yuv2rgb_altivec.c.svn-base

mediastreamer2是开源的网络传输媒体流的库

            U = (vector signed short)
                vec_perm (uyvy, (vector unsigned char)AVV(0), demux_u);

            V = (vector signed short)
                vec_perm (uyvy, (vector unsigned char)AVV(0), demux_v);

            Y = (vector signed short)
                vec_perm (uyvy, (vector unsigned char)AVV(0), demux_y);

            cvtyuvtoRGB (c, Y,U,V,&R0,&G0,&B0);

            uyvy = vec_ld (16, img);
            U = (vector signed short)
                vec_perm (uyvy, (vector unsigned char)AVV(0), demux_u);

            V = (vector signed short)
                vec_perm (uyvy, (vector unsigned char)AVV(0), demux_v);

            Y = (vector signed short)
                vec_perm (uyvy, (vector unsigned char)AVV(0), demux_y);

            cvtyuvtoRGB (c, Y,U,V,&R1,&G1,&B1);

            R  = vec_packclp (R0,R1);
            G  = vec_packclp (G0,G1);
            B  = vec_packclp (B0,B1);

            //      vec_mstbgr24 (R,G,B, out);
            out_rgba (R,G,B,out);

            img += 32;
        }
    }
    return srcSliceH;
}



/* Ok currently the acceleration routine only supports
   inputs of widths a multiple of 16
   and heights a multiple 2

   So we just fall back to the C codes for this.
*/
SwsFunc yuv2rgb_init_altivec (SwsContext *c)
{
    if (!(c->flags & SWS_CPU_CAPS_ALTIVEC))
        return NULL;

    /*
      and this seems not to matter too much I tried a bunch of
      videos with abnormal widths and mplayer crashes else where.
      mplayer -vo x11 -rawvideo on:w=350:h=240 raw-350x240.eyuv
      boom with X11 bad match.

    */
    if ((c->srcW & 0xf) != 0)    return NULL;

    switch (c->srcFormat) {
    case PIX_FMT_YUV410P:
    case PIX_FMT_YUV420P:
    /*case IMGFMT_CLPL:        ??? */
    case PIX_FMT_GRAY8:
    case PIX_FMT_NV12:
    case PIX_FMT_NV21:
        if ((c->srcH & 0x1) != 0)
            return NULL;

        switch(c->dstFormat){
        case PIX_FMT_RGB24:
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space RGB24\n");
            return altivec_yuv2_rgb24;
        case PIX_FMT_BGR24:
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space BGR24\n");
            return altivec_yuv2_bgr24;
        case PIX_FMT_ARGB:
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space ARGB\n");
            return altivec_yuv2_argb;
        case PIX_FMT_ABGR:
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space ABGR\n");
            return altivec_yuv2_abgr;
        case PIX_FMT_RGBA:
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space RGBA\n");
            return altivec_yuv2_rgba;
        case PIX_FMT_BGRA:
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space BGRA\n");
            return altivec_yuv2_bgra;
        default: return NULL;
        }
        break;

    case PIX_FMT_UYVY422:
        switch(c->dstFormat){
        case PIX_FMT_BGR32:
            av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space UYVY -> RGB32\n");
            return altivec_uyvy_rgb32;
        default: return NULL;
        }
        break;

    }
    return NULL;
}

static uint16_t roundToInt16(int64_t f){
    int r= (f + (1<<15))>>16;
         if (r<-0x7FFF) return 0x8000;
    else if (r> 0x7FFF) return 0x7FFF;
    else                return r;
}

void yuv2rgb_altivec_init_tables (SwsContext *c, const int inv_table[4],int brightness,int contrast, int saturation)
{
    union {
        signed short tmp[8] __attribute__ ((aligned(16)));
        vector signed short vec;
    } buf;

    buf.tmp[0] =  ((0xffffLL) * contrast>>8)>>9;                        //cy
    buf.tmp[1] =  -256*brightness;                                      //oy
    buf.tmp[2] =  (inv_table[0]>>3) *(contrast>>16)*(saturation>>16);   //crv
    buf.tmp[3] =  (inv_table[1]>>3) *(contrast>>16)*(saturation>>16);   //cbu
    buf.tmp[4] = -((inv_table[2]>>1)*(contrast>>16)*(saturation>>16));  //cgu
    buf.tmp[5] = -((inv_table[3]>>1)*(contrast>>16)*(saturation>>16));  //cgv


