comparison h264.c @ 5528:4c3b1222ff57 libavcodec

Don't check the return value of decode_cabac_residual since it always returns 0. This leads to a 0.4% speed-up. Patch by Alexander Strange astrange at_ ithinksw dot com Original thread: Date: Aug 11, 2007 11:44 PM Subject: [FFmpeg-devel] [PATCH] h264: don't check decode_cabac_residual return

5527:8b6ffd245ee2

    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
    3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
    5, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8
};

static int decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff) {
    const int mb_xy  = h->s.mb_x + h->s.mb_y*h->s.mb_stride;
    static const int significant_coeff_flag_offset[2][6] = {
      { 105+0, 105+15, 105+29, 105+44, 105+47, 402 },
      { 277+0, 277+15, 277+29, 277+44, 277+47, 436 }
    };

#ifdef CABAC_ON_STACK
            h->cabac.range     = cc.range     ;
            h->cabac.low       = cc.low       ;
            h->cabac.bytestream= cc.bytestream;
#endif
            return 0;
        }
    }

    significant_coeff_ctx_base = h->cabac_state
        + significant_coeff_flag_offset[MB_FIELD][cat];

#ifdef CABAC_ON_STACK
            h->cabac.range     = cc.range     ;
            h->cabac.low       = cc.low       ;
            h->cabac.bytestream= cc.bytestream;
#endif
    return 0;
}

static inline void compute_mb_neighbors(H264Context *h)
{
    MpegEncContext * const s = &h->s;

        h->chroma_qp[1] = get_chroma_qp(h, 1, s->qscale);

        if( IS_INTRA16x16( mb_type ) ) {
            int i;
            //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 DC\n" );
            if( decode_cabac_residual( h, h->mb, 0, 0, dc_scan, NULL, 16) < 0)
                return -1;
            if( cbp&15 ) {
                for( i = 0; i < 16; i++ ) {
                    //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 AC:%d\n", i );
                    if( decode_cabac_residual(h, h->mb + 16*i, 1, i, scan + 1, h->dequant4_coeff[0][s->qscale], 15) < 0 )
                        return -1;
                }
            } else {
                fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1);
            }
        } else {
            int i8x8, i4x4;
            for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
                if( cbp & (1<<i8x8) ) {
                    if( IS_8x8DCT(mb_type) ) {
                        if( decode_cabac_residual(h, h->mb + 64*i8x8, 5, 4*i8x8,
                            scan8x8, h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 64) < 0 )
                            return -1;
                    } else
                    for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
                        const int index = 4*i8x8 + i4x4;
                        //av_log( s->avctx, AV_LOG_ERROR, "Luma4x4: %d\n", index );
//START_TIMER
                        if( decode_cabac_residual(h, h->mb + 16*index, 2, index, scan, h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale], 16) < 0 )
                            return -1;
//STOP_TIMER("decode_residual")
                    }
                } else {
                    uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
                    nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;

        if( cbp&0x30 ){
            int c;
            for( c = 0; c < 2; c++ ) {
                //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c );
                if( decode_cabac_residual(h, h->mb + 256 + 16*4*c, 3, c, chroma_dc_scan, NULL, 4) < 0)
                    return -1;
            }
        }

        if( cbp&0x20 ) {
            int c, i;
            for( c = 0; c < 2; c++ ) {
                const uint32_t *qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[c]];
                for( i = 0; i < 4; i++ ) {
                    const int index = 16 + 4 * c + i;
                    //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 );
                    if( decode_cabac_residual(h, h->mb + 16*index, 4, index - 16, scan + 1, qmul, 15) < 0)
                        return -1;
                }
            }
        } else {
            uint8_t * const nnz= &h->non_zero_count_cache[0];
            nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =

5528:4c3b1222ff57

    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
    3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
    5, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8
};

static void decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff) {
    const int mb_xy  = h->s.mb_x + h->s.mb_y*h->s.mb_stride;
    static const int significant_coeff_flag_offset[2][6] = {
      { 105+0, 105+15, 105+29, 105+44, 105+47, 402 },
      { 277+0, 277+15, 277+29, 277+44, 277+47, 436 }
    };

#ifdef CABAC_ON_STACK
            h->cabac.range     = cc.range     ;
            h->cabac.low       = cc.low       ;
            h->cabac.bytestream= cc.bytestream;
#endif
            return;
        }
    }

    significant_coeff_ctx_base = h->cabac_state
        + significant_coeff_flag_offset[MB_FIELD][cat];

#ifdef CABAC_ON_STACK
            h->cabac.range     = cc.range     ;
            h->cabac.low       = cc.low       ;
            h->cabac.bytestream= cc.bytestream;
#endif

}

static inline void compute_mb_neighbors(H264Context *h)
{
    MpegEncContext * const s = &h->s;

        h->chroma_qp[1] = get_chroma_qp(h, 1, s->qscale);

        if( IS_INTRA16x16( mb_type ) ) {
            int i;
            //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 DC\n" );
            decode_cabac_residual( h, h->mb, 0, 0, dc_scan, NULL, 16);

            if( cbp&15 ) {
                for( i = 0; i < 16; i++ ) {
                    //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 AC:%d\n", i );
                    decode_cabac_residual(h, h->mb + 16*i, 1, i, scan + 1, h->dequant4_coeff[0][s->qscale], 15);
                }
            } else {
                fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1);
            }
        } else {
            int i8x8, i4x4;
            for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
                if( cbp & (1<<i8x8) ) {
                    if( IS_8x8DCT(mb_type) ) {
                        decode_cabac_residual(h, h->mb + 64*i8x8, 5, 4*i8x8,
                            scan8x8, h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 64);
                    } else
                    for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
                        const int index = 4*i8x8 + i4x4;
                        //av_log( s->avctx, AV_LOG_ERROR, "Luma4x4: %d\n", index );
//START_TIMER
                        decode_cabac_residual(h, h->mb + 16*index, 2, index, scan, h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale], 16);
//STOP_TIMER("decode_residual")
                    }
                } else {
                    uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
                    nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;

        if( cbp&0x30 ){
            int c;
            for( c = 0; c < 2; c++ ) {
                //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c );
                decode_cabac_residual(h, h->mb + 256 + 16*4*c, 3, c, chroma_dc_scan, NULL, 4);
            }
        }

        if( cbp&0x20 ) {
            int c, i;
            for( c = 0; c < 2; c++ ) {
                const uint32_t *qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[c]];
                for( i = 0; i < 4; i++ ) {
                    const int index = 16 + 4 * c + i;
                    //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 );
                    decode_cabac_residual(h, h->mb + 16*index, 4, index - 16, scan + 1, qmul, 15);
                }
            }
        } else {
            uint8_t * const nnz= &h->non_zero_count_cache[0];
            nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =