h264dsputil.c

君正早期ucos系统(只有早期的才不没有打包成库),MPLAYER,文件系统,图片解码,浏览,电子书,录音,想学ucos,识货的人就下吧 russblock fmradio explore set

        S32SFL (xr0,  xr15, xr15, xr15, ptn3);  // xr15: tc0[i]
//
        tpix = pix;
        S32LDD  (xr2, tpix,  0);          // xr2: q3,q2,q1,q0
        S32LDD  (xr1, tpix, -4);          // xr1: p0,p1,p2,p3
        S32LDIV (xr4, tpix, stride, 0);    // xr4: q3',q2',q1',q0'
        S32LDD  (xr3, tpix, -4);          // xr3: p0',p1',p2',p3'
//
	S32SFL  (xr6, xr4, xr2, xr8, ptn0); // xr6: q3'q3q2'q2, xr8:q1'q1q0'q0
	S32SFL  (xr5, xr3, xr1, xr7, ptn0); // xr5: p0'p0p1'p1, xr7:p2'p2p3'p3
//
        S32LDIV (xr2, tpix, stride, 0);    // xr2: q3",q2",q1",q0"
        S32LDD  (xr1, tpix, -4);          // xr1: p0",p1",p2",p3"
        S32LDIV (xr4, tpix, stride, 0);    // xr4: q3^,q2^,q1^,q0^
        S32LDD  (xr3, tpix, -4);          // xr3: p0^,p1^,p2^,p3^
//
	S32SFL  (xr4, xr4, xr2, xr2, ptn0); // xr4: q3^q3"q2^q2", xr2:q1^q1"q0^q0"
	S32SFL  (xr3, xr3, xr1, xr1, ptn0); // xr3: p0^p0"p1^p1", xr1:p2^p2"p3^p3"
//
        S32SFL  (xr10, xr3, xr5, xr11, ptn3); // xr10: p0^p0"p0'p0, xr11:p1^p1"p1'p1
        S32SFL  (xr12, xr1, xr7, xr0,  ptn3); // xr12: p2^p2"p2'p2
        S32SFL  (xr1,  xr2, xr8, xr9,  ptn3); // xr1: q1^q1"q1'q1, xr9:q0^q0"q0'q0
        S32SFL  (xr0,  xr4, xr6, xr2,  ptn3); // xr2: q2^q2"q2'q2
//
        Q8ABD(xr3, xr10, xr9 );   // FFABS (p0 - q0)
        Q8ABD(xr4, xr11, xr10);   // FFABS (p1 - p0)
        Q8ABD(xr5, xr1,  xr9 );   // FFABS (q1 - q0)
// FFABS(p0 - q0) - alpha, FFABS(p1 - p0) - beta
        Q8ADDE_SS (xr6, xr3, xr13,  xr7);  // FFABS(p0 - q0) - alpha
        Q8ADDE_SS (xr3, xr4, xr14,  xr4);  // FFABS(p1 - p0) - beta
        Q16SLR (xr6, xr6, xr7, xr7, 15);   // 1: < 0 (FFABS( p0 - q0 ) < alpha)
        Q16SLR (xr3, xr3, xr4, xr4, 15);   // 1: < 0 (FFABS( p1 - p0 ) < beta)
        Q16SAT (xr4, xr3, xr4);   // xr4: 1: < 0 (FFABS( p1 - p0 ) < beta)
        Q16SAT (xr3, xr6, xr7);   // xr3: 1: < 0 (FFABS( p0 - q0 ) < alpha)
// FFABS(q1 - q0) - beta
        Q8ADDE_SS (xr6, xr5, xr14,  xr7);  // FFABS(q1 - q0) - beta
        Q16SLR (xr6, xr6, xr7, xr7, 15);   // 1: < 0 (FFABS( q1 - q0 ) < beta)
        Q16SAT (xr5, xr6, xr7);   // xr5: 1: < 0 (FFABS( q1 - q0 ) < beta)

// if( FFABS( p0 - q0 ) < alpha && ..)
        Q8MADL_AA (xr0, xr3, xr4, xr3);
        Q8ABD(xr6, xr12, xr10 );  // FFABS (p2 - p0)
        Q8ABD(xr7, xr2,  xr9 );   // FFABS (q2 - q0)
        Q8MADL_AA (xr0, xr3, xr5, xr3);    // xr3: (FFABS( p0 - q0 ) < alpha
                                           //      && FFABS( p1 - p0 ) < beta && ...)
        Q8ADDE_SS (xr4, xr6, xr14,  xr5);  // FFABS(p2 - p0) - beta
        Q8ADDE_SS (xr6, xr7, xr14,  xr7);  // FFABS(q2 - q0) - beta
        Q8MADL_AA (xr0, xr3, xr15, xr15);  // xr15: new clip value
        Q16SLR (xr4, xr4, xr5, xr5, 15);   // 1: < 0  FFABS(p2 - p0) - beta
        Q16SLR (xr6, xr6, xr7, xr7, 15);   // 1: < 0  FFABS(q2 - q0) - beta
        Q16SAT (xr4, xr4, xr5);   // xr4: 1: < 0 (FFABS( p2 - p0 ) < beta)
        Q16SAT (xr5, xr6, xr7);   // xr5: 1: < 0 (FFABS( q2 - q0 ) < beta)
// calculate clip value
        Q8ADD_AA  (xr6, xr4, xr5);         // pre-calculate for tc++, tc++
        Q8MADL_AA (xr0, xr6, xr3, xr6);    // xr6: new clip differ
        Q8MADL_AA (xr0, xr4, xr15, xr4);   // xr4: new clip value for p1
        Q8MADL_AA (xr0, xr5, xr15, xr5);   // xr5: new clip value for q1
        Q8ADD_AA  (xr3, xr6, xr15);        // xr3: new clip value for p0,q0
//
        Q8AVGR (xr6,  xr9, xr10);  // (p0 + q0 + 1) >> 1
        Q8AVG  (xr12, xr6, xr12);  // (p2 + (p0 + q0 + 1) >> 1) >> 1
        Q8AVG  (xr2,  xr6, xr2);   // (q2 + (p0 + q0 + 1) >> 1) >> 1
        Q8ADDE_SS (xr12, xr12, xr11, xr7); // (p2 + (p0 + q0 + 1) >> 1) >> 1 - p1
        Q8ADDE_SS (xr2,  xr2,  xr1,  xr8); // (q2 + (p0 + q0 + 1) >> 1) >> 1 - q1
        Q8ADDE_SS (xr6,  xr0, xr4, xr15);      //-tc
//av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
        D16MAX (xr12, xr12, xr6);
        D16MAX (xr7,  xr7,  xr15);
        Q8ADDE_AA (xr6,  xr0, xr4, xr15);      //tc
        D16MIN (xr12, xr12, xr6);
        D16MIN (xr7,  xr7,  xr15);
        Q8ADDE_SS (xr6,  xr0, xr5, xr15);      //-tc
        S32SFL (xr0, xr12, xr7, xr12, ptn1);
//av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
        D16MAX (xr2, xr2, xr6);
        D16MAX (xr8, xr8, xr15);
        Q8ADDE_AA (xr6,  xr0, xr5, xr15);      //tc
        D16MIN (xr2, xr2, xr6);
        D16MIN (xr8, xr8, xr15);
// p1 + ...; q1 + ...
        Q8ADD_AA  (xr4, xr12, xr11); // p1 + ...     //final p1
        S32SFL (xr0, xr2, xr8, xr2, ptn1);
        Q8ADD_AA  (xr5, xr2, xr1);    // q1 + ...    //final q1
//  (q0 - p0 ) << 2 + (p1 - q1)
        Q8ADDE_SS (xr2, xr9, xr10, xr12);   // q0 - p0
        Q16SLL (xr2, xr2, xr12, xr12, 2);   // (q0 - p0) << 2
        Q8ACCE_SS (xr2, xr11, xr1, xr12);    // (q0 - p0) << 2 + (p1 - q1)
//  -tc, tc
        Q8ADDE_AA (xr11, xr0, xr3, xr15);     //+tc (xr11, xr15)
        Q8ADDE_SS (xr1,  xr0, xr3, xr6);      //-tc (xr1,  xr6)
//  i_delta = av_clip (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3 ....)
        Q16SAR (xr2, xr2, xr12, xr12, 2);   // ((q0 - p0) << 2 + (p1 - q1) + 0) >> 2
        D16AVGR(xr2,  xr2,  xr0);
        D16AVGR(xr12, xr12, xr0);           // ((q0 - p0) << 2 + (p1 - q1) + 1) >> 1
        D16MAX (xr2,  xr2,  xr1);
        D16MAX (xr12, xr12, xr6);
        D16MIN (xr2,  xr2,  xr11);
        D16MIN (xr12, xr12, xr15);
        Q16ADD_SS_WW (xr1,  xr0, xr2,  xr0);    // xr1  = -xr2
        Q16ADD_SS_WW (xr11, xr0, xr12, xr0);    // xr11 = -xr12
//
        Q8ACCE_AA (xr2, xr0, xr10, xr12);    // (p0 + i_delta)
        Q8ACCE_AA (xr1, xr0, xr9,  xr11);    // (q0 - i_delta)
        Q16SAT    (xr10, xr2, xr12);         // final p0
        Q16SAT    (xr9, xr1, xr11);          // final q0
// store
        S32SFL    (xr2, xr10, xr4, xr1, ptn0);   // xr2: p0^p1^p0"p1", xr1:p0'p1'p0p1
        S32SFL    (xr4, xr5,  xr9, xr3, ptn0);   // xr4: q1^q0^q1"q0", xr3:q1'q0'q1q0
        t1 = S32M2I(xr1);
        t2 = S32M2I(xr2);
        t3 = S32M2I(xr3);
        t4 = S32M2I(xr4);
        *((unsigned short *)pix - 1) = t1;
        *((unsigned short *)pix + 0) = t3;
        pix += stride;
        *((unsigned short *)pix - 1) = (t1 >> 16);
        *((unsigned short *)pix + 0) = (t3 >> 16);
        pix += stride;
        *((unsigned short *)pix - 1) = t2;
        *((unsigned short *)pix + 0) = t4;
        pix += stride;
        *((unsigned short *)pix - 1) = (t2 >> 16);
        *((unsigned short *)pix + 0) = (t4 >> 16);
        pix += stride;
    }
}
#else
static inline void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
{
    int i, d;
    for( i = 0; i < 4; i++ ) {
        if( tc0[i] < 0 ) {
            pix += 4*ystride;
            continue;
        }
        for( d = 0; d < 4; d++ ) {
            const int p0 = pix[-1*xstride];
            const int p1 = pix[-2*xstride];
            const int p2 = pix[-3*xstride];
            const int q0 = pix[0];
            const int q1 = pix[1*xstride];
            const int q2 = pix[2*xstride];

