Incorrect use of getComponentScaleX() for LM, DMVR, and QPA

[wangyang.cs]

1. In IntraPrediction::xGetLMParameters()

  const int unitWidth    = baseUnitSize >> getComponentScaleX(chromaArea.compID, nChromaFormat);
  const int unitHeight   = baseUnitSize >> getComponentScaleX(chromaArea.compID, nChromaFormat);

should be corrected as following:

  const int unitWidth    = baseUnitSize >> getComponentScaleX(chromaArea.compID, nChromaFormat);
  const int unitHeight   = baseUnitSize >> getComponentScaleY(chromaArea.compID, nChromaFormat);

2. In InterPrediction::xPrefetch()

    int mvshiftTemp = mvShift + getComponentScaleX((ComponentID)compID, pu.chromaFormat);
    width += (filtersize - 1);
    height += (filtersize - 1);
    cMv += Mv(-(((filtersize >> 1) - 1) << mvshiftTemp),
      -(((filtersize >> 1) - 1) << mvshiftTemp));
...
      Position Rec_offset = pu.blocks[compID].pos().offset(cMv.getHor() >> mvshiftTemp, cMv.getVer() >> mvshiftTemp);

should be corrected as following:

    int mvshiftTempHor = mvShift + getComponentScaleX((ComponentID)compID, pu.chromaFormat);
    int mvshiftTempVer = mvShift + getComponentScaleY((ComponentID)compID, pu.chromaFormat);
    width += (filtersize - 1);
    height += (filtersize - 1);
    cMv += Mv(-(((filtersize >> 1) - 1) << mvshiftTempHor),
      -(((filtersize >> 1) - 1) << mvshiftTempVer));
...
      Position Rec_offset = pu.blocks[compID].pos().offset(cMv.getHor() >> mvshiftTempHor, cMv.getVer() >> mvshiftTempVer);

3. In InterPrediction::xPad()

Different pad sizes should be used for width and height in 4:2:2 sequences. The simplest way to solve it is to use the larger one.

padsize = (DMVR_NUM_ITERATION) >> getComponentScaleX((ComponentID)compID, pu.chromaFormat);

should be corrected as following:

padsize = (DMVR_NUM_ITERATION) >> getComponentScaleY((ComponentID)compID, pu.chromaFormat);

4. In InterPrediction::xFinalPaddedMCForDMVR()

        int mvshiftTemp = mvShift + getComponentScaleX((ComponentID)compID, pu.chromaFormat);
        int leftPixelExtra;
        if (compID == COMPONENT_Y)
        {
          leftPixelExtra = (NTAPS_LUMA >> 1) - 1;
        }
        else
        {
          leftPixelExtra = (NTAPS_CHROMA >> 1) - 1;
        }
        PelBuf &srcBuf = pcPadTemp.bufs[compID];
        deltaIntMvX = (cMv.getHor() >> mvshiftTemp) -
          (startMv.getHor() >> mvshiftTemp);
        deltaIntMvY = (cMv.getVer() >> mvshiftTemp) -
          (startMv.getVer() >> mvshiftTemp);

should be corrected as following:

        int mvshiftTempHor = mvShift + getComponentScaleX((ComponentID)compID, pu.chromaFormat);
        int mvshiftTempVer = mvShift + getComponentScaleY((ComponentID)compID, pu.chromaFormat);
        int leftPixelExtra;
        if (compID == COMPONENT_Y)
        {
          leftPixelExtra = (NTAPS_LUMA >> 1) - 1;
        }
        else
        {
          leftPixelExtra = (NTAPS_CHROMA >> 1) - 1;
        }
        PelBuf &srcBuf = pcPadTemp.bufs[compID];
        deltaIntMvX = (cMv.getHor() >> mvshiftTempHor) -
          (startMv.getHor() >> mvshiftTempHor);
        deltaIntMvY = (cMv.getVer() >> mvshiftTempVer) -
          (startMv.getVer() >> mvshiftTempVer);

5. In EncGOP::xFindDistortionPlane()

In JVET-H0047, upper limit B_max (i.e. 128 >> chromaShift) equals 64 for 4:2:0 downsampled chroma components and 128 otherwise.

uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1, const uint32_t rshift
#if ENABLE_QPA
                                    , const uint32_t chromaShift /*= 0*/
#endif
                                      )
...
const uint32_t   B = Clip3<uint32_t>(0, 128 >> chromaShift, 4 * uint32_t(16.0 * sqrt(R) + 0.5)); // WPSNR block size in integer multiple of 4 (for SIMD, = 64 at full-HD)
...
sumAct = 16.0 * sqrt ((3840.0 * 2160.0) / double((W << chromaShift) * (H << chromaShift))) * double(1 << BD);

should be corrected as following:

uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1, const uint32_t rshift
#if ENABLE_QPA
                                    , const uint32_t chromaShiftHor /*= 0*/, const uint32_t chromaShiftVer /*= 0*/
#endif
                                      )
...
const uint32_t   B = Clip3<uint32_t>(0, 128 >> chromaShiftVer, 4 * uint32_t(16.0 * sqrt(R) + 0.5)); // WPSNR block size in integer multiple of 4 (for SIMD, = 64 at full-HD)
...
sumAct = 16.0 * sqrt ((3840.0 * 2160.0) / double((W << chromaShiftHor) * (H << chromaShiftVer))) * double(1 << BD);