void WelsIChromaPredDc_c (uint8_t* pPred, uint8_t* pRef, const int32_t kiStride) {
const int32_t kuiL1 = kiStride - 1;
const int32_t kuiL2 = kuiL1 + kiStride;
const int32_t kuiL3 = kuiL2 + kiStride;
const int32_t kuiL4 = kuiL3 + kiStride;
const int32_t kuiL5 = kuiL4 + kiStride;
const int32_t kuiL6 = kuiL5 + kiStride;
const int32_t kuiL7 = kuiL6 + kiStride;
/*caculate the iMean value*/
const uint8_t kuiMean1 = (pRef[-kiStride] + pRef[1 - kiStride] + pRef[2 - kiStride] + pRef[3 - kiStride] +
pRef[-1] + pRef[kuiL1] + pRef[kuiL2] + pRef[kuiL3] + 4) >> 3;
const uint32_t kuiSum2 = pRef[4 - kiStride] + pRef[5 - kiStride] + pRef[6 - kiStride] + pRef[7 - kiStride];
const uint32_t kuiSum3 = pRef[kuiL4] + pRef[kuiL5] + pRef[kuiL6] + pRef[kuiL7];
const uint8_t kuiMean2 = (kuiSum2 + 2) >> 2;
const uint8_t kuiMean3 = (kuiSum3 + 2) >> 2;
const uint8_t kuiMean4 = (kuiSum2 + kuiSum3 + 4) >> 3;
const uint8_t kuiTopMean[8] = {kuiMean1, kuiMean1, kuiMean1, kuiMean1, kuiMean2, kuiMean2, kuiMean2, kuiMean2};
const uint8_t kuiBottomMean[8] = {kuiMean3, kuiMean3, kuiMean3, kuiMean3, kuiMean4, kuiMean4, kuiMean4, kuiMean4};
const uint64_t kuiTopMean64 = LD64 (kuiTopMean);
const uint64_t kuiBottomMean64 = LD64 (kuiBottomMean);
ST64 (pPred , kuiTopMean64);
ST64 (pPred + 8 , kuiTopMean64);
ST64 (pPred + 16, kuiTopMean64);
ST64 (pPred + 24, kuiTopMean64);
ST64 (pPred + 32, kuiBottomMean64);
ST64 (pPred + 40, kuiBottomMean64);
ST64 (pPred + 48, kuiBottomMean64);
ST64 (pPred + 56, kuiBottomMean64);
}
void WelsIChromaPredDc_c (uint8_t* pPred, const int32_t kiStride) {
const int32_t kiL1 = kiStride - 1;
const int32_t kiL2 = kiL1 + kiStride;
const int32_t kiL3 = kiL2 + kiStride;
const int32_t kiL4 = kiL3 + kiStride;
const int32_t kiL5 = kiL4 + kiStride;
const int32_t kiL6 = kiL5 + kiStride;
const int32_t kiL7 = kiL6 + kiStride;
/*caculate the kMean value*/
const uint8_t kuiM1 = (pPred[-kiStride] + pPred[1 - kiStride] + pPred[2 - kiStride] + pPred[3 - kiStride] +
pPred[-1] + pPred[kiL1] + pPred[kiL2] + pPred[kiL3] + 4) >> 3 ;
const uint32_t kuiSum2 = pPred[4 - kiStride] + pPred[5 - kiStride] + pPred[6 - kiStride] + pPred[7 - kiStride];
const uint32_t kuiSum3 = pPred[kiL4] + pPred[kiL5] + pPred[kiL6] + pPred[kiL7];
const uint8_t kuiM2 = (kuiSum2 + 2) >> 2;
const uint8_t kuiM3 = (kuiSum3 + 2) >> 2;
const uint8_t kuiM4 = (kuiSum2 + kuiSum3 + 4) >> 3;
const uint8_t kuiMUP[8] = {kuiM1, kuiM1, kuiM1, kuiM1, kuiM2, kuiM2, kuiM2, kuiM2};
const uint8_t kuiMDown[8] = {kuiM3, kuiM3, kuiM3, kuiM3, kuiM4, kuiM4, kuiM4, kuiM4};
const uint64_t kuiUP64 = LD64 (kuiMUP);
const uint64_t kuiDN64 = LD64 (kuiMDown);
ST64 (pPred , kuiUP64);
ST64 (pPred + kiL1 + 1, kuiUP64);
ST64 (pPred + kiL2 + 1, kuiUP64);
ST64 (pPred + kiL3 + 1, kuiUP64);
ST64 (pPred + kiL4 + 1, kuiDN64);
ST64 (pPred + kiL5 + 1, kuiDN64);
ST64 (pPred + kiL6 + 1, kuiDN64);
ST64 (pPred + kiL7 + 1, kuiDN64);
}
static inline void McCopyWidthEq16_c (uint8_t* pSrc, int32_t iSrcStride, uint8_t* pDst, int32_t iDstStride,
int32_t iHeight) {
int32_t i;
for (i = 0; i < iHeight; i++) {
ST64 (pDst , LD64 (pSrc));
ST64 (pDst + 8, LD64 (pSrc + 8));
pDst += iDstStride;
