re-structure SWelsDecoderContext and others and prepare for the support of multi-threading support.

xiaotiansf committed 7 years ago

Commit 2f5ecec1 parent 9ed97c729 files · +2285 −1568Browse files

Showing the changes to codec/decoder/core/src/mv_pred.cpp (1 of 29 files). Show all

Unified · Split

codec/decoder/core/src/mv_pred.cpp+244−198

@@ -155,7 +155,7 @@ void CopyRectBlock4Cols (void* vdst, void* vsrc, const int32_t stride_dst, const · 154 unmodified lines
155	155	memcpy (&dst[stride_dst * 3], &src[stride_src * 3], 16);
156	156	}
157	157	}
158		−void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) {
	158	+void PredPSkipMvFromNeighbor (PDqLayer pCurDqLayer, int16_t iMvp[2]) {
159	159	bool bTopAvail, bLeftTopAvail, bRightTopAvail, bLeftAvail;
160	160
161	161	int32_t iCurSliceIdc, iTopSliceIdc, iLeftTopSliceIdc, iRightTopSliceIdc, iLeftSliceIdc;
@@ -170,14 +170,14 @@ void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) { · 8 unmodified lines
170	170	int8_t iMatchRef;
171	171	int16_t iMvA[2], iMvB[2], iMvC[2], iMvD[2];
172	172
173		− iCurXy = pCurLayer->iMbXyIndex;
174		− iCurX = pCurLayer->iMbX;
175		− iCurY = pCurLayer->iMbY;
176		− iCurSliceIdc = pCurLayer->pSliceIdc[iCurXy];
	173	+ iCurXy = pCurDqLayer->iMbXyIndex;
	174	+ iCurX = pCurDqLayer->iMbX;
	175	+ iCurY = pCurDqLayer->iMbY;
	176	+ iCurSliceIdc = pCurDqLayer->pSliceIdc[iCurXy];
177	177
178	178	if (iCurX != 0) {
179	179	iLeftXy = iCurXy - 1;
180		− iLeftSliceIdc = pCurLayer->pSliceIdc[iLeftXy];
	180	+ iLeftSliceIdc = pCurDqLayer->pSliceIdc[iLeftXy];
181	181	bLeftAvail = (iLeftSliceIdc == iCurSliceIdc);
182	182	} else {
183	183	bLeftAvail = 0;
@@ -185,19 +185,19 @@ void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) { · 1 unmodified line
185	185	}
186	186
187	187	if (iCurY != 0) {
188		− iTopXy = iCurXy - pCurLayer->iMbWidth;
189		− iTopSliceIdc = pCurLayer->pSliceIdc[iTopXy];
	188	+ iTopXy = iCurXy - pCurDqLayer->iMbWidth;
	189	+ iTopSliceIdc = pCurDqLayer->pSliceIdc[iTopXy];
190	190	bTopAvail = (iTopSliceIdc == iCurSliceIdc);
191	191	if (iCurX != 0) {
192	192	iLeftTopXy = iTopXy - 1;
193		− iLeftTopSliceIdc = pCurLayer->pSliceIdc[iLeftTopXy];
	193	+ iLeftTopSliceIdc = pCurDqLayer->pSliceIdc[iLeftTopXy];
194	194	bLeftTopAvail = (iLeftTopSliceIdc == iCurSliceIdc);
195	195	} else {
196	196	bLeftTopAvail = 0;
197	197	}
198		− if (iCurX != (pCurLayer->iMbWidth - 1)) {
	198	+ if (iCurX != (pCurDqLayer->iMbWidth - 1)) {
199	199	iRightTopXy = iTopXy + 1;
200		− iRightTopSliceIdc = pCurLayer->pSliceIdc[iRightTopXy];
	200	+ iRightTopSliceIdc = pCurDqLayer->pSliceIdc[iRightTopXy];
201	201	bRightTopAvail = (iRightTopSliceIdc == iCurSliceIdc);
202	202	} else {
203	203	bRightTopAvail = 0;
@@ -208,18 +208,18 @@ void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) { · 4 unmodified lines
208	208	bRightTopAvail = 0;
209	209	}
210	210
211		− iLeftType = ((iCurX != 0 && bLeftAvail) ? pCurLayer->pMbType[iLeftXy] : 0);
212		− iTopType = ((iCurY != 0 && bTopAvail) ? pCurLayer->pMbType[iTopXy] : 0);
	211	+ iLeftType = ((iCurX != 0 && bLeftAvail) ? GetMbType (pCurDqLayer)[iLeftXy] : 0);
	212	+ iTopType = ((iCurY != 0 && bTopAvail) ? GetMbType (pCurDqLayer)[iTopXy] : 0);
213	213	iLeftTopType = ((iCurX != 0 && iCurY != 0 && bLeftTopAvail)
214		− ? pCurLayer->pMbType[iLeftTopXy] : 0);
215		− iRightTopType = ((iCurX != pCurLayer->iMbWidth - 1 && iCurY != 0 && bRightTopAvail)
216		− ? pCurLayer->pMbType[iRightTopXy] : 0);
	214	+ ? GetMbType (pCurDqLayer)[iLeftTopXy] : 0);
	215	+ iRightTopType = ((iCurX != pCurDqLayer->iMbWidth - 1 && iCurY != 0 && bRightTopAvail)
	216	+ ? GetMbType (pCurDqLayer)[iRightTopXy] : 0);
217	217
218	218	/get neb mv&iRefIdxArray/
219	219	/left/
220	220	if (bLeftAvail && IS_INTER (iLeftType)) {
221		− ST32 (iMvA, LD32 (pCurLayer->pMv[0][iLeftXy][3]));
222		− iLeftRef = pCurLayer->pRefIndex[0][iLeftXy][3];
	221	+ ST32 (iMvA, LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[0][iLeftXy][3] : pCurDqLayer->pMv[0][iLeftXy][3]));
	222	+ iLeftRef = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[0][iLeftXy][3] : pCurDqLayer->pRefIndex[0][iLeftXy][3];
223	223	} else {
224	224	ST32 (iMvA, 0);
225	225	if (0 == bLeftAvail) { //not available
@@ -236,8 +236,8 @@ void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) { · 10 unmodified lines
236	236
237	237	/top/
238	238	if (bTopAvail && IS_INTER (iTopType)) {
239		− ST32 (iMvB, LD32 (pCurLayer->pMv[0][iTopXy][12]));
240		− iTopRef = pCurLayer->pRefIndex[0][iTopXy][12];
	239	+ ST32 (iMvB, LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[0][iTopXy][12] : pCurDqLayer->pMv[0][iTopXy][12]));
	240	+ iTopRef = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[0][iTopXy][12] : pCurDqLayer->pRefIndex[0][iTopXy][12];
241	241	} else {
242	242	ST32 (iMvB, 0);
243	243	if (0 == bTopAvail) { //not available
@@ -254,8 +254,10 @@ void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) { · 10 unmodified lines
254	254
255	255	/right_top/
256	256	if (bRightTopAvail && IS_INTER (iRightTopType)) {
257		− ST32 (iMvC, LD32 (pCurLayer->pMv[0][iRightTopXy][12]));
258		− iRightTopRef = pCurLayer->pRefIndex[0][iRightTopXy][12];
	257	+ ST32 (iMvC, LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[0][iRightTopXy][12] :
	258	+ pCurDqLayer->pMv[0][iRightTopXy][12]));
	259	+ iRightTopRef = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[0][iRightTopXy][12] :
	260	+ pCurDqLayer->pRefIndex[0][iRightTopXy][12];
259	261	} else {
260	262	ST32 (iMvC, 0);
261	263	if (0 == bRightTopAvail) { //not available
