umc_h264_mc.cpp

audio-video-codecs.rar语音编解码器

        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset
            + curr_slice->m_InitialOffset[pFields1[block_ref]];
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, pitchPixels, iXType, iYType);

        // interpolate copy the 8x16 block
#ifdef NEW_INTERPOLATE
        Interpolate(mb_decoding_flag, pMCRef, pitchPixels,pDst, 16, iXType, iYType, size8x16, pMVBwd->mvy, uMBy, 0);
#else
        ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pDst, sizeof(PixType)*16, iXType, iYType, size8x16, m_PicParamSet.bit_depth_luma);
#endif

        // Copy the stored 2nd 8x16 prediction data from the BiPred Mode Buffer
        pBiPred = curr_slice->m_pPred4BiPred + 8;
        pMCDst = pDst + 8;
        for (uRow=0; uRow<16; uRow++) {
            memcpy(pMCDst, pBiPred, 8*sizeof(PixType));
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }
        break;

    case MBTYPE_INTER_8x8_REF0:
    case MBTYPE_INTER_8x8:
    case MBTYPE_B_8x8:
        {
        Ipp32s ref_off = 0;
        Ipp32s sb_pos = 0;

        // 4 8x8 blocks
        for (uBlock=0; uBlock < 4; uBlock++)
        {
            block_ref = pRefIdxL0[ref_off];
            prev_frame = block_ref >= 0 ? pRefPicList0[block_ref]->m_pYPlane + uOffset
                + curr_slice->m_InitialOffset[pFields0[block_ref]]: 0;
            block_ref = pRefIdxL1[ref_off];
            futr_frame = block_ref >= 0 ? pRefPicList1[block_ref] ?
                pRefPicList1[block_ref]->m_pYPlane + uOffset
                + curr_slice->m_InitialOffset[pFields1[block_ref]]: 0 : 0;

            switch (cur_mb.GlobalMacroblockInfo->sbtype[uBlock])
            {
                case SBTYPE_8x8:
                case SBTYPE_FORWARD_8x8:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, pitchPixels, iXType, iYType);

                    // interpolate copy the 8x8 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels,pDst, 16, iXType, iYType, size8x8, pMVFwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pDst, sizeof(PixType)*16, iXType, iYType, size8x8, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_BACKWARD_8x8:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, pitchPixels, iXType, iYType);

                    // interpolate copy the 8x8 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels,pDst, 16, iXType, iYType, size8x8, pMVBwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pDst, sizeof(PixType)*16, iXType, iYType, size8x8, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_8x4:
                case SBTYPE_FORWARD_8x4:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, pitchPixels, iXType, iYType);

                    // interpolate copy the first 8x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels,pDst, 16, iXType, iYType, size8x4, pMVFwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pDst, sizeof(PixType)*16, iXType, iYType, size8x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVFwd + 4;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + pitchPixels*4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4;

                    // interpolate copy the second 8x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels,pMCDst, 16, iXType, iYType, size8x4, pMCMV->mvy, uMBy, sb_pos + 4);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size8x4, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_BACKWARD_8x4:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, pitchPixels, iXType, iYType);

                    // interpolate copy the first 8x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pDst, 16, iXType, iYType, size8x4, pMVBwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType), pDst, sizeof(PixType)*16, iXType, iYType, size8x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVBwd + 4;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + pitchPixels*4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4;

                    // interpolate copy the second 8x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size8x4, pMCMV->mvy, uMBy, sb_pos + 4);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType), pMCDst, sizeof(PixType)*16, iXType, iYType, size8x4, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_4x8:
                case SBTYPE_FORWARD_4x8:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, pitchPixels, iXType, iYType);

                    // interpolate copy the first 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pDst, 16, iXType, iYType, size4x8, pMVFwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType), pDst, sizeof(PixType)*16, iXType, iYType, size4x8, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVFwd + 1;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + 4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + 4;