            if( FFABS( p0 - q0 ) < alpha &&
                FFABS( p1 - p0 ) < beta &&
                FFABS( q1 - q0 ) < beta ) {

                int tc = tc0[i];
                int i_delta;

                if( FFABS( p2 - p0 ) < beta ) {
                    pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
                    tc++;
                }
                if( FFABS( q2 - q0 ) < beta ) {
                    pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
                    tc++;
                }

                i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                pix[-xstride] = av_clip_uint8( p0 + i_delta );    /* p0' */
                pix[0]        = av_clip_uint8( q0 - i_delta );    /* q0' */
            }
            pix += ystride;
        }
    }
}
static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0);
}
static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
}
#endif

#ifdef JZ4740_MXU_OPT
static void h264_v_loop_filter_chroma_mxu(uint8_t *pix, int xstride, int alpha, int beta, int8_t *tc0)
{
    uint8_t *tpix;
    int i, d;

    S32I2M (xr14, beta);
    S32SFL (xr0,  xr14, xr14, xr14, ptn0);
    S32SFL (xr0,  xr14, xr14, xr14, ptn3);   // xr14: beta
    S32I2M (xr13, alpha);
    S32SFL (xr0,  xr13, xr13, xr13, ptn0);
    S32SFL (xr0,  xr13, xr13, xr13, ptn3);   // xr13: alpha
    S32I2M (xr8,  0x01010101);

