comparison vc1dsp.c @ 9859:7a116de63777 libavcodec

idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall. Includes mmx2 asm for the various functions. Note that the actual idct still does not have an x86 SIMD implemtation. For wmv3 files using regular idct, the decoder just falls back to simple_idct, since simple_idct_dc doesn't exist (yet).

9858:53d5914a30ef

    vc1_loop_filter(src, stride, 1, 16, pq);
}

/** Do inverse transform on 8x8 block
*/
static void vc1_inv_trans_8x8_c(DCTELEM block[64])
{
    int i;
    register int t1,t2,t3,t4,t5,t6,t7,t8;
    DCTELEM *src, *dst;

    }
}

/** Do inverse transform on 8x4 part of block
*/
static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    register int t1,t2,t3,t4,t5,t6,t7,t8;
    DCTELEM *src, *dst;

    }
}

/** Do inverse transform on 4x8 parts of block
*/
static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    register int t1,t2,t3,t4,t5,t6,t7,t8;
    DCTELEM *src, *dst;

    }
}

/** Do inverse transform on 4x4 part of block
*/
static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    register int t1,t2,t3,t4;
    DCTELEM *src, *dst;

void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
    dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
    dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
    dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
    dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
    dsp->vc1_h_overlap = vc1_h_overlap_c;
    dsp->vc1_v_overlap = vc1_v_overlap_c;
    dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;
    dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;
    dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;

9859:7a116de63777

    vc1_loop_filter(src, stride, 1, 16, pq);
}

/** Do inverse transform on 8x8 block
*/
static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    int dc = block[0];
    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
    dc = (3 * dc +  1) >> 1;
    dc = (3 * dc + 16) >> 5;
    for(i = 0; i < 8; i++){
        dest[0] = cm[dest[0]+dc];
        dest[1] = cm[dest[1]+dc];
        dest[2] = cm[dest[2]+dc];
        dest[3] = cm[dest[3]+dc];
        dest[4] = cm[dest[4]+dc];
        dest[5] = cm[dest[5]+dc];
        dest[6] = cm[dest[6]+dc];
        dest[7] = cm[dest[7]+dc];
        dest += linesize;
    }
}

static void vc1_inv_trans_8x8_c(DCTELEM block[64])
{
    int i;
    register int t1,t2,t3,t4,t5,t6,t7,t8;
    DCTELEM *src, *dst;

    }
}

/** Do inverse transform on 8x4 part of block
*/
static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    int dc = block[0];
    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
    dc = ( 3 * dc +  1) >> 1;
    dc = (17 * dc + 64) >> 7;
    for(i = 0; i < 4; i++){
        dest[0] = cm[dest[0]+dc];
        dest[1] = cm[dest[1]+dc];
        dest[2] = cm[dest[2]+dc];
        dest[3] = cm[dest[3]+dc];
        dest[4] = cm[dest[4]+dc];
        dest[5] = cm[dest[5]+dc];
        dest[6] = cm[dest[6]+dc];
        dest[7] = cm[dest[7]+dc];
        dest += linesize;
    }
}

static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    register int t1,t2,t3,t4,t5,t6,t7,t8;
    DCTELEM *src, *dst;

    }
}

/** Do inverse transform on 4x8 parts of block
*/
static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    int dc = block[0];
    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
    dc = (17 * dc +  4) >> 3;
    dc = (12 * dc + 64) >> 7;
    for(i = 0; i < 8; i++){
        dest[0] = cm[dest[0]+dc];
        dest[1] = cm[dest[1]+dc];
        dest[2] = cm[dest[2]+dc];
        dest[3] = cm[dest[3]+dc];
        dest += linesize;
    }
}

static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    register int t1,t2,t3,t4,t5,t6,t7,t8;
    DCTELEM *src, *dst;

    }
}

/** Do inverse transform on 4x4 part of block
*/
static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    int dc = block[0];
    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
    dc = (17 * dc +  4) >> 3;
    dc = (17 * dc + 64) >> 7;
    for(i = 0; i < 4; i++){
        dest[0] = cm[dest[0]+dc];
        dest[1] = cm[dest[1]+dc];
        dest[2] = cm[dest[2]+dc];
        dest[3] = cm[dest[3]+dc];
        dest += linesize;
    }
}

static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block)
{
    int i;
    register int t1,t2,t3,t4;
    DCTELEM *src, *dst;

void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
    dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
    dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
    dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
    dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
    dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c;
    dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c;
    dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c;
    dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c;
    dsp->vc1_h_overlap = vc1_h_overlap_c;
    dsp->vc1_v_overlap = vc1_v_overlap_c;
    dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;
    dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;
    dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;