@@ -267,8 +269,9 @@ void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) { · 5 unmodified lines
267	269
268	270	/left_top/
269	271	if (bLeftTopAvail && IS_INTER (iLeftTopType)) {
270		− ST32 (iMvD, LD32 (pCurLayer->pMv[0][iLeftTopXy][15]));
271		− iLeftTopRef = pCurLayer->pRefIndex[0][iLeftTopXy][15];
	272	+ ST32 (iMvD, LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[0][iLeftTopXy][15] : pCurDqLayer->pMv[0][iLeftTopXy][15]));
	273	+ iLeftTopRef = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[0][iLeftTopXy][15] :
	274	+ pCurDqLayer->pRefIndex[0][iLeftTopXy][15];
272	275	} else {
273	276	ST32 (iMvD, 0);
274	277	if (0 == bLeftTopAvail) { //not available
@@ -305,11 +308,11 @@ void PredPSkipMvFromNeighbor (PDqLayer pCurLayer, int16_t iMvp[2]) { · 30 unmodified lines
305	308	}
306	309
307	310	int32_t GetColocatedMb (PWelsDecoderContext pCtx, MbType& mbType, SubMbType& subMbType) {
308		− PDqLayer pCurLayer = pCtx->pCurDqLayer;
309		− int32_t iMbXy = pCurLayer->iMbXyIndex;
	311	+ PDqLayer pCurDqLayer = pCtx->pCurDqLayer;
	312	+ int32_t iMbXy = pCurDqLayer->iMbXyIndex;
310	313
311		− uint32_t is8x8 = IS_Inter_8x8 (pCurLayer->pMbType[iMbXy]);
312		− mbType = pCurLayer->pMbType[iMbXy];
	314	+ uint32_t is8x8 = IS_Inter_8x8 (GetMbType (pCurDqLayer)[iMbXy]);
	315	+ mbType = GetMbType (pCurDqLayer)[iMbXy];
313	316
314	317	PPicture colocPic = pCtx->sRefPic.pRefList[LIST_1][0];
315	318
@@ -336,43 +339,43 @@ int32_t GetColocatedMb (PWelsDecoderContext pCtx, MbType& mbType, SubMbType& sub · 20 unmodified lines
336	339	}
337	340
338	341	if (IS_INTRA (coloc_mbType)) {
339		− SetRectBlock (pCurLayer->iColocIntra, 4, 4, 4 * sizeof (int8_t), 1, sizeof (int8_t));
	342	+ SetRectBlock (pCurDqLayer->iColocIntra, 4, 4, 4 * sizeof (int8_t), 1, sizeof (int8_t));
340	343	return ERR_NONE;
341	344	}
342		− SetRectBlock (pCurLayer->iColocIntra, 4, 4, 4 * sizeof (int8_t), 0, sizeof (int8_t));
	345	+ SetRectBlock (pCurDqLayer->iColocIntra, 4, 4, 4 * sizeof (int8_t), 0, sizeof (int8_t));
343	346
344	347	if (IS_INTER_16x16 (mbType)) {
345	348	int16_t iMVZero[2] = { 0 };
346	349	int16_t* pMv = IS_TYPE_L1 (coloc_mbType) ? colocPic->pMv[LIST_1][iMbXy][0] : iMVZero;
347		− ST32 (pCurLayer->iColocMv[LIST_0][0], LD32 (colocPic->pMv[LIST_0][iMbXy][0]));
348		− ST32 (pCurLayer->iColocMv[LIST_1][0], LD32 (pMv));
349		− pCurLayer->iColocRefIndex[LIST_0][0] = colocPic->pRefIndex[LIST_0][iMbXy][0];
350		− pCurLayer->iColocRefIndex[LIST_1][0] = IS_TYPE_L1 (coloc_mbType) ? colocPic->pRefIndex[LIST_1][iMbXy][0] :
351		− REF_NOT_IN_LIST;
	350	+ ST32 (pCurDqLayer->iColocMv[LIST_0][0], LD32 (colocPic->pMv[LIST_0][iMbXy][0]));
	351	+ ST32 (pCurDqLayer->iColocMv[LIST_1][0], LD32 (pMv));
	352	+ pCurDqLayer->iColocRefIndex[LIST_0][0] = colocPic->pRefIndex[LIST_0][iMbXy][0];
	353	+ pCurDqLayer->iColocRefIndex[LIST_1][0] = IS_TYPE_L1 (coloc_mbType) ? colocPic->pRefIndex[LIST_1][iMbXy][0] :
	354	+ REF_NOT_IN_LIST;
352	355	} else {
353	356	if (!pCtx->pSps->bDirect8x8InferenceFlag) {
354		− CopyRectBlock4Cols (pCurLayer->iColocMv[LIST_0], colocPic->pMv[LIST_0][iMbXy], 16, 16, 4, 4);
355		− CopyRectBlock4Cols (pCurLayer->iColocRefIndex[LIST_0], colocPic->pRefIndex[LIST_0][iMbXy], 4, 4, 4, 1);
	357	+ CopyRectBlock4Cols (pCurDqLayer->iColocMv[LIST_0], colocPic->pMv[LIST_0][iMbXy], 16, 16, 4, 4);
	358	+ CopyRectBlock4Cols (pCurDqLayer->iColocRefIndex[LIST_0], colocPic->pRefIndex[LIST_0][iMbXy], 4, 4, 4, 1);
356	359	if (IS_TYPE_L1 (coloc_mbType)) {
357		− CopyRectBlock4Cols (pCurLayer->iColocMv[LIST_1], colocPic->pMv[LIST_1][iMbXy], 16, 16, 4, 4);
358		− CopyRectBlock4Cols (pCurLayer->iColocRefIndex[LIST_1], colocPic->pRefIndex[LIST_1][iMbXy], 4, 4, 4, 1);
	360	+ CopyRectBlock4Cols (pCurDqLayer->iColocMv[LIST_1], colocPic->pMv[LIST_1][iMbXy], 16, 16, 4, 4);
	361	+ CopyRectBlock4Cols (pCurDqLayer->iColocRefIndex[LIST_1], colocPic->pRefIndex[LIST_1][iMbXy], 4, 4, 4, 1);
359	362	} else { // only forward prediction
360		− SetRectBlock (pCurLayer->iColocRefIndex[LIST_1], 4, 4, 4, (uint8_t)REF_NOT_IN_LIST, 1);
	363	+ SetRectBlock (pCurDqLayer->iColocRefIndex[LIST_1], 4, 4, 4, (uint8_t)REF_NOT_IN_LIST, 1);
361	364	}
362	365	} else {
363	366	for (int32_t listIdx = 0; listIdx < 1 + !! (coloc_mbType & MB_TYPE_L1); listIdx++) {
364		− SetRectBlock (pCurLayer->iColocMv[listIdx][0], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][0]), 4);
365		− SetRectBlock (pCurLayer->iColocMv[listIdx][2], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][3]), 4);
366		− SetRectBlock (pCurLayer->iColocMv[listIdx][8], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][12]), 4);
367		− SetRectBlock (pCurLayer->iColocMv[listIdx][10], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][15]), 4);
368		−
369		− SetRectBlock (&pCurLayer->iColocRefIndex[listIdx][0], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][0], 1);
370		− SetRectBlock (&pCurLayer->iColocRefIndex[listIdx][2], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][3], 1);
371		− SetRectBlock (&pCurLayer->iColocRefIndex[listIdx][8], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][12], 1);
372		− SetRectBlock (&pCurLayer->iColocRefIndex[listIdx][10], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][15], 1);
	367	+ SetRectBlock (pCurDqLayer->iColocMv[listIdx][0], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][0]), 4);
	368	+ SetRectBlock (pCurDqLayer->iColocMv[listIdx][2], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][3]), 4);
	369	+ SetRectBlock (pCurDqLayer->iColocMv[listIdx][8], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][12]), 4);