    for(i = 0; i < 2; i++) {
      const int tc = *((short *)tc0 + i);
//
      tpix = pix - 2*xstride;
      S32LDD (xr11, tpix, 0);            // p1
      S32LDIV(xr10, tpix, xstride, 0);   // p0
      S32LDIV(xr9,  tpix, xstride, 0);   // q0
      S32LDIV(xr1,  tpix, xstride, 0);   // q1
      Q8ABD(xr3, xr10, xr9 );   // FFABS (p0 - q0)
      Q8ABD(xr4, xr11, xr10);   // FFABS (p1 - p0)
      Q8ABD(xr5, xr1,  xr9 );   // FFABS (q1 - q0)
// FFABS(p0 - q0) - alpha, FFABS(p1 - p0) - beta
      Q8ADDE_SS (xr6, xr3, xr13,  xr7);  // FFABS(p0 - q0) - alpha
      Q8ADDE_SS (xr3, xr4, xr14,  xr4);  // FFABS(p1 - p0) - beta
      Q16SLR (xr6, xr6, xr7, xr7, 15);   // 1: < 0 (FFABS( p0 - q0 ) < alpha)
      Q16SLR (xr3, xr3, xr4, xr4, 15);   // 1: < 0 (FFABS( p1 - p0 ) < beta)
      Q16SAT (xr4, xr3, xr4);   // xr4: 1: < 0 (FFABS( p1 - p0 ) < beta
      Q16SAT (xr3, xr6, xr7);   // xr3: 1: < 0 (FFABS( p0 - q0 ) < alpha)
// FFABS(q1 - q0) - beta
      Q8ADDE_SS (xr6, xr5, xr14,  xr7);  // FFABS(q1 - q0) - beta
      Q16SLR (xr6, xr6, xr7, xr7, 15);   // 1: < 0 (FFABS( q1 - q0 ) < beta)
      Q16SAT (xr5, xr6, xr7);   // xr5: 1: < 0 (FFABS( q1 - q0 ) < beta)
// if( FFABS( p0 - q0 ) < alpha && ..)
      Q8MADL_AA (xr0, xr3, xr4, xr3);
// xr15 = tc
      S32I2M (xr2, tc);
      S32SFL (xr0, xr2, xr2, xr15, ptn0);
//
      Q8ADDE_SS (xr2, xr9, xr10, xr12);   // q0 - p0
      Q8ACCE_AA (xr2, xr0, xr8, xr12);    // (q0 - p0 + 1)
      Q8MADL_AA (xr0, xr3, xr5, xr3);    // xr3: (FFABS( p0 - q0 ) < alpha
                                         //      && FFABS( p1 - p0 ) < beta && ...)
      Q8MAX  (xr15, xr0, xr15);
      Q16SLL (xr2, xr2, xr12, xr12, 2);   // (q0 - p0) << 2 + 4
      Q8MADL_AA (xr0, xr3, xr15, xr15);    // xr15: (FFABS( p0 - q0 ) < alpha .... ) * tc
      Q8ACCE_SS (xr2, xr11, xr1, xr12);    // (q0 - p0) << 2 + (p1 - q1) + 4
//  -tc, tc
      Q8ADDE_AA (xr7, xr0, xr15, xr6);     //  tc1,  tc0
      Q8ADDE_SS (xr5, xr0, xr15, xr4);     // -tc1, -tc0
      Q16SAR (xr2, xr2, xr12, xr12, 3);    
      D16MAX (xr2,  xr2,  xr5);
      D16MAX (xr12, xr12, xr4);
      D16MIN (xr2,  xr2,  xr7);
      D16MIN (xr12, xr12, xr6);
// p0 + delta, p1 - delta
      Q16ADD_SS_WW (xr1,  xr0, xr2,  xr0);    // xr1  = -xr2     -delta
      Q16ADD_SS_WW (xr11, xr0, xr12, xr0);    // xr11 = -xr12    -delta
      Q8ACCE_AA (xr2, xr0, xr10, xr12);    // (p0 + i_delta)
      Q8ACCE_AA (xr1, xr0, xr9,  xr11);    // (q0 - i_delta)
      tpix = pix - 2*xstride;
      Q16SAT    (xr2, xr2, xr12);
      Q16SAT    (xr1, xr1, xr11);
      S32SDIV(xr2,  tpix, xstride, 0);      //-1, p0
      S32SDIV(xr1,  tpix, xstride, 0);      //0, q0
      pix += 4;
    }
}

static void h264_h_loop_filter_chroma_mxu(uint8_t *pix, int ystride, int alpha, int beta, int8_t *tc0)
{
    int i, d;
    uint8_t *tpix;
    unsigned int p0, q0;
    
    S32I2M (xr14, beta);
    S32SFL (xr0,  xr14, xr14, xr14, ptn0);
    S32SFL (xr0,  xr14, xr14, xr14, ptn3);   // xr14: beta
    S32I2M (xr13, alpha);
    S32SFL (xr0,  xr13, xr13, xr13, ptn0);
    S32SFL (xr0,  xr13, xr13, xr13, ptn3);   // xr13: alpha
    S32I2M (xr8,  0x01010101);

