dsputil.c

    qpel_v_lowpass(halfHV, halfH, 8, 8, 8, 16-r);\
    avg2_block(dst, halfV, halfHV, dstStride, 8, 1-r);\
}\
static void qpel_mc22_c ## name (UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my)\
{\
    UINT8 halfH[72];\
    qpel_h_lowpass(halfH, src, 8, srcStride, 9, 16-r);\
    qpel_v_lowpass(dst, halfH, dstStride, 8, 8, 16-r);\
}\
qpel_mc_func qpel_mc ## name ## _tab[16]={ \
    qpel_mc00_c ## name,                                                                   \
    qpel_mc10_c ## name,                                                                   \
    qpel_mc20_c ## name,                                                                   \
    qpel_mc30_c ## name,                                                                   \
    qpel_mc01_c ## name,                                                                   \
    qpel_mc11_c ## name,                                                                   \
    qpel_mc21_c ## name,                                                                   \
    qpel_mc31_c ## name,                                                                   \
    qpel_mc02_c ## name,                                                                   \
    qpel_mc12_c ## name,                                                                   \
    qpel_mc22_c ## name,                                                                   \
    qpel_mc32_c ## name,                                                                   \
    qpel_mc03_c ## name,                                                                   \
    qpel_mc13_c ## name,                                                                   \
    qpel_mc23_c ## name,                                                                   \
    qpel_mc33_c ## name,                                                                   \
};

QPEL_MC(0, _rnd)
QPEL_MC(1, _no_rnd)

int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);
        s += abs(pix1[5] - pix2[5]);
        s += abs(pix1[6] - pix2[6]);
        s += abs(pix1[7] - pix2[7]);
        s += abs(pix1[8] - pix2[8]);
        s += abs(pix1[9] - pix2[9]);
        s += abs(pix1[10] - pix2[10]);
        s += abs(pix1[11] - pix2[11]);
        s += abs(pix1[12] - pix2[12]);
        s += abs(pix1[13] - pix2[13]);
        s += abs(pix1[14] - pix2[14]);
        s += abs(pix1[15] - pix2[15]);
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

int pix_abs16x16_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
        s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
        s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
        s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
        s += abs(pix1[8] - avg2(pix2[8], pix2[9]));
        s += abs(pix1[9] - avg2(pix2[9], pix2[10]));
        s += abs(pix1[10] - avg2(pix2[10], pix2[11]));
        s += abs(pix1[11] - avg2(pix2[11], pix2[12]));
        s += abs(pix1[12] - avg2(pix2[12], pix2[13]));
        s += abs(pix1[13] - avg2(pix2[13], pix2[14]));
        s += abs(pix1[14] - avg2(pix2[14], pix2[15]));
        s += abs(pix1[15] - avg2(pix2[15], pix2[16]));
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

int pix_abs16x16_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;
    UINT8 *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
        s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
        s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
        s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
        s += abs(pix1[8] - avg2(pix2[8], pix3[8]));
        s += abs(pix1[9] - avg2(pix2[9], pix3[9]));
        s += abs(pix1[10] - avg2(pix2[10], pix3[10]));
        s += abs(pix1[11] - avg2(pix2[11], pix3[11]));
        s += abs(pix1[12] - avg2(pix2[12], pix3[12]));
        s += abs(pix1[13] - avg2(pix2[13], pix3[13]));
        s += abs(pix1[14] - avg2(pix2[14], pix3[14]));
        s += abs(pix1[15] - avg2(pix2[15], pix3[15]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

int pix_abs16x16_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;
    UINT8 *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
        s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
        s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
        s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
        s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9]));
        s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10]));
        s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11]));
        s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12]));
        s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13]));
        s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14]));
        s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15]));
        s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

int pix_abs8x8_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);
        s += abs(pix1[5] - pix2[5]);
        s += abs(pix1[6] - pix2[6]);
        s += abs(pix1[7] - pix2[7]);
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

int pix_abs8x8_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
        s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
        s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
        s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

int pix_abs8x8_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;
    UINT8 *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
        s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
        s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
        s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

int pix_abs8x8_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
{
    int s, i;
    UINT8 *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
        s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
        s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
        s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

/* permute block according so that it corresponds to the MMX idct
   order */
#ifdef SIMPLE_IDCT
 /* general permutation, but perhaps slightly slower */
void block_permute(INT16 *block)
{
	int i;
	INT16 temp[64];

	for(i=0; i<64; i++) temp[ block_permute_op(i) ] = block[i];

	for(i=0; i<64; i++) block[i] = temp[i];
}
#else

void block_permute(INT16 *block)
{
    int tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
    int i;

    for(i=0;i<8;i++) {
        tmp1 = block[1];
        tmp2 = block[2];
        tmp3 = block[3];
        tmp4 = block[4];
        tmp5 = block[5];
        tmp6 = block[6];
        block[1] = tmp2;
        block[2] = tmp4;
        block[3] = tmp6;
        block[4] = tmp1;
        block[5] = tmp3;
        block[6] = tmp5;
        block += 8;
    }
}
#endif

void clear_blocks_c(DCTELEM *blocks)
{
    memset(blocks, 0, sizeof(DCTELEM)*6*64);
}