	370	+ SetRectBlock (pCurDqLayer->iColocMv[listIdx][10], 2, 2, 16, LD32 (colocPic->pMv[listIdx][iMbXy][15]), 4);
	371	+
	372	+ SetRectBlock (&pCurDqLayer->iColocRefIndex[listIdx][0], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][0], 1);
	373	+ SetRectBlock (&pCurDqLayer->iColocRefIndex[listIdx][2], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][3], 1);
	374	+ SetRectBlock (&pCurDqLayer->iColocRefIndex[listIdx][8], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][12], 1);
	375	+ SetRectBlock (&pCurDqLayer->iColocRefIndex[listIdx][10], 2, 2, 4, colocPic->pRefIndex[listIdx][iMbXy][15], 1);
373	376	}
374	377	if (! (coloc_mbType & MB_TYPE_L1)) // only forward prediction
375		− SetRectBlock (&pCurLayer->iColocRefIndex[1][0], 4, 4, 4, (uint8_t)REF_NOT_IN_LIST, 1);
	378	+ SetRectBlock (&pCurDqLayer->iColocRefIndex[1][0], 4, 4, 4, (uint8_t)REF_NOT_IN_LIST, 1);
376	379	}
377	380	}
378	381	return ERR_NONE;
@@ -382,9 +385,9 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 3 unmodified lines
382	385	SubMbType& subMbType) {
383	386
384	387	int32_t ret = ERR_NONE;
385		− PDqLayer pCurLayer = pCtx->pCurDqLayer;
386		− int32_t iMbXy = pCurLayer->iMbXyIndex;
387		− bool bSkipOrDirect = (IS_SKIP (pCurLayer->pMbType[iMbXy]) \| IS_DIRECT (pCurLayer->pMbType[iMbXy])) > 0;
	388	+ PDqLayer pCurDqLayer = pCtx->pCurDqLayer;
	389	+ int32_t iMbXy = pCurDqLayer->iMbXyIndex;
	390	+ bool bSkipOrDirect = (IS_SKIP (GetMbType (pCurDqLayer)[iMbXy]) \| IS_DIRECT (GetMbType (pCurDqLayer)[iMbXy])) > 0;
388	391
389	392	MbType mbType;
390	393	ret = GetColocatedMb (pCtx, mbType, subMbType);
@@ -404,15 +407,15 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 13 unmodified lines
404	407	int8_t iDiagonalRef[LIST_A];
405	408	int16_t iMvA[LIST_A][2], iMvB[LIST_A][2], iMvC[LIST_A][2], iMvD[LIST_A][2];
406	409
407		− iCurXy = pCurLayer->iMbXyIndex;
	410	+ iCurXy = pCurDqLayer->iMbXyIndex;
408	411
409		− iCurX = pCurLayer->iMbX;
410		− iCurY = pCurLayer->iMbY;
411		− iCurSliceIdc = pCurLayer->pSliceIdc[iCurXy];
	412	+ iCurX = pCurDqLayer->iMbX;
	413	+ iCurY = pCurDqLayer->iMbY;
	414	+ iCurSliceIdc = pCurDqLayer->pSliceIdc[iCurXy];
412	415
413	416	if (iCurX != 0) {
414	417	iLeftXy = iCurXy - 1;
415		− iLeftSliceIdc = pCurLayer->pSliceIdc[iLeftXy];
	418	+ iLeftSliceIdc = pCurDqLayer->pSliceIdc[iLeftXy];
416	419	bLeftAvail = (iLeftSliceIdc == iCurSliceIdc);
417	420	} else {
418	421	bLeftAvail = 0;
@@ -420,19 +423,19 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 1 unmodified line
420	423	}
421	424
422	425	if (iCurY != 0) {
423		− iTopXy = iCurXy - pCurLayer->iMbWidth;
424		− iTopSliceIdc = pCurLayer->pSliceIdc[iTopXy];
	426	+ iTopXy = iCurXy - pCurDqLayer->iMbWidth;
	427	+ iTopSliceIdc = pCurDqLayer->pSliceIdc[iTopXy];
425	428	bTopAvail = (iTopSliceIdc == iCurSliceIdc);
426	429	if (iCurX != 0) {
427	430	iLeftTopXy = iTopXy - 1;
428		− iLeftTopSliceIdc = pCurLayer->pSliceIdc[iLeftTopXy];
	431	+ iLeftTopSliceIdc = pCurDqLayer->pSliceIdc[iLeftTopXy];
429	432	bLeftTopAvail = (iLeftTopSliceIdc == iCurSliceIdc);
430	433	} else {
431	434	bLeftTopAvail = 0;
432	435	}
433		− if (iCurX != (pCurLayer->iMbWidth - 1)) {
	436	+ if (iCurX != (pCurDqLayer->iMbWidth - 1)) {
434	437	iRightTopXy = iTopXy + 1;
435		− iRightTopSliceIdc = pCurLayer->pSliceIdc[iRightTopXy];
	438	+ iRightTopSliceIdc = pCurDqLayer->pSliceIdc[iRightTopXy];
436	439	bRightTopAvail = (iRightTopSliceIdc == iCurSliceIdc);
437	440	} else {
438	441	bRightTopAvail = 0;
@@ -443,20 +446,22 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 4 unmodified lines
443	446	bRightTopAvail = 0;
444	447	}
445	448
446		− iLeftType = ((iCurX != 0 && bLeftAvail) ? pCurLayer->pMbType[iLeftXy] : 0);
447		− iTopType = ((iCurY != 0 && bTopAvail) ? pCurLayer->pMbType[iTopXy] : 0);
	449	+ iLeftType = ((iCurX != 0 && bLeftAvail) ? GetMbType (pCurDqLayer)[iLeftXy] : 0);
	450	+ iTopType = ((iCurY != 0 && bTopAvail) ? GetMbType (pCurDqLayer)[iTopXy] : 0);
448	451	iLeftTopType = ((iCurX != 0 && iCurY != 0 && bLeftTopAvail)
449		− ? pCurLayer->pMbType[iLeftTopXy] : 0);
450		− iRightTopType = ((iCurX != pCurLayer->iMbWidth - 1 && iCurY != 0 && bRightTopAvail)
451		− ? pCurLayer->pMbType[iRightTopXy] : 0);
	452	+ ? GetMbType (pCurDqLayer)[iLeftTopXy] : 0);
	453	+ iRightTopType = ((iCurX != pCurDqLayer->iMbWidth - 1 && iCurY != 0 && bRightTopAvail)
	454	+ ? GetMbType (pCurDqLayer)[iRightTopXy] : 0);
452	455
453	456	/get neb mv&iRefIdxArray/
454	457	for (int32_t listIdx = LIST_0; listIdx < LIST_A; ++listIdx) {
455	458
456	459	/left/
457	460	if (bLeftAvail && IS_INTER (iLeftType)) {
458		− ST32 (iMvA[listIdx], LD32 (pCurLayer->pMv[listIdx][iLeftXy][3]));
459		− iLeftRef[listIdx] = pCurLayer->pRefIndex[listIdx][iLeftXy][3];
	461	+ ST32 (iMvA[listIdx], LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[listIdx][iLeftXy][3] :
	462	+ pCurDqLayer->pMv[listIdx][iLeftXy][3]));
	463	+ iLeftRef[listIdx] = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[listIdx][iLeftXy][3] :
	464	+ pCurDqLayer->pRefIndex[listIdx][iLeftXy][3];
460	465	} else {
461	466	ST32 (iMvA[listIdx], 0);
462	467	if (0 == bLeftAvail) { //not available
@@ -468,8 +473,10 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 5 unmodified lines
468	473
469	474	/top/
470	475	if (bTopAvail && IS_INTER (iTopType)) {
471		− ST32 (iMvB[listIdx], LD32 (pCurLayer->pMv[listIdx][iTopXy][12]));
472		− iTopRef[listIdx] = pCurLayer->pRefIndex[listIdx][iTopXy][12];
	476	+ ST32 (iMvB[listIdx], LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[listIdx][iTopXy][12] :
	477	+ pCurDqLayer->pMv[listIdx][iTopXy][12]));