    for(i = 0; i < 2; i++) {
      const int tc = *((short *)tc0 + i);
//
      tpix = pix;
      S32LDD  (xr2, tpix,  0);          // xr2: q3,q2,q1,q0
      S32LDD  (xr1, tpix, -4);          // xr1: p0,p1,p2,p3
      S32LDIV (xr4, tpix, ystride, 0);  // xr4: q3',q2',q1',q0'
      S32LDD  (xr3, tpix, -4);          // xr3: p0',p1',p2',p3'
//
      S32SFL  (xr6, xr4, xr2, xr12, ptn0); // xr6: q3'q3q2'q2, xr12:q1'q1q0'q0
      S32SFL  (xr5, xr3, xr1, xr7, ptn0); // xr5: p0'p0p1'p1, xr7:p2'p2p3'p
//
      S32LDIV (xr2, tpix, ystride, 0);    // xr2: q3",q2",q1",q0"
      S32LDD  (xr1, tpix, -4);          // xr1: p0",p1",p2",p3"
      S32LDIV (xr4, tpix, ystride, 0);    // xr4: q3^,q2^,q1^,q0^
      S32LDD  (xr3, tpix, -4);          // xr3: p0^,p1^,p2^,p3^
//
      S32SFL  (xr4, xr4, xr2, xr2, ptn0); // xr4: q3^q3"q2^q2", xr2:q1^q1"q0^q0"
      S32SFL  (xr3, xr3, xr1, xr1, ptn0); // xr3: p0^p0"p1^p1", xr1:p2^p2"p3^p3"
//
      S32SFL  (xr10, xr3, xr5, xr11, ptn3); // xr10: p0^p0"p0'p0, xr11:p1^p1"p1'p1
      S32SFL  (xr1,  xr2, xr12, xr9,  ptn3); // xr1: q1^q1"q1'q1, xr9:q0^q0"q0'q0
//
      Q8ABD(xr3, xr10, xr9 );   // FFABS (p0 - q0)
      Q8ABD(xr4, xr11, xr10);   // FFABS (p1 - p0)
      Q8ABD(xr5, xr1,  xr9 );   // FFABS (q1 - q0)
//    FFABS(p0 - q0) - alpha, FFABS(p1 - p0) - beta
      Q8ADDE_SS (xr6, xr3, xr13,  xr7);  // FFABS(p0 - q0) - alpha
      Q8ADDE_SS (xr3, xr4, xr14,  xr4);  // FFABS(p1 - p0) - beta
      Q16SLR (xr6, xr6, xr7, xr7, 15);   // 1: < 0 (FFABS( p0 - q0 ) < alpha)
      Q16SLR (xr3, xr3, xr4, xr4, 15);   // 1: < 0 (FFABS( p1 - p0 ) < beta)
      Q16SAT (xr4, xr3, xr4);   // xr4: 1: < 0 (FFABS( p1 - p0 ) < beta
      Q16SAT (xr3, xr6, xr7);   // xr3: 1: < 0 (FFABS( p0 - q0 ) < alpha)
// FFABS(q1 - q0) - beta
      Q8ADDE_SS (xr6, xr5, xr14,  xr7);  // FFABS(q1 - q0) - beta
      Q16SLR (xr6, xr6, xr7, xr7, 15);   // 1: < 0 (FFABS( q1 - q0 ) < beta)
      Q16SAT (xr5, xr6, xr7);   // xr5: 1: < 0 (FFABS( q1 - q0 ) < beta)
// if( FFABS( p0 - q0 ) < alpha && ..)
      Q8MADL_AA (xr0, xr3, xr4, xr3);
// xr15 = tc
      S32I2M (xr2, tc);
      S32SFL (xr0, xr2, xr2, xr15, ptn0);
//
      Q8ADDE_SS (xr2, xr9, xr10, xr12);   // q0 - p0
      Q8ACCE_AA (xr2, xr0, xr8, xr12);    // (q0 - p0 + 1)
      Q8MADL_AA (xr0, xr3, xr5, xr3);     // xr3: (FFABS( p0 - q0 ) < alpha
                                          //      && FFABS( p1 - p0 ) < beta && ...)      
      Q8MAX  (xr15, xr0, xr15);
      Q16SLL (xr2, xr2, xr12, xr12, 2);   // (q0 - p0) << 2 + 4
      Q8MADL_AA (xr0, xr3, xr15, xr15);    // xr15: (FFABS( p0 - q0 ) < alpha .... ) * tc
      Q8ACCE_SS (xr2, xr11, xr1, xr12);    // (q0 - p0) << 2 + (p1 - q1) + 4
//  -tc, tc
      Q8ADDE_AA (xr7, xr0, xr15, xr6);     //  tc1,  tc0
      Q8ADDE_SS (xr5, xr0, xr15, xr4);     // -tc1, -tc0
      Q16SAR (xr2, xr2, xr12, xr12, 3);
      D16MAX (xr2,  xr2,  xr5);
      D16MAX (xr12, xr12, xr4);
      D16MIN (xr2,  xr2,  xr7);
      D16MIN (xr12, xr12, xr6);
// p0 + delta, p1 - delta