/* XXX: those functions should be suppressed ASAP when all IDCTs are
   converted */
void gen_idct_put(UINT8 *dest, int line_size, DCTELEM *block)
{
    ff_idct (block);
    put_pixels_clamped(block, dest, line_size);
}

void gen_idct_add(UINT8 *dest, int line_size, DCTELEM *block)
{
    ff_idct (block);
    add_pixels_clamped(block, dest, line_size);
}

void dsputil_init(void)
{
    int i, j;
    int use_permuted_idct;

    for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i;
    for(i=0;i<MAX_NEG_CROP;i++) {
        cropTbl[i] = 0;
        cropTbl[i + MAX_NEG_CROP + 256] = 255;
    }

    for(i=0;i<512;i++) {
        squareTbl[i] = (i - 256) * (i - 256);
    }

#ifdef SIMPLE_IDCT
    ff_idct = NULL;
#else
    ff_idct = j_rev_dct;
#endif
    get_pixels = get_pixels_c;
    diff_pixels = diff_pixels_c;
    put_pixels_clamped = put_pixels_clamped_c;
    add_pixels_clamped = add_pixels_clamped_c;
    gmc1= gmc1_c;
    clear_blocks= clear_blocks_c;
    pix_sum= pix_sum_c;
    pix_norm1= pix_norm1_c;

    pix_abs16x16     = pix_abs16x16_c;
    pix_abs16x16_x2  = pix_abs16x16_x2_c;
    pix_abs16x16_y2  = pix_abs16x16_y2_c;
    pix_abs16x16_xy2 = pix_abs16x16_xy2_c;
    pix_abs8x8     = pix_abs8x8_c;
    pix_abs8x8_x2  = pix_abs8x8_x2_c;
    pix_abs8x8_y2  = pix_abs8x8_y2_c;
    pix_abs8x8_xy2 = pix_abs8x8_xy2_c;
    av_fdct = fdct_ifast;

    use_permuted_idct = 1;

#ifdef HAVE_MMX
    dsputil_init_mmx();
#endif
#ifdef ARCH_ARMV4L
    dsputil_init_armv4l();
#endif
#ifdef HAVE_MLIB
    dsputil_init_mlib();
    use_permuted_idct = 0;
#endif
#ifdef ARCH_ALPHA
    dsputil_init_alpha();
    use_permuted_idct = 0;
#endif
#ifdef ARCH_POWERPC
    dsputil_init_altivec();
#endif

#ifdef SIMPLE_IDCT
    if (ff_idct == NULL) {
        ff_idct_put = simple_idct_put;
        ff_idct_add = simple_idct_add;
        use_permuted_idct=0;
    }
#endif
    if(ff_idct != NULL) {
        ff_idct_put = gen_idct_put;
        ff_idct_add = gen_idct_add;
    }

    if(use_permuted_idct)
#ifdef SIMPLE_IDCT
        for(i=0; i<64; i++) permutation[i]= simple_mmx_permutation[i];
#else
        for(i=0; i<64; i++) permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);
#endif
    else
        for(i=0; i<64; i++) permutation[i]=i;

    for(i=0; i<64; i++) inv_zigzag_direct16[zigzag_direct[i]]= i+1;
    for(i=0; i<64; i++) zigzag_direct_noperm[i]= zigzag_direct[i];
    
    if (use_permuted_idct) {
        /* permute for IDCT */
        for(i=0;i<64;i++) {
            j = zigzag_direct[i];
            zigzag_direct[i] = block_permute_op(j);
            j = ff_alternate_horizontal_scan[i];
            ff_alternate_horizontal_scan[i] = block_permute_op(j);
            j = ff_alternate_vertical_scan[i];
            ff_alternate_vertical_scan[i] = block_permute_op(j);
        }
        block_permute(ff_mpeg1_default_intra_matrix);
        block_permute(ff_mpeg1_default_non_intra_matrix);
        block_permute(ff_mpeg4_default_intra_matrix);
        block_permute(ff_mpeg4_default_non_intra_matrix);
    }
    
    build_zigzag_end();
}

/* remove any non bit exact operation (testing purpose) */
void avcodec_set_bit_exact(void)
{
#ifdef HAVE_MMX
    dsputil_set_bit_exact_mmx();
#endif
}

void get_psnr(UINT8 *orig_image[3], UINT8 *coded_image[3],
              int orig_linesize[3], int coded_linesize,
              AVCodecContext *avctx)
{
    int quad, diff, x, y;
    UINT8 *orig, *coded;
    UINT32 *sq = squareTbl + 256;
    
    quad = 0;
    diff = 0;
    
    /* Luminance */
    orig = orig_image[0];
    coded = coded_image[0];
    
    for (y=0;y<avctx->height;y++) {
        for (x=0;x<avctx->width;x++) {
            diff = *(orig + x) - *(coded + x);
            quad += sq[diff];
        }
        orig += orig_linesize[0];
        coded += coded_linesize;
    }
   
    avctx->psnr_y = (float) quad / (float) (avctx->width * avctx->height);
    
    if (avctx->psnr_y) {
        avctx->psnr_y = (float) (255 * 255) / avctx->psnr_y;
        avctx->psnr_y = 10 * (float) log10 (avctx->psnr_y); 
    } else
        avctx->psnr_y = 99.99;
}