	478	+ iTopRef[listIdx] = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[listIdx][iTopXy][12] :
	479	+ pCurDqLayer->pRefIndex[listIdx][iTopXy][12];
473	480	} else {
474	481	ST32 (iMvB[listIdx], 0);
475	482	if (0 == bTopAvail) { //not available
@@ -481,8 +488,10 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 5 unmodified lines
481	488
482	489	/right_top/
483	490	if (bRightTopAvail && IS_INTER (iRightTopType)) {
484		− ST32 (iMvC[listIdx], LD32 (pCurLayer->pMv[listIdx][iRightTopXy][12]));
485		− iRightTopRef[listIdx] = pCurLayer->pRefIndex[listIdx][iRightTopXy][12];
	491	+ ST32 (iMvC[listIdx], LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[listIdx][iRightTopXy][12] :
	492	+ pCurDqLayer->pMv[listIdx][iRightTopXy][12]));
	493	+ iRightTopRef[listIdx] = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[listIdx][iRightTopXy][12] :
	494	+ pCurDqLayer->pRefIndex[listIdx][iRightTopXy][12];
486	495	} else {
487	496	ST32 (iMvC[listIdx], 0);
488	497	if (0 == bRightTopAvail) { //not available
@@ -493,8 +502,10 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 4 unmodified lines
493	502	}
494	503	/left_top/
495	504	if (bLeftTopAvail && IS_INTER (iLeftTopType)) {
496		− ST32 (iMvD[listIdx], LD32 (pCurLayer->pMv[listIdx][iLeftTopXy][15]));
497		− iLeftTopRef[listIdx] = pCurLayer->pRefIndex[listIdx][iLeftTopXy][15];
	505	+ ST32 (iMvD[listIdx], LD32 (pCurDqLayer->pDec ? pCurDqLayer->pDec->pMv[listIdx][iLeftTopXy][15] :
	506	+ pCurDqLayer->pMv[listIdx][iLeftTopXy][15]));
	507	+ iLeftTopRef[listIdx] = pCurDqLayer->pDec ? pCurDqLayer->pDec->pRefIndex[listIdx][iLeftTopXy][15] :
	508	+ pCurDqLayer->pRefIndex[listIdx][iLeftTopXy][15];
498	509	} else {
499	510	ST32 (iMvD[listIdx], 0);
500	511	if (0 == bLeftTopAvail) { //not available
@@ -543,7 +554,7 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 42 unmodified lines
543	554	mbType &= ~MB_TYPE_L0;
544	555	subMbType &= ~MB_TYPE_L0;
545	556	}
546		− pCurLayer->pMbType[iMbXy] = mbType;
	557	+ GetMbType (pCurDqLayer)[iMbXy] = mbType;
547	558
548	559	int16_t pMvd[4] = { 0 };
549	560
@@ -551,31 +562,31 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 1 unmodified line
551	562
552	563	if (IS_INTER_16x16 (mbType)) {
553	564	if ((* (int32_t)iMvp[LIST_0] \| (int32_t*)iMvp[LIST_1])) {
554		− if (0 == pCurLayer->iColocIntra[0] && !bIsLongRef
555		− && ((pCurLayer->iColocRefIndex[LIST_0][0] == 0 && (unsigned) (pCurLayer->iColocMv[LIST_0][0][0] + 1) <= 2
556		− && (unsigned) (pCurLayer->iColocMv[LIST_0][0][1] + 1) <= 2)
557		− \|\| (pCurLayer->iColocRefIndex[LIST_0][0] < 0 && pCurLayer->iColocRefIndex[LIST_1][0] == 0
558		− && (unsigned) (pCurLayer->iColocMv[LIST_1][0][0] + 1) <= 2
559		− && (unsigned) (pCurLayer->iColocMv[LIST_1][0][1] + 1) <= 2))) {
	565	+ if (0 == pCurDqLayer->iColocIntra[0] && !bIsLongRef
	566	+ && ((pCurDqLayer->iColocRefIndex[LIST_0][0] == 0 && (unsigned) (pCurDqLayer->iColocMv[LIST_0][0][0] + 1) <= 2
	567	+ && (unsigned) (pCurDqLayer->iColocMv[LIST_0][0][1] + 1) <= 2)
	568	+ \|\| (pCurDqLayer->iColocRefIndex[LIST_0][0] < 0 && pCurDqLayer->iColocRefIndex[LIST_1][0] == 0
	569	+ && (unsigned) (pCurDqLayer->iColocMv[LIST_1][0][0] + 1) <= 2
	570	+ && (unsigned) (pCurDqLayer->iColocMv[LIST_1][0][1] + 1) <= 2))) {
560	571	if (0 >= ref[0]) * (uint32_t*)iMvp[LIST_0] = 0;
561	572	if (0 >= ref[1]) * (uint32_t*)iMvp[LIST_1] = 0;
562	573	}
563	574	}
564		− UpdateP16x16DirectCabac (pCurLayer);
	575	+ UpdateP16x16DirectCabac (pCurDqLayer);
565	576	for (int32_t listIdx = LIST_0; listIdx < LIST_A; ++listIdx) {
566		− UpdateP16x16MotionInfo (pCurLayer, listIdx, ref[listIdx], iMvp[listIdx]);
567		− UpdateP16x16MvdCabac (pCurLayer, pMvd, listIdx);
	577	+ UpdateP16x16MotionInfo (pCurDqLayer, listIdx, ref[listIdx], iMvp[listIdx]);
	578	+ UpdateP16x16MvdCabac (pCurDqLayer, pMvd, listIdx);
568	579	}
569	580	} else {
570	581	if (bSkipOrDirect) {
571	582	int8_t pSubPartCount[4], pPartW[4];
572	583	for (int32_t i = 0; i < 4; i++) { //Direct 8x8 Ref and mv
573	584	int16_t iIdx8 = i << 2;
574		− pCurLayer->pSubMbType[iMbXy][i] = subMbType;
	585	+ pCurDqLayer->pSubMbType[iMbXy][i] = subMbType;
575	586	int8_t pRefIndex[LIST_A][30];
576		− UpdateP8x8RefIdxCabac (pCurLayer, pRefIndex, iIdx8, ref[LIST_0], LIST_0);
577		− UpdateP8x8RefIdxCabac (pCurLayer, pRefIndex, iIdx8, ref[LIST_1], LIST_1);
578		− UpdateP8x8DirectCabac (pCurLayer, iIdx8);
	587	+ UpdateP8x8RefIdxCabac (pCurDqLayer, pRefIndex, iIdx8, ref[LIST_0], LIST_0);
	588	+ UpdateP8x8RefIdxCabac (pCurDqLayer, pRefIndex, iIdx8, ref[LIST_1], LIST_1);
	589	+ UpdateP8x8DirectCabac (pCurDqLayer, iIdx8);
579	590
580	591	pSubPartCount[i] = g_ksInterBSubMbTypeInfo[0].iPartCount;
581	592	pPartW[i] = g_ksInterBSubMbTypeInfo[0].iPartWidth;
@@ -584,7 +595,7 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 2 unmodified lines
584	595	pSubPartCount[i] = 4;
585	596	pPartW[i] = 1;
586	597	}
587		− FillSpatialDirect8x8Mv (pCurLayer, iIdx8, pSubPartCount[i], pPartW[i], subMbType, bIsLongRef, iMvp, ref, NULL, NULL);
	598	+ FillSpatialDirect8x8Mv (pCurDqLayer, iIdx8, pSubPartCount[i], pPartW[i], subMbType, bIsLongRef, iMvp, ref, NULL, NULL);
588	599	}
589	600	}
590	601	}
@@ -594,9 +605,9 @@ int32_t PredMvBDirectSpatial (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 3 unmodified lines
594	605	int32_t PredBDirectTemporal (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], int8_t ref[LIST_A],
595	606	SubMbType& subMbType) {
596	607	int32_t ret = ERR_NONE;
597		− PDqLayer pCurLayer = pCtx->pCurDqLayer;
598		− int32_t iMbXy = pCurLayer->iMbXyIndex;