pSrc += iSrcStride;
}
}
void WelsI16x16LumaPredV_c (uint8_t* pPred, uint8_t* pRef, const int32_t kiStride) {
uint8_t i = 15;
const int8_t* kpSrc = (int8_t*)&pRef[-kiStride];
const uint64_t kuiT1 = LD64 (kpSrc);
const uint64_t kuiT2 = LD64 (kpSrc + 8);
uint8_t* pDst = pPred;
do {
ST64 (pDst , kuiT1);
ST64 (pDst + 8, kuiT2);
pDst += 16;
} while (i-- > 0);
}
void WelsI16x16LumaPredV_c (uint8_t* pPred, const int32_t kiStride) {
int32_t iTmp = (kiStride << 4) - kiStride;
const uint64_t kuiTop1 = LD64 (pPred - kiStride);
const uint64_t kuiTop2 = LD64 (pPred - kiStride + 8);
uint8_t i = 15;
do {
ST64 (pPred + iTmp , kuiTop1);
ST64 (pPred + iTmp + 8, kuiTop2);
iTmp -= kiStride;
} while (i-- > 0);
}
void WelsIChromaPredV_c (uint8_t* pPred, uint8_t* pRef, const int32_t kiStride) {
const uint64_t kuiSrc64 = LD64 (&pRef[-kiStride]);
ST64 (pPred , kuiSrc64);
ST64 (pPred + 8 , kuiSrc64);
ST64 (pPred + 16, kuiSrc64);
ST64 (pPred + 24, kuiSrc64);
ST64 (pPred + 32, kuiSrc64);
ST64 (pPred + 40, kuiSrc64);
ST64 (pPred + 48, kuiSrc64);
ST64 (pPred + 56, kuiSrc64);
}
void WelsIChromaPredDcTop_c (uint8_t* pPred, uint8_t* pRef, const int32_t kiStride) {
/*caculate the iMean value*/
const uint8_t kuiMean1 = (pRef[-kiStride] + pRef[1 - kiStride] + pRef[2 - kiStride] + pRef[3 - kiStride] + 2) >> 2;
const uint8_t kuiMean2 = (pRef[4 - kiStride] + pRef[5 - kiStride] + pRef[6 - kiStride] + pRef[7 - kiStride] + 2) >> 2;
const uint8_t kuiMean[8] = {kuiMean1, kuiMean1, kuiMean1, kuiMean1, kuiMean2, kuiMean2, kuiMean2, kuiMean2};
const uint64_t kuiMean64 = LD64 (kuiMean);
ST64 (pPred , kuiMean64);
ST64 (pPred + 8 , kuiMean64);
ST64 (pPred + 16, kuiMean64);
ST64 (pPred + 24, kuiMean64);
ST64 (pPred + 32, kuiMean64);
ST64 (pPred + 40, kuiMean64);
ST64 (pPred + 48, kuiMean64);
ST64 (pPred + 56, kuiMean64);
}
void WelsIChromaPredDcTop_c (uint8_t* pPred, const int32_t kiStride) {
int32_t iTmp = (kiStride << 3) - kiStride;
/*caculate the kMean value*/
const uint8_t kuiM1 = (pPred[-kiStride] + pPred[1 - kiStride] + pPred[2 - kiStride] + pPred[3 - kiStride] + 2) >> 2;
const uint8_t kuiM2 = (pPred[4 - kiStride] + pPred[5 - kiStride] + pPred[6 - kiStride] + pPred[7 - kiStride] + 2) >>
2;
const uint8_t kuiM[8] = {kuiM1, kuiM1, kuiM1, kuiM1, kuiM2, kuiM2, kuiM2, kuiM2};
uint8_t i = 7;
do {
ST64 (pPred + iTmp, LD64 (kuiM));
iTmp -= kiStride;
} while (i-- > 0);
}
void WelsFillCacheInter (PNeighAvail pNeighAvail, uint8_t* pNonZeroCount,
int16_t iMvArray[LIST_A][30][MV_A], int8_t iRefIdxArray[LIST_A][30], PDqLayer pCurLayer) {
int32_t iCurXy = pCurLayer->iMbXyIndex;
int32_t iTopXy = 0;
int32_t iLeftXy = 0;
int32_t iLeftTopXy = 0;
int32_t iRightTopXy = 0;
//stuff non_zero_coeff_count from pNeighAvail(left and top)
WelsFillCacheNonZeroCount (pNeighAvail, pNonZeroCount, pCurLayer);
if (pNeighAvail->iTopAvail) {
iTopXy = iCurXy - pCurLayer->iMbWidth;
}
if (pNeighAvail->iLeftAvail) {
iLeftXy = iCurXy - 1;
}
if (pNeighAvail->iLeftTopAvail) {
iLeftTopXy = iCurXy - 1 - pCurLayer->iMbWidth;
}
if (pNeighAvail->iRightTopAvail) {
iRightTopXy = iCurXy + 1 - pCurLayer->iMbWidth;
}
//stuff mv_cache and iRefIdxArray from left and top (inter)
