umc_h264_mc.cpp

这是在PCA下的基于IPP库示例代码例子,在网上下了IPP的库之后,设置相关参数就可以编译该代码.

        // Copy the stored 2nd 16x8 prediction data from the BiPred Mode Buffer
        pBiPred = m_pCurrentFrame->pPred4BiPred + 128;
        pMCDst = pDst + DEST_PITCH*8;
        for (uRow=0; uRow<8; uRow++) {
            memcpy(pMCDst, pBiPred, 16);
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }
        break;

    case MBTYPE_BIDIR_BWD_16x8:
        // Copy the stored 16x8 prediction data from the BiPred Mode Buffer
        pBiPred = m_pCurrentFrame->pPred4BiPred;
        pMCDst = pDst;
        for (uRow=0; uRow<8; uRow++) {
            memcpy(pMCDst, pBiPred, 16);
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }

        // Do other 16x8 block
        pMVBwd += uWidthIn4x4Blocks * 2;      // Reposition MV pointer to next 16x8
        pDst += DEST_PITCH*8;

        block_ref = pRefIdxL1[uWidthIn4x4Blocks * 2];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset + uPitch*8;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 16x8 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size16x8);
        break;

    case MBTYPE_BWD_BIDIR_16x8:
        // 2 vectors for the MB
        block_ref = pRefIdxL1[0];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 16x8 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size16x8);

        // Copy the stored 2nd 16x8 prediction data from the BiPred Mode Buffer

        pBiPred = m_pCurrentFrame->pPred4BiPred + 128;
        pMCDst = pDst + DEST_PITCH*8;
        for (uRow=0; uRow<8; uRow++) {
            memcpy(pMCDst, pBiPred, 16);
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }
        break;

    case MBTYPE_INTER_8x16:
    case MBTYPE_FWD_FWD_8x16:
        // 2 vectors for the MB
        block_ref = pRefIdxL0[0];
        prev_frame = pRefPicList0[block_ref]->m_pYPlane + uOffset;
        pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);

        // Do other 8x16 block
        pMVFwd += 2;                    // Reposition MV pointer to next 8x16
        pDst += 8;

        block_ref = pRefIdxL0[2];
        prev_frame = pRefPicList0[block_ref]->m_pYPlane + uOffset + 8;
        pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);
        break;

    case MBTYPE_FWD_BWD_8x16:
        // 2 vectors for the MB
        block_ref = pRefIdxL0[0];
        prev_frame = pRefPicList0[block_ref]->m_pYPlane + uOffset;
        pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);

        // Do other 8x16 block
        pMVBwd += 2;                    // Reposition MV pointer to next 8x16
        pDst += 8;

        block_ref = pRefIdxL1[2];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset + 8;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);
        break;

    case MBTYPE_BWD_FWD_8x16:
        // 2 vectors for the MB
        block_ref = pRefIdxL1[0];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);

        // Do other 8x16 block
        pMVFwd += 2;                    // Reposition MV pointer to next 8x16
        pDst += 8;

        block_ref = pRefIdxL0[2];
        prev_frame = pRefPicList0[block_ref]->m_pYPlane + uOffset + 8;
        pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);
        break;

    case MBTYPE_BWD_BWD_8x16:
        // 2 vectors for the MB
        block_ref = pRefIdxL1[0];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);

        // Do other 8x16 block
        pMVBwd += 2;                    // Reposition MV pointer to next 8x16
        pDst += 8;

        block_ref = pRefIdxL1[2];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset + 8;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);
        break;

    case MBTYPE_BIDIR_FWD_8x16:
        // Copy the stored 8x16 prediction data from the BiPred Mode Buffer
        pBiPred = m_pCurrentFrame->pPred4BiPred;
        pMCDst = pDst;
        for (uRow=0; uRow<16; uRow++) {
            memcpy(pMCDst, pBiPred, 8);
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }

        // Do other 8x16 block
        pMVFwd += 2;                    // Reposition MV pointer to next 8x16
        pDst += 8;

        block_ref = pRefIdxL0[2];
        prev_frame = pRefPicList0[block_ref]->m_pYPlane + uOffset + 8;
        pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);
        break;

    case MBTYPE_FWD_BIDIR_8x16:
        // 2 vectors for the MB
        block_ref = pRefIdxL0[0];
        prev_frame = pRefPicList0[block_ref]->m_pYPlane + uOffset;
        pMCRef = prev_frame + SubpelMVAdjust(pMVFwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);

        // Copy the stored 2nd 8x16 prediction data from the BiPred Mode Buffer
        pBiPred = m_pCurrentFrame->pPred4BiPred + 8;
        pMCDst = pDst + 8;
        for (uRow=0; uRow<16; uRow++) {
            memcpy(pMCDst, pBiPred, 8);
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }
        break;

    case MBTYPE_BIDIR_BWD_8x16:
        // Copy the stored 8x16 prediction data from the BiPred Mode Buffer
        pBiPred = m_pCurrentFrame->pPred4BiPred;
        pMCDst = pDst;
        for (uRow=0; uRow<16; uRow++) {
            memcpy(pMCDst, pBiPred, 8);
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }

        // Do other 8x16 block
        pMVBwd += 2;                    // Reposition MV pointer to next 8x16
        pDst += 8;

        block_ref = pRefIdxL1[2];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset + 8;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);
        break;

    case MBTYPE_BWD_BIDIR_8x16:
        // 2 vectors for the MB
        block_ref = pRefIdxL1[0];
        futr_frame = pRefPicList1[block_ref]->m_pYPlane + uOffset;
        pMCRef = futr_frame + SubpelMVAdjust(pMVBwd, uPitch, iXType, iYType);

        // interpolate copy the 8x16 block
        ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x16);

        // Copy the stored 2nd 8x16 prediction data from the BiPred Mode Buffer
        pBiPred = m_pCurrentFrame->pPred4BiPred + 8;
        pMCDst = pDst + 8;
        for (uRow=0; uRow<16; uRow++) {
            memcpy(pMCDst, pBiPred, 8);
            pMCDst += DEST_PITCH;
            pBiPred += 16;
        }
        break;

    case MBTYPE_INTER_8x8_REF0:
    case MBTYPE_INTER_8x8:
    case MBTYPE_B_8x8:
        {
        int ref_off = 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 : 0;
            block_ref = pRefIdxL1[ref_off];
            futr_frame = block_ref >= 0 ? pRefPicList1[block_ref] ?
                pRefPicList1[block_ref]->m_pYPlane + uOffset : 0 : 0;

            switch (pMBData->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, uPitch, iXType, iYType);

                    // interpolate copy the 8x8 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x8);
                    break;

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

                    // interpolate copy the 8x8 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x8);
                    break;

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

                    // interpolate copy the first 8x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x4);

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVFwd + uWidthIn4x4Blocks;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = prev_frame + uPitch*4 + SubpelMVAdjust(pMCMV, uPitch, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4;

                    // interpolate copy the second 8x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pMCDst, 16, iXType, iYType, size8x4);
                    break;

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

                    // interpolate copy the first 8x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size8x4);

                    // Update MV, Ref & Dst pointers for block 1
                    pMCMV = pMVBwd + uWidthIn4x4Blocks;
                    // convert 1/4 pel vector to pel vector and interpolation type
                    pMCRef = futr_frame + uPitch*4 + SubpelMVAdjust(pMCMV, uPitch, iXType, iYType);
                    pMCDst = pDst + DEST_PITCH*4;

                    // interpolate copy the second 8x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pMCDst, 16, iXType, iYType, size8x4);
                    break;

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

                    // interpolate copy the first 4x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size4x8);

                    // 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, uPitch, iXType, iYType);
                    pMCDst = pDst + 4;

                    // interpolate copy the second 4x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pMCDst, 16, iXType, iYType, size4x8);
                    break;

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

                    // interpolate copy the first 4x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pDst, 16, iXType, iYType, size4x8);

                    // 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, uPitch, iXType, iYType);
                    pMCDst = pDst + 4;

                    // interpolate copy the second 4x4 block
                    ippiInterpolateLuma_H264_8u_C1R(pMCRef, uPitch,pMCDst, 16, iXType, iYType, size4x8);
                    break;

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