599		− bool bSkipOrDirect = (IS_SKIP (pCurLayer->pMbType[iMbXy]) \| IS_DIRECT (pCurLayer->pMbType[iMbXy])) > 0;
	608	+ PDqLayer pCurDqLayer = pCtx->pCurDqLayer;
	609	+ int32_t iMbXy = pCurDqLayer->iMbXyIndex;
	610	+ bool bSkipOrDirect = (IS_SKIP (GetMbType (pCurDqLayer)[iMbXy]) \| IS_DIRECT (GetMbType (pCurDqLayer)[iMbXy])) > 0;
600	611
601	612	MbType mbType;
602	613	ret = GetColocatedMb (pCtx, mbType, subMbType);
@@ -604,42 +615,42 @@ int32_t PredBDirectTemporal (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 1 unmodified line
604	615	return ret;
605	616	}
606	617
607		− pCurLayer->pMbType[iMbXy] = mbType;
	618	+ GetMbType (pCurDqLayer)[iMbXy] = mbType;
608	619
609		− PSlice pSlice = &pCurLayer->sLayerInfo.sSliceInLayer;
	620	+ PSlice pSlice = &pCurDqLayer->sLayerInfo.sSliceInLayer;
610	621	PSliceHeader pSliceHeader = &pSlice->sSliceHeaderExt.sSliceHeader;
611	622	int16_t pMvd[4] = { 0 };
612	623	const int32_t ref0Count = WELS_MIN (pSliceHeader->uiRefCount[LIST_0], pCtx->sRefPic.uiRefCount[LIST_0]);
613	624	if (IS_INTER_16x16 (mbType)) {
614	625	ref[LIST_0] = 0;
615	626	ref[LIST_1] = 0;
616		− UpdateP16x16DirectCabac (pCurLayer);
617		− UpdateP16x16RefIdx (pCurLayer, LIST_1, ref[LIST_1]);
	627	+ UpdateP16x16DirectCabac (pCurDqLayer);
	628	+ UpdateP16x16RefIdx (pCurDqLayer, LIST_1, ref[LIST_1]);
618	629	ST64 (iMvp, 0);
619		− if (pCurLayer->iColocIntra[0]) {
620		− UpdateP16x16MotionOnly (pCurLayer, LIST_0, iMvp[LIST_0]);
621		− UpdateP16x16MotionOnly (pCurLayer, LIST_1, iMvp[LIST_1]);
622		− UpdateP16x16RefIdx (pCurLayer, LIST_0, ref[LIST_0]);
	630	+ if (pCurDqLayer->iColocIntra[0]) {
	631	+ UpdateP16x16MotionOnly (pCurDqLayer, LIST_0, iMvp[LIST_0]);
	632	+ UpdateP16x16MotionOnly (pCurDqLayer, LIST_1, iMvp[LIST_1]);
	633	+ UpdateP16x16RefIdx (pCurDqLayer, LIST_0, ref[LIST_0]);
623	634	} else {
624	635	ref[LIST_0] = 0;
625		− int16_t* mv = pCurLayer->iColocMv[LIST_0][0];
626		− int8_t colocRefIndexL0 = pCurLayer->iColocRefIndex[LIST_0][0];
	636	+ int16_t* mv = pCurDqLayer->iColocMv[LIST_0][0];
	637	+ int8_t colocRefIndexL0 = pCurDqLayer->iColocRefIndex[LIST_0][0];
627	638	if (colocRefIndexL0 >= 0) {
628	639	ref[LIST_0] = MapColToList0 (pCtx, colocRefIndexL0, ref0Count);
629	640	} else {
630		− mv = pCurLayer->iColocMv[LIST_1][0];
	641	+ mv = pCurDqLayer->iColocMv[LIST_1][0];
631	642	}
632		− UpdateP16x16RefIdx (pCurLayer, LIST_0, ref[LIST_0]);
	643	+ UpdateP16x16RefIdx (pCurDqLayer, LIST_0, ref[LIST_0]);
633	644
634	645	iMvp[LIST_0][0] = (pSlice->iMvScale[LIST_0][ref[LIST_0]] * mv[0] + 128) >> 8;
635	646	iMvp[LIST_0][1] = (pSlice->iMvScale[LIST_0][ref[LIST_0]] * mv[1] + 128) >> 8;
636		− UpdateP16x16MotionOnly (pCurLayer, LIST_0, iMvp[LIST_0]);
	647	+ UpdateP16x16MotionOnly (pCurDqLayer, LIST_0, iMvp[LIST_0]);
637	648	iMvp[LIST_1][0] = iMvp[LIST_0][0] - mv[0];
638	649	iMvp[LIST_1][1] = iMvp[LIST_0][1] - mv[1];
639		− UpdateP16x16MotionOnly (pCurLayer, LIST_1, iMvp[LIST_1]);
	650	+ UpdateP16x16MotionOnly (pCurDqLayer, LIST_1, iMvp[LIST_1]);
640	651	}
641		− UpdateP16x16MvdCabac (pCurLayer, pMvd, LIST_0);
642		− UpdateP16x16MvdCabac (pCurLayer, pMvd, LIST_1);
	652	+ UpdateP16x16MvdCabac (pCurDqLayer, pMvd, LIST_0);
	653	+ UpdateP16x16MvdCabac (pCurDqLayer, pMvd, LIST_1);
643	654	} else {
644	655	if (bSkipOrDirect) {
645	656	int8_t pSubPartCount[4], pPartW[4];
@@ -647,27 +658,27 @@ int32_t PredBDirectTemporal (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 1 unmodified line
647	658	for (int32_t i = 0; i < 4; i++) {
648	659	int16_t iIdx8 = i << 2;
649	660	const uint8_t iScan4Idx = g_kuiScan4[iIdx8];
650		− pCurLayer->pSubMbType[iMbXy][i] = subMbType;
	661	+ pCurDqLayer->pSubMbType[iMbXy][i] = subMbType;
651	662
652		− int16_t (*mvColoc)[2] = pCurLayer->iColocMv[LIST_0];
	663	+ int16_t (*mvColoc)[2] = pCurDqLayer->iColocMv[LIST_0];
653	664
654	665	ref[LIST_1] = 0;
655		− UpdateP8x8RefIdxCabac (pCurLayer, pRefIndex, iIdx8, ref[LIST_1], LIST_1);
656		− if (pCurLayer->iColocIntra[iScan4Idx]) {
	666	+ UpdateP8x8RefIdxCabac (pCurDqLayer, pRefIndex, iIdx8, ref[LIST_1], LIST_1);
	667	+ if (pCurDqLayer->iColocIntra[iScan4Idx]) {
657	668	ref[LIST_0] = 0;
658		− UpdateP8x8RefIdxCabac (pCurLayer, pRefIndex, iIdx8, ref[LIST_0], LIST_0);
	669	+ UpdateP8x8RefIdxCabac (pCurDqLayer, pRefIndex, iIdx8, ref[LIST_0], LIST_0);
659	670	ST64 (iMvp, 0);
660	671	} else {
661	672	ref[LIST_0] = 0;
662		− int8_t colocRefIndexL0 = pCurLayer->iColocRefIndex[LIST_0][iScan4Idx];
	673	+ int8_t colocRefIndexL0 = pCurDqLayer->iColocRefIndex[LIST_0][iScan4Idx];
663	674	if (colocRefIndexL0 >= 0) {
664	675	ref[LIST_0] = MapColToList0 (pCtx, colocRefIndexL0, ref0Count);
665	676	} else {
666		− mvColoc = pCurLayer->iColocMv[LIST_1];
	677	+ mvColoc = pCurDqLayer->iColocMv[LIST_1];
667	678	}
668		− UpdateP8x8RefIdxCabac (pCurLayer, pRefIndex, iIdx8, ref[LIST_0], LIST_0);
	679	+ UpdateP8x8RefIdxCabac (pCurDqLayer, pRefIndex, iIdx8, ref[LIST_0], LIST_0);
669	680	}
670		− UpdateP8x8DirectCabac (pCurLayer, iIdx8);
	681	+ UpdateP8x8DirectCabac (pCurDqLayer, iIdx8);
671	682
672	683	pSubPartCount[i] = g_ksInterBSubMbTypeInfo[0].iPartCount;
673	684	pPartW[i] = g_ksInterBSubMbTypeInfo[0].iPartWidth;
@@ -676,7 +687,7 @@ int32_t PredBDirectTemporal (PWelsDecoderContext pCtx, int16_t iMvp[LIST_A][2], · 2 unmodified lines
676	687	pSubPartCount[i] = 4;
677	688	pPartW[i] = 1;
678	689	}
679		− FillTemporalDirect8x8Mv (pCurLayer, iIdx8, pSubPartCount[i], pPartW[i], subMbType, ref, mvColoc, NULL, NULL);
	690	+ FillTemporalDirect8x8Mv (pCurDqLayer, iIdx8, pSubPartCount[i], pPartW[i], subMbType, ref, mvColoc, NULL, NULL);
680	691	}
