dsputil.c

Trolltech公司发布的图形界面操作系统。可在qt-embedded-2.3.7平台上编译为嵌入式图形界面操作系统。

        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;
}

static 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;
}

static 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;
}

static 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;
}

static 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;
}

static 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;
}

static 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;
}

void ff_block_permute(INT16 *block, UINT8 *permutation, const UINT8 *scantable, int last)
{
    int i;
    INT16 temp[64];
    
    if(last<=0) return;

    for(i=0; i<=last; i++){
        const int j= scantable[i];
        temp[j]= block[j];
        block[j]=0;
    }
    
    for(i=0; i<=last; i++){
        const int j= scantable[i];
        const int perm_j= permutation[j];
        block[perm_j]= temp[j];
    }
}

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

static void add_bytes_c(uint8_t *dst, uint8_t *src, int w){
    int i;
    for(i=0; i+7<w; i++){
        dst[i+0] += src[i+0];
        dst[i+1] += src[i+1];
        dst[i+2] += src[i+2];
        dst[i+3] += src[i+3];
        dst[i+4] += src[i+4];
        dst[i+5] += src[i+5];
        dst[i+6] += src[i+6];
        dst[i+7] += src[i+7];
    }
    for(; i<w; i++)
        dst[i+0] += src[i+0];
}

static void diff_bytes_c(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
    int i;
    for(i=0; i+7<w; i++){
        dst[i+0] = src1[i+0]-src2[i+0];
        dst[i+1] = src1[i+1]-src2[i+1];
        dst[i+2] = src1[i+2]-src2[i+2];
        dst[i+3] = src1[i+3]-src2[i+3];
        dst[i+4] = src1[i+4]-src2[i+4];
        dst[i+5] = src1[i+5]-src2[i+5];
        dst[i+6] = src1[i+6]-src2[i+6];
        dst[i+7] = src1[i+7]-src2[i+7];
    }
    for(; i<w; i++)
        dst[i+0] = src1[i+0]-src2[i+0];
}

void dsputil_init(DSPContext* c, unsigned mask)
{
    static int init_done = 0;
    int i;

    if (!init_done) {
	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);
	}

	for(i=0; i<64; i++) inv_zigzag_direct16[ff_zigzag_direct[i]]= i+1;

	init_done = 1;
    }

    c->get_pixels = get_pixels_c;
    c->diff_pixels = diff_pixels_c;
    c->put_pixels_clamped = put_pixels_clamped_c;
    c->add_pixels_clamped = add_pixels_clamped_c;
    c->gmc1 = gmc1_c;
    c->gmc = gmc_c;
    c->clear_blocks = clear_blocks_c;
    c->pix_sum = pix_sum_c;
    c->pix_norm1 = pix_norm1_c;

    /* TODO [0] 16  [1] 8 */
    c->pix_abs16x16     = pix_abs16x16_c;
    c->pix_abs16x16_x2  = pix_abs16x16_x2_c;
    c->pix_abs16x16_y2  = pix_abs16x16_y2_c;
    c->pix_abs16x16_xy2 = pix_abs16x16_xy2_c;
    c->pix_abs8x8     = pix_abs8x8_c;
    c->pix_abs8x8_x2  = pix_abs8x8_x2_c;
    c->pix_abs8x8_y2  = pix_abs8x8_y2_c;
    c->pix_abs8x8_xy2 = pix_abs8x8_xy2_c;

#define dspfunc(PFX, IDX, NUM) \
    c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## NUM ## _c;     \
    c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## NUM ## _x2_c;  \
    c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## NUM ## _y2_c;  \
    c->PFX ## _pixels_tab[IDX][3] = PFX ## _pixels ## NUM ## _xy2_c

    dspfunc(put, 0, 16);
    dspfunc(put_no_rnd, 0, 16);
    dspfunc(put, 1, 8);
    dspfunc(put_no_rnd, 1, 8);

    dspfunc(avg, 0, 16);
    dspfunc(avg_no_rnd, 0, 16);
    dspfunc(avg, 1, 8);
    dspfunc(avg_no_rnd, 1, 8);
#undef dspfunc

#define dspfunc(PFX, IDX, NUM) \
    c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \
    c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \
    c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_c; \
    c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_c; \
    c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_c; \
    c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_c; \
    c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_c; \
    c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_c; \
    c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_c; \
    c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_c; \
    c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_c; \
    c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_c; \
    c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_c; \
    c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_c; \
    c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_c; \
    c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_c

    dspfunc(put_qpel, 0, 16);
    dspfunc(put_no_rnd_qpel, 0, 16);

    dspfunc(avg_qpel, 0, 16);
    /* dspfunc(avg_no_rnd_qpel, 0, 16); */

    dspfunc(put_qpel, 1, 8);
    dspfunc(put_no_rnd_qpel, 1, 8);

    dspfunc(avg_qpel, 1, 8);
    /* dspfunc(avg_no_rnd_qpel, 1, 8); */
#undef dspfunc

    c->add_bytes= add_bytes_c;
    c->diff_bytes= diff_bytes_c;

#ifdef HAVE_MMX
    dsputil_init_mmx(c, mask);
    if (ff_bit_exact)
    {
        /* FIXME - AVCodec context should have flag for bitexact match */
	/* fprintf(stderr, "\n\n\nff_bit_exact %d\n\n\n\n", ff_bit_exact); */
	dsputil_set_bit_exact_mmx(c, mask);
    }
#endif
#ifdef ARCH_ARMV4L
    dsputil_init_armv4l(c, mask);
#endif
#ifdef HAVE_MLIB
    dsputil_init_mlib(c, mask);
#endif
#ifdef ARCH_ALPHA
    dsputil_init_alpha(c, mask);
#endif
#ifdef ARCH_POWERPC
    dsputil_init_ppc(c, mask);
#endif
#ifdef HAVE_MMI
    dsputil_init_mmi(c, mask);
#endif

}

/* remove any non bit exact operation (testing purpose) */
void avcodec_set_bit_exact(void)
{
    ff_bit_exact=1;
#ifdef HAVE_MMX
// FIXME - better set_bit_exact
//    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;
}