    c->CSHIFT = (vector unsigned short)vec_splat_u16(2);
    c->CY   = vec_splat ((vector signed short)buf.vec, 0);
    c->OY   = vec_splat ((vector signed short)buf.vec, 1);
    c->CRV  = vec_splat ((vector signed short)buf.vec, 2);
    c->CBU  = vec_splat ((vector signed short)buf.vec, 3);
    c->CGU  = vec_splat ((vector signed short)buf.vec, 4);
    c->CGV  = vec_splat ((vector signed short)buf.vec, 5);
#if 0
    {
    int i;
    char *v[6]={"cy","oy","crv","cbu","cgu","cgv"};
    for (i=0; i<6; i++)
        printf("%s %d ", v[i],buf.tmp[i] );
        printf("\n");
    }
#endif
    return;
}


void
altivec_yuv2packedX (SwsContext *c,
                     int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
                     int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
                     uint8_t *dest, int dstW, int dstY)
{
    int i,j;
    vector signed short X,X0,X1,Y0,U0,V0,Y1,U1,V1,U,V;
    vector signed short R0,G0,B0,R1,G1,B1;

    vector unsigned char R,G,B;
    vector unsigned char *out,*nout;

    vector signed short   RND = vec_splat_s16(1<<3);
    vector unsigned short SCL = vec_splat_u16(4);
    unsigned long scratch[16] __attribute__ ((aligned (16)));

    vector signed short *YCoeffs, *CCoeffs;

    YCoeffs = c->vYCoeffsBank+dstY*lumFilterSize;
    CCoeffs = c->vCCoeffsBank+dstY*chrFilterSize;

    out = (vector unsigned char *)dest;

    for (i=0; i<dstW; i+=16){
        Y0 = RND;
        Y1 = RND;
        /* extract 16 coeffs from lumSrc */
        for (j=0; j<lumFilterSize; j++) {
            X0 = vec_ld (0,  &lumSrc[j][i]);
            X1 = vec_ld (16, &lumSrc[j][i]);
            Y0 = vec_mradds (X0, YCoeffs[j], Y0);
            Y1 = vec_mradds (X1, YCoeffs[j], Y1);
        }

        U = RND;
        V = RND;
        /* extract 8 coeffs from U,V */
        for (j=0; j<chrFilterSize; j++) {
            X  = vec_ld (0, &chrSrc[j][i/2]);
            U  = vec_mradds (X, CCoeffs[j], U);
            X  = vec_ld (0, &chrSrc[j][i/2+2048]);
            V  = vec_mradds (X, CCoeffs[j], V);
        }

        /* scale and clip signals */
        Y0 = vec_sra (Y0, SCL);
        Y1 = vec_sra (Y1, SCL);
        U  = vec_sra (U,  SCL);
        V  = vec_sra (V,  SCL);

        Y0 = vec_clip_s16 (Y0);
        Y1 = vec_clip_s16 (Y1);
        U  = vec_clip_s16 (U);
        V  = vec_clip_s16 (V);

        /* now we have
          Y0= y0 y1 y2 y3 y4 y5 y6 y7     Y1= y8 y9 y10 y11 y12 y13 y14 y15
          U= u0 u1 u2 u3 u4 u5 u6 u7      V= v0 v1 v2 v3 v4 v5 v6 v7

          Y0= y0 y1 y2 y3 y4 y5 y6 y7    Y1= y8 y9 y10 y11 y12 y13 y14 y15
          U0= u0 u0 u1 u1 u2 u2 u3 u3    U1= u4 u4 u5 u5 u6 u6 u7 u7
          V0= v0 v0 v1 v1 v2 v2 v3 v3    V1= v4 v4 v5 v5 v6 v6 v7 v7
        */

        U0 = vec_mergeh (U,U);
        V0 = vec_mergeh (V,V);

        U1 = vec_mergel (U,U);
        V1 = vec_mergel (V,V);

        cvtyuvtoRGB (c, Y0,U0,V0,&R0,&G0,&B0);
        cvtyuvtoRGB (c, Y1,U1,V1,&R1,&G1,&B1);