681	692	}
682	693	}
@@ -785,14 +796,23 @@ void UpdateP16x16MotionInfo (PDqLayer pCurDqLayer, int32_t listIdx, int8_t iRef, · 102 unmodified lines
785	796	//mb
786	797	const uint8_t kuiScan4Idx = g_kuiScan4[i];
787	798	const uint8_t kuiScan4IdxPlus4 = 4 + kuiScan4Idx;
	799	+ if (pCurDqLayer->pDec != NULL) {
	800	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
	801	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
	802	+
	803	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	804	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	805	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	806	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	807	+ } else {
	808	+ ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
	809	+ ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
788	810
789		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4Idx ], kiRef2);
790		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
791		−
792		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][ kuiScan4Idx ], kiMV32);
793		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx ], kiMV32);
794		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][ kuiScan4IdxPlus4], kiMV32);
795		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	811	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	812	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	813	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	814	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	815	+ }
796	816	}
797	817	}
798	818
@@ -808,8 +828,8 @@ void UpdateP16x16RefIdx (PDqLayer pCurDqLayer, int32_t listIdx, int8_t iRef) { · 9 unmodified lines
808	828	const uint8_t kuiScan4Idx = g_kuiScan4[i];
809	829	const uint8_t kuiScan4IdxPlus4 = 4 + kuiScan4Idx;
810	830
811		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
812		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
	831	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
	832	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
813	833	}
814	834	}
815	835
@@ -824,11 +844,17 @@ void UpdateP16x16MotionOnly (PDqLayer pCurDqLayer, int32_t listIdx, int16_t iMVs · 8 unmodified lines
824	844	//mb
825	845	const uint8_t kuiScan4Idx = g_kuiScan4[i];
826	846	const uint8_t kuiScan4IdxPlus4 = 4 + kuiScan4Idx;
827		−
828		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
829		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
830		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
831		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	847	+ if (pCurDqLayer->pDec != NULL) {
	848	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	849	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	850	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	851	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	852	+ } else {
	853	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	854	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	855	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	856	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	857	+ }
832	858	}
833	859	}
834	860
@@ -848,12 +874,21 @@ void UpdateP16x8MotionInfo (PDqLayer pCurDqLayer, int16_t iMotionVector[LIST_A][ · 13 unmodified lines
848	874	const uint8_t kuiCacheIdxPlus6 = 6 + kuiCacheIdx;
849	875
850	876	//mb
851		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4Idx ], kiRef2);
852		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
853		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][ kuiScan4Idx ], kiMV32);
854		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx ], kiMV32);
855		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][ kuiScan4IdxPlus4], kiMV32);
856		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	877	+ if (pCurDqLayer->pDec != NULL) {
	878	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
	879	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
	880	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	881	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	882	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	883	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	884	+ } else {
	885	+ ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
	886	+ ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
	887	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	888	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	889	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	890	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	891	+ }
857	892	//cache
858	893	ST16 (&iRefIndex[listIdx][kuiCacheIdx ], kiRef2);
859	894	ST16 (&iRefIndex[listIdx][kuiCacheIdxPlus6], kiRef2);
@@ -879,12 +914,21 @@ void UpdateP8x16MotionInfo (PDqLayer pCurDqLayer, int16_t iMotionVector[LIST_A][ · 19 unmodified lines
879	914	const uint8_t kuiCacheIdxPlus6 = 6 + kuiCacheIdx;
880	915
881	916	//mb
882		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4Idx ], kiRef2);
883		− ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
884		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][ kuiScan4Idx ], kiMV32);
885		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx ], kiMV32);
886		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][ kuiScan4IdxPlus4], kiMV32);
887		− ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	917	+ if (pCurDqLayer->pDec != NULL) {
	918	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
	919	+ ST16 (&pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