if (pNeighAvail->iLeftAvail && IS_INTER (pNeighAvail->iLeftType)) {
ST32 (iMvArray[0][ 6], LD32 (pCurLayer->pMv[0][iLeftXy][ 3]));
ST32 (iMvArray[0][12], LD32 (pCurLayer->pMv[0][iLeftXy][ 7]));
ST32 (iMvArray[0][18], LD32 (pCurLayer->pMv[0][iLeftXy][11]));
ST32 (iMvArray[0][24], LD32 (pCurLayer->pMv[0][iLeftXy][15]));
iRefIdxArray[0][ 6] = pCurLayer->pRefIndex[0][iLeftXy][ 3];
iRefIdxArray[0][12] = pCurLayer->pRefIndex[0][iLeftXy][ 7];
iRefIdxArray[0][18] = pCurLayer->pRefIndex[0][iLeftXy][11];
iRefIdxArray[0][24] = pCurLayer->pRefIndex[0][iLeftXy][15];
} else {
ST32 (iMvArray[0][ 6], 0);
ST32 (iMvArray[0][12], 0);
ST32 (iMvArray[0][18], 0);
ST32 (iMvArray[0][24], 0);
if (0 == pNeighAvail->iLeftAvail) { //not available
iRefIdxArray[0][ 6] =
iRefIdxArray[0][12] =
iRefIdxArray[0][18] =
iRefIdxArray[0][24] = REF_NOT_AVAIL;
} else { //available but is intra mb type
iRefIdxArray[0][ 6] =
iRefIdxArray[0][12] =
iRefIdxArray[0][18] =
iRefIdxArray[0][24] = REF_NOT_IN_LIST;
}
}
if (pNeighAvail->iLeftTopAvail && IS_INTER (pNeighAvail->iLeftTopType)) {
ST32 (iMvArray[0][0], LD32 (pCurLayer->pMv[0][iLeftTopXy][15]));
iRefIdxArray[0][0] = pCurLayer->pRefIndex[0][iLeftTopXy][15];
} else {
ST32 (iMvArray[0][0], 0);
if (0 == pNeighAvail->iLeftTopAvail) { //not available
iRefIdxArray[0][0] = REF_NOT_AVAIL;
} else { //available but is intra mb type
iRefIdxArray[0][0] = REF_NOT_IN_LIST;
}
}
if (pNeighAvail->iTopAvail && IS_INTER (pNeighAvail->iTopType)) {
ST64 (iMvArray[0][1], LD64 (pCurLayer->pMv[0][iTopXy][12]));
ST64 (iMvArray[0][3], LD64 (pCurLayer->pMv[0][iTopXy][14]));
ST32 (&iRefIdxArray[0][1], LD32 (&pCurLayer->pRefIndex[0][iTopXy][12]));
} else {
ST64 (iMvArray[0][1], 0);
ST64 (iMvArray[0][3], 0);
if (0 == pNeighAvail->iTopAvail) { //not available
iRefIdxArray[0][1] =
iRefIdxArray[0][2] =
iRefIdxArray[0][3] =
iRefIdxArray[0][4] = REF_NOT_AVAIL;
} else { //available but is intra mb type
iRefIdxArray[0][1] =
iRefIdxArray[0][2] =
iRefIdxArray[0][3] =
iRefIdxArray[0][4] = REF_NOT_IN_LIST;
}
}
if (pNeighAvail->iRightTopAvail && IS_INTER (pNeighAvail->iRightTopType)) {
ST32 (iMvArray[0][5], LD32 (pCurLayer->pMv[0][iRightTopXy][12]));
iRefIdxArray[0][5] = pCurLayer->pRefIndex[0][iRightTopXy][12];
} else {
ST32 (iMvArray[0][5], 0);
if (0 == pNeighAvail->iRightTopAvail) { //not available
iRefIdxArray[0][5] = REF_NOT_AVAIL;
} else { //available but is intra mb type
iRefIdxArray[0][5] = REF_NOT_IN_LIST;
}
}
//right-top 4*4 block unavailable
ST32 (iMvArray[0][ 9], 0);
ST32 (iMvArray[0][21], 0);
ST32 (iMvArray[0][11], 0);
ST32 (iMvArray[0][17], 0);
ST32 (iMvArray[0][23], 0);
iRefIdxArray[0][ 9] =
iRefIdxArray[0][21] =
iRefIdxArray[0][11] =
iRefIdxArray[0][17] =
iRefIdxArray[0][23] = REF_NOT_AVAIL;
}
static inline void WelsFillingPred1to16_c (uint8_t* pPred, const uint8_t kuiSrc) {
const uint8_t kuiSrc8[8] = { kuiSrc, kuiSrc, kuiSrc, kuiSrc, kuiSrc, kuiSrc, kuiSrc, kuiSrc };
ST64 (pPred , LD64 (kuiSrc8));
ST64 (pPred + 8, LD64 (kuiSrc8));
}
static inline void WelsFillingPred8x2to16_c (uint8_t* pPred, uint8_t* pSrc) {
ST64 (pPred , LD64 (pSrc));
ST64 (pPred + 8, LD64 (pSrc + 8));
}