        R  = vec_packclp (R0,R1);
        G  = vec_packclp (G0,G1);
        B  = vec_packclp (B0,B1);

        switch(c->dstFormat) {
            case PIX_FMT_ABGR:  out_abgr  (R,G,B,out); break;
            case PIX_FMT_BGRA:  out_bgra  (R,G,B,out); break;
            case PIX_FMT_RGBA:  out_rgba  (R,G,B,out); break;
            case PIX_FMT_ARGB:  out_argb  (R,G,B,out); break;
            case PIX_FMT_RGB24: out_rgb24 (R,G,B,out); break;
            case PIX_FMT_BGR24: out_bgr24 (R,G,B,out); break;
            default:
            {
                /* If this is reached, the caller should have called yuv2packedXinC
                   instead. */
                static int printed_error_message;
                if (!printed_error_message) {
                    av_log(c, AV_LOG_ERROR, "altivec_yuv2packedX doesn't support %s output\n",
                           sws_format_name(c->dstFormat));
                    printed_error_message=1;
                }
                return;
            }
        }
    }

    if (i < dstW) {
        i -= 16;

        Y0 = RND;
        Y1 = RND;
        /* extract 16 coeffs from lumSrc */
        for (j=0; j<lumFilterSize; j++) {
            X0 = vec_ld (0,  &lumSrc[j][i]);
            X1 = vec_ld (16, &lumSrc[j][i]);
            Y0 = vec_mradds (X0, YCoeffs[j], Y0);
            Y1 = vec_mradds (X1, YCoeffs[j], Y1);
        }

        U = RND;
        V = RND;
        /* extract 8 coeffs from U,V */
        for (j=0; j<chrFilterSize; j++) {
            X  = vec_ld (0, &chrSrc[j][i/2]);
            U  = vec_mradds (X, CCoeffs[j], U);
            X  = vec_ld (0, &chrSrc[j][i/2+2048]);
            V  = vec_mradds (X, CCoeffs[j], V);
        }

        /* scale and clip signals */
        Y0 = vec_sra (Y0, SCL);
        Y1 = vec_sra (Y1, SCL);
        U  = vec_sra (U,  SCL);
        V  = vec_sra (V,  SCL);

        Y0 = vec_clip_s16 (Y0);
        Y1 = vec_clip_s16 (Y1);
        U  = vec_clip_s16 (U);
        V  = vec_clip_s16 (V);

        /* now we have
           Y0= y0 y1 y2 y3 y4 y5 y6 y7     Y1= y8 y9 y10 y11 y12 y13 y14 y15
           U = u0 u1 u2 u3 u4 u5 u6 u7     V = v0 v1 v2 v3 v4 v5 v6 v7

           Y0= y0 y1 y2 y3 y4 y5 y6 y7    Y1= y8 y9 y10 y11 y12 y13 y14 y15
           U0= u0 u0 u1 u1 u2 u2 u3 u3    U1= u4 u4 u5 u5 u6 u6 u7 u7
           V0= v0 v0 v1 v1 v2 v2 v3 v3    V1= v4 v4 v5 v5 v6 v6 v7 v7
        */

        U0 = vec_mergeh (U,U);
        V0 = vec_mergeh (V,V);

        U1 = vec_mergel (U,U);
        V1 = vec_mergel (V,V);

        cvtyuvtoRGB (c, Y0,U0,V0,&R0,&G0,&B0);
        cvtyuvtoRGB (c, Y1,U1,V1,&R1,&G1,&B1);

        R  = vec_packclp (R0,R1);
        G  = vec_packclp (G0,G1);
        B  = vec_packclp (B0,B1);

        nout = (vector unsigned char *)scratch;
        switch(c->dstFormat) {
            case PIX_FMT_ABGR:  out_abgr  (R,G,B,nout); break;
            case PIX_FMT_BGRA:  out_bgra  (R,G,B,nout); break;
            case PIX_FMT_RGBA:  out_rgba  (R,G,B,nout); break;
            case PIX_FMT_ARGB:  out_argb  (R,G,B,nout); break;
            case PIX_FMT_RGB24: out_rgb24 (R,G,B,nout); break;
            case PIX_FMT_BGR24: out_bgr24 (R,G,B,nout); break;
            default:
                /* Unreachable, I think. */
                av_log(c, AV_LOG_ERROR, "altivec_yuv2packedX doesn't support %s output\n",
                       sws_format_name(c->dstFormat));
                return;
        }

        memcpy (&((uint32_t*)dest)[i], scratch, (dstW-i)/4);
    }

}