	920	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	921	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	922	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	923	+ ST32 (pCurDqLayer->pDec->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	924	+ } else {
	925	+ ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4Idx], kiRef2);
	926	+ ST16 (&pCurDqLayer->pRefIndex[listIdx][iMbXy][kuiScan4IdxPlus4], kiRef2);
	927	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4Idx], kiMV32);
	928	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4Idx], kiMV32);
	929	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][kuiScan4IdxPlus4], kiMV32);
	930	+ ST32 (pCurDqLayer->pMv[listIdx][iMbXy][1 + kuiScan4IdxPlus4], kiMV32);
	931	+ }
888	932	//cache
889	933	ST16 (&iRefIndex[listIdx][kuiCacheIdx ], kiRef2);
890	934	ST16 (&iRefIndex[listIdx][kuiCacheIdxPlus6], kiRef2);
@@ -895,10 +939,10 @@ void UpdateP8x16MotionInfo (PDqLayer pCurDqLayer, int16_t iMotionVector[LIST_A][ · 4 unmodified lines
895	939	}
896	940	}
897	941
898		−void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int8_t& iPartCount, const int8_t& iPartW,
	942	+void FillSpatialDirect8x8Mv (PDqLayer pCurDqLayer, const int16_t& iIdx8, const int8_t& iPartCount, const int8_t& iPartW,
899	943	const SubMbType& subMbType, const bool& bIsLongRef, int16_t pMvDirect[LIST_A][2], int8_t iRef[LIST_A],
900	944	int16_t pMotionVector[LIST_A][30][MV_A], int16_t pMvdCache[LIST_A][30][MV_A]) {
901		− int32_t iMbXy = pCurLayer->iMbXyIndex;
	945	+ int32_t iMbXy = pCurDqLayer->iMbXyIndex;
902	946	for (int32_t j = 0; j < iPartCount; j++) {
903	947	int8_t iPartIdx = iIdx8 + j * iPartW;
904	948	uint8_t iScan4Idx = g_kuiScan4[iPartIdx];
@@ -909,10 +953,10 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 4 unmodified lines
909	953	if (IS_SUB_8x8 (subMbType)) {
910	954	* (uint32_t)pMV = (uint32_t*)pMvDirect[LIST_0];
911	955	ST32 ((pMV + 2), LD32 (pMV));
912		− ST64 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx], LD64 (pMV));
913		− ST64 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx + 4], LD64 (pMV));
914		− ST64 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
915		− ST64 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx + 4], 0);
	956	+ ST64 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx], LD64 (pMV));
	957	+ ST64 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx + 4], LD64 (pMV));
	958	+ ST64 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
	959	+ ST64 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx + 4], 0);
916	960	if (pMotionVector != NULL) {
917	961	ST64 (pMotionVector[LIST_0][iCacheIdx], LD64 (pMV));
918	962	ST64 (pMotionVector[LIST_0][iCacheIdx + 6], LD64 (pMV));
@@ -923,10 +967,10 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 4 unmodified lines
923	967	}
924	968	* (uint32_t)pMV = (uint32_t*)pMvDirect[LIST_1];
925	969	ST32 ((pMV + 2), LD32 (pMV));
926		− ST64 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx], LD64 (pMV));
927		− ST64 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx + 4], LD64 (pMV));
928		− ST64 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
929		− ST64 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx + 4], 0);
	970	+ ST64 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx], LD64 (pMV));
	971	+ ST64 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx + 4], LD64 (pMV));
	972	+ ST64 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
	973	+ ST64 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx + 4], 0);
930	974	if (pMotionVector != NULL) {
931	975	ST64 (pMotionVector[LIST_1][iCacheIdx], LD64 (pMV));
932	976	ST64 (pMotionVector[LIST_1][iCacheIdx + 6], LD64 (pMV));
@@ -937,8 +981,8 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 4 unmodified lines
937	981	}
938	982	} else { //SUB_4x4
939	983	* (uint32_t)pMV = (uint32_t*)pMvDirect[LIST_0];
940		− ST32 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx], LD32 (pMV));
941		− ST32 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
	984	+ ST32 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx], LD32 (pMV));
	985	+ ST32 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
942	986	if (pMotionVector != NULL) {
943	987	ST32 (pMotionVector[LIST_0][iCacheIdx], LD32 (pMV));
944	988	}
@@ -946,8 +990,8 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 1 unmodified line
946	990	ST32 (pMvdCache[LIST_0][iCacheIdx], 0);
947	991	}
948	992	* (uint32_t)pMV = (uint32_t*)pMvDirect[LIST_1];
949		− ST32 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx], LD32 (pMV));
950		− ST32 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
	993	+ ST32 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx], LD32 (pMV));
	994	+ ST32 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
951	995	if (pMotionVector != NULL) {
952	996	ST32 (pMotionVector[LIST_1][iCacheIdx], LD32 (pMV));
953	997	}
@@ -956,19 +1000,19 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 2 unmodified lines
956	1000	}
957	1001	}
958	1002	if ((* (int32_t)pMvDirect[LIST_0] \| (int32_t*)pMvDirect[LIST_1])) {
959		− uint32_t uiColZeroFlag = (0 == pCurLayer->iColocIntra[iColocIdx]) && !bIsLongRef &&
960		− (pCurLayer->iColocRefIndex[LIST_0][iColocIdx] == 0 \|\| (pCurLayer->iColocRefIndex[LIST_0][iColocIdx] < 0
961		− && pCurLayer->iColocRefIndex[LIST_1][iColocIdx] == 0));
962		− const int16_t (*mvColoc)[2] = 0 == pCurLayer->iColocRefIndex[LIST_0][iColocIdx] ? pCurLayer->iColocMv[LIST_0] :
963		− pCurLayer->iColocMv[LIST_1];
	1003	+ uint32_t uiColZeroFlag = (0 == pCurDqLayer->iColocIntra[iColocIdx]) && !bIsLongRef &&
	1004	+ (pCurDqLayer->iColocRefIndex[LIST_0][iColocIdx] == 0 \|\| (pCurDqLayer->iColocRefIndex[LIST_0][iColocIdx] < 0