                    // interpolate copy the second 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x8, pMCMV->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x8, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_BACKWARD_4x8:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, pitchPixels, iXType, iYType);

                    // interpolate copy the first 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pDst, 16, iXType, iYType, size4x8, pMVBwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType), pDst, sizeof(PixType)*16, iXType, iYType, size4x8, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVBwd + 1;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + 4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + 4;

                    // interpolate copy the second 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x8, pMCMV->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x8, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_4x4:
                case SBTYPE_FORWARD_4x4:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, pitchPixels, iXType, iYType);

                    // interpolate copy the first 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pDst, 16, iXType, iYType, size4x4, pMVFwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVFwd + 1;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + 4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + 4;

                    // interpolate copy the second 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x4, pMCMV->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 2
                    pMCMV = pMVFwd + 4;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + pitchPixels*4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4;

                    // interpolate copy the third 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x4, pMCMV->mvy, uMBy, sb_pos + 2);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 3
                    pMCMV = pMVFwd + 4 + 1;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + pitchPixels*4 + 4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4 + 4;

                    // interpolate copy the fourth 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x4, pMCMV->mvy, uMBy, sb_pos + 2);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_BACKWARD_4x4:
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, pitchPixels, iXType, iYType);

                    // interpolate copy the first 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pDst, 16, iXType, iYType, size4x4, pMVBwd->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVBwd + 1;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + 4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + 4;

                    // interpolate copy the second 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x4, pMCMV->mvy, uMBy, sb_pos);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 2
                    pMCMV = pMVBwd + 4;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + pitchPixels*4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4;

                    // interpolate copy the third 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x4, pMCMV->mvy, uMBy, sb_pos + 2);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif

                    // Update MV, Ref & Dst pointers for block 3
                    pMCMV = pMVBwd + 4 + 1;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + pitchPixels*4 + 4 + SubpelMVAdjust(pMCMV, pitchPixels, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4 + 4;

                    // interpolate copy the fourth 4x4 block
#ifdef NEW_INTERPOLATE
                    Interpolate(mb_decoding_flag, pMCRef, pitchPixels, pMCDst, 16, iXType, iYType, size4x4, pMCMV->mvy, uMBy, sb_pos + 2);
#else
                    ippiInterpolateLuma_H264(pMCRef, pitchPixels*sizeof(PixType),pMCDst, sizeof(PixType)*16, iXType, iYType, size4x4, m_PicParamSet.bit_depth_luma);
#endif
                    break;

                case SBTYPE_BIDIR_8x8:
                case SBTYPE_BIDIR_8x4:
                case SBTYPE_BIDIR_4x8:
                case SBTYPE_BIDIR_4x4: {
                    // Copy the stored 8x8 prediction data from the BiPred Mode Buffer
                    Ipp32u uRow;
                    PixType* pBiPred = curr_slice->m_pPred4BiPred + (uBlock&1)*8 + (uBlock&2)*16*4;
                    pMCDst = pDst;

                    for (uRow=0; uRow<8; uRow++) {
                        memcpy(pMCDst, pBiPred, 8*sizeof(PixType));
                        pMCDst += DEST_PITCH;
                        pBiPred += 16;
                    }
                }
                break;

                case SBTYPE_DIRECT: {
                    // Copy the stored 8x8 prediction data from the Direct Mode Buffer
                    Ipp32u uRow;
                    PixType* pDirect = curr_slice->m_pPred4DirectB + (uBlock&1)*8 + (uBlock&2)*16*4;
                    pMCDst = pDst;

                    for (uRow=0; uRow<8; uRow++) {
                        memcpy(pMCDst, pDirect, 8*sizeof(PixType));
                        pMCDst += DEST_PITCH;
                        pDirect += 16;
                    }
                }
                break;

                default:
                    break;

            }

            // advance MV pointers and source ptrs to next 8x8 block
            if (1 == uBlock)