = _packlh2(r1_7654, r1_3210);
        r2_5432 = _packlh2(r2_7654, r2_3210);
        r3_5432 = _packlh2(r3_7654, r3_3210);

        sum2 = (_dotpsu4(mask1_0, r1_5432) + _dotpsu4(mask2_0, r2_5432)
                + _dotpsu4(mask3_0, r3_5432) + 8 ) >> 4;

        sum3 = (_dotpsu4(mask1_1, r1_5432) + _dotpsu4(mask2_1, r2_5432)
                + _dotpsu4(mask3_1, r3_5432) + 8 ) >> 4;

//        sum32 = _spack2(sum3, sum2);
        sum32 = _spack2(sum2, sum1);

        sum3210 = _spacku4(sum32, sum10);

        _mem4(&outLine[j]) = sum3210;
    }

    // process the rest
    outLine[j] = sum3;
    for( ; j<outLen-1; j++ )
    {
        outLine[j] = ( in1[j-1] + in1[j]*2 + in1[j+1] + in2[j-1]*2 + in2[j]*4 + in2[j+1]*2 + in3[j-1] + in3[j]*2 + in3[j+1] + 8 ) / 16;
    }
    outLine[j] = in2[j];

    return;
}

void smooth_pingpang( unsigned char* inPtr, short inW, unsigned char* outPtr, short outW,short outH )
{
    // ...
    short i,t;
    unsigned char *p1, *p2, *p3, *p4,*pTmp;
    unsigned char *pout1, *pout2;
    unsigned char *pS, *pD;

    // ...
    p1 = inPtr;
    p2 = inBuf;
    DMA_FAST1DTO1D(p2,p1,outW);
    p2 += MAX_W;
    p1 += inW;
    DMA_FAST1DTO1D(p2,p1,outW);
    p2 += MAX_W;
    p1 += inW;
    DMA_FAST1DTO1D(p2,p1,outW);

    //TODO wait


    p1 = inBuf;
    p2 = p1+MAX_W;
    p3 = p2+MAX_W;
    p4 = p3+MAX_W;
    pS = inPtr+3*inW;
    pD = outPtr;
    pout1 = outBuf;
    pout2 = pout1+MAX_W;
    // p1 --> out frist
    DMA_FAST1DTO1D(pD,p1,outW);
    pD += inW;
    t = (int)((outH-3)/2);
    // process p1,p2,p3 => out 1
    smooth_one_line( p1, p2, p3, pout1, outW );

    for( i=0; i<t; i++ ) //循环次数保证 剩余偶数 如果单数 外面多拷贝一行 TODO?????
    {
        // DMA, ext mem -> inBuff2, outW
        DMA_FAST1DTO1D(p4,pS,inW);
        pS+=inW;
        // DMA, out 1 -> ext mem, outW
        DMA_FAST1DTO1D(pD,pout1,outW);
        pD += inW;
        pTmp = p1;
        p1 = p2;
        p2 = p3;
        p3 = p4;
        p4 = pTmp;
        // process( 1, 2, 3 ) >>out 2
        smooth_one_line( p1, p2, p3, pout2, outW );

        // DMA, ext mem -> inBufA
        DMA_FAST1DTO1D(p4,pS,inW);
        pS += inW;
        // DMA, out2 -> ext mem
        DMA_FAST1DTO1D(pD,pout2,outW);
        pD += inW;
        pTmp = p1;
        p1 = p2;
        p2 = p3;
        p3 = p4;
        p4 = pTmp;

        // process( 1, 2, 3 ), >> out 1
        smooth_one_line( p1, p2, p3, pout1, outW);

    }   
    if(0==((outH-3)%2))
    {
        //拷贝最后一行
        // p1 -> out last 1最后一行的拷贝
        pD += inW;
        DMA_FAST1DTO1D( pD, p3, outW );
    }
    else
    {
        DMA_FAST1DTO1D(pD,pout1,outW);
        pD += inW;
        //处理倒数第二行
        //process( 1, 2, 3 );
        smooth_one_line( p1, p2, p3, pout2, inW );
        // p1 -> out last 1最后一行的拷贝
        //拷贝最后两行
        DMA_FAST1DTO1D( pD, pout2, outW );
        pD += inW;
        DMA_FAST1DTO1D( pD, p3, outW );
    }

    ITG_DMA_FastWait( 63 );

    return;
}

void smooth_pingpang_upsetdown( unsigned char* inPtr, short inW, unsigned char* outPtr, short outW, short outH )
{
    // ...
    short i,t;
    unsigned char *p1, *p2, *p3, *p4,*pTmp;
    unsigned char *pout1, *pout2;
    unsigned char *pS, *pD;

    // ...
    p1 = inPtr;
    p2 = inBuf;
    DMA_FAST1DTO1D(p2,p1,outW);
    p2 += MAX_W;
    p1 += inW;
    DMA_FAST1DTO1D(p2,p1,outW);
    p2 += MAX_W;
    p1 += inW;
    DMA_FAST1DTO1D(p2,p1,outW);

    //TODO wait
    p1 = inBuf;
    p2 = p1+MAX_W;
    p3 = p2+MAX_W;
    p4 = p3+MAX_W;
    pS = inPtr+3*inW;
    pD = outPtr + outH*outW;//.........
    pout1 = outBuf;
    pout2 = pout1+MAX_W;
    // p1 --> out frist
    pD -= outW;//.........

    t = (int)((outH-2)/2);
    //TODO process p1,p2,p3 => out 1

    for( i=0; i<t; i+=2 ) //循环次数保证 剩余偶数 如果单数 外面多拷贝一行 TODO?????
    {
        // DMA, ext mem -> inBuff2, outW
        DMA_FAST1DTO1D(p4,pS,outW);
        pS+=inW;
        // DMA, out 1 -> ext mem, outW
        DMA_FAST1DTO1D(pD,pout1,outW);
        pD -= outW;//.........
        pTmp = p1;
        p1 = p2;
        p2 = p3;
        p3 = p4;
        p4 = pTmp;
        // process( 1, 2, 3 ) >>out 2
        // DMA, ext mem -> inBufA
        DMA_FAST1DTO1D(p4,pS,outW);
        pS += inW;
        // DMA, out2 -> ext mem
        DMA_FAST1DTO1D(pD,pout2,outW);
        pD -= outW;//.........
        pTmp = p1;
        p1 = p2;
        p2 = p3;
        p3 = p4;
        p4 = pTmp;

        // process( 1, 2, 3 ), >> out 1

    }
    DMA_FAST1DTO1D(pD,p1,outW);

    if(0==((outH-2)%2))
    {
        DMA_FAST1DTO1D( pD, p4, outW );//拷贝最后一行
        // p1 -> out last 1最后一行的拷贝
    }
    else
    {
        //处理倒数第二行


        // p1 -> out last 1最后一行的拷贝
        //拷贝最后两行
        DMA_FAST1DTO1D( pD, p2, outW );
        pD+=outW;
        DMA_FAST1DTO1D( pD, p3, outW );

    }

    ITG_DMA_FastWait( 63 );

    return;
}