	1005	+ && pCurDqLayer->iColocRefIndex[LIST_1][iColocIdx] == 0));
	1006	+ const int16_t (*mvColoc)[2] = 0 == pCurDqLayer->iColocRefIndex[LIST_0][iColocIdx] ? pCurDqLayer->iColocMv[LIST_0] :
	1007	+ pCurDqLayer->iColocMv[LIST_1];
964	1008	const int16_t* mv = mvColoc[iColocIdx];
965	1009	if (IS_SUB_8x8 (subMbType)) {
966	1010	if (uiColZeroFlag && ((unsigned) (mv[0] + 1) <= 2 && (unsigned) (mv[1] + 1) <= 2)) {
967	1011	if (iRef[LIST_0] == 0) {
968		− ST64 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx], 0);
969		− ST64 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx + 4], 0);
970		− ST64 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
971		− ST64 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx + 4], 0);
	1012	+ ST64 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx], 0);
	1013	+ ST64 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx + 4], 0);
	1014	+ ST64 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
	1015	+ ST64 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx + 4], 0);
972	1016	if (pMotionVector != NULL) {
973	1017	ST64 (pMotionVector[LIST_0][iCacheIdx], 0);
974	1018	ST64 (pMotionVector[LIST_0][iCacheIdx + 6], 0);
@@ -980,10 +1024,10 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 5 unmodified lines
980	1024	}
981	1025
982	1026	if (iRef[LIST_1] == 0) {
983		− ST64 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx], 0);
984		− ST64 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx + 4], 0);
985		− ST64 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
986		− ST64 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx + 4], 0);
	1027	+ ST64 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx], 0);
	1028	+ ST64 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx + 4], 0);
	1029	+ ST64 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
	1030	+ ST64 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx + 4], 0);
987	1031	if (pMotionVector != NULL) {
988	1032	ST64 (pMotionVector[LIST_1][iCacheIdx], 0);
989	1033	ST64 (pMotionVector[LIST_1][iCacheIdx + 6], 0);
@@ -997,8 +1041,8 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 7 unmodified lines
997	1041	} else {
998	1042	if (uiColZeroFlag && ((unsigned) (mv[0] + 1) <= 2 && (unsigned) (mv[1] + 1) <= 2)) {
999	1043	if (iRef[LIST_0] == 0) {
1000		− ST32 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx], 0);
1001		− ST32 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
	1044	+ ST32 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx], 0);
	1045	+ ST32 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
1002	1046	if (pMotionVector != NULL) {
1003	1047	ST32 (pMotionVector[LIST_0][iCacheIdx], 0);
1004	1048	}
@@ -1007,8 +1051,8 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 2 unmodified lines
1007	1051	}
1008	1052	}
1009	1053	if (iRef[LIST_1] == 0) {
1010		− ST32 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx], 0);
1011		− ST32 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
	1054	+ ST32 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx], 0);
	1055	+ ST32 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
1012	1056	if (pMotionVector != NULL) {
1013	1057	ST32 (pMotionVector[LIST_1][iCacheIdx], 0);
1014	1058	}
@@ -1022,11 +1066,12 @@ void FillSpatialDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int · 7 unmodified lines
1022	1066	}
1023	1067	}
1024	1068
1025		−void FillTemporalDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const int8_t& iPartCount, const int8_t& iPartW,
	1069	+void FillTemporalDirect8x8Mv (PDqLayer pCurDqLayer, const int16_t& iIdx8, const int8_t& iPartCount,
	1070	+ const int8_t& iPartW,
1026	1071	const SubMbType& subMbType, int8_t iRef[LIST_A], int16_t (*mvColoc)[2], int16_t pMotionVector[LIST_A][30][MV_A],
1027	1072	int16_t pMvdCache[LIST_A][30][MV_A]) {
1028		− PSlice pSlice = &pCurLayer->sLayerInfo.sSliceInLayer;
1029		− int32_t iMbXy = pCurLayer->iMbXyIndex;
	1073	+ PSlice pSlice = &pCurDqLayer->sLayerInfo.sSliceInLayer;
	1074	+ int32_t iMbXy = pCurDqLayer->iMbXyIndex;
1030	1075	int16_t pMvDirect[LIST_A][2] = { { 0, 0 }, { 0, 0 } };
1031	1076	for (int32_t j = 0; j < iPartCount; j++) {
1032	1077	int8_t iPartIdx = iIdx8 + j * iPartW;
@@ -1038,16 +1083,16 @@ void FillTemporalDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const in · 5 unmodified lines
1038	1083
1039	1084	int16_t pMV[4] = { 0 };
1040	1085	if (IS_SUB_8x8 (subMbType)) {
1041		− if (!pCurLayer->iColocIntra[iColocIdx]) {
	1086	+ if (!pCurDqLayer->iColocIntra[iColocIdx]) {
1042	1087	pMvDirect[LIST_0][0] = (pSlice->iMvScale[LIST_0][iRef[LIST_0]] * mv[0] + 128) >> 8;
1043	1088	pMvDirect[LIST_0][1] = (pSlice->iMvScale[LIST_0][iRef[LIST_0]] * mv[1] + 128) >> 8;
1044	1089	}
1045	1090	ST32 (pMV, LD32 (pMvDirect[LIST_0]));
1046	1091	ST32 ((pMV + 2), LD32 (pMvDirect[LIST_0]));
1047		− ST64 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx], LD64 (pMV));
1048		− ST64 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx + 4], LD64 (pMV));
1049		− ST64 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
1050		− ST64 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx + 4], 0);
	1092	+ ST64 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx], LD64 (pMV));
	1093	+ ST64 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx + 4], LD64 (pMV));
	1094	+ ST64 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
	1095	+ ST64 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx + 4], 0);
1051	1096	if (pMotionVector != NULL) {
1052	1097	ST64 (pMotionVector[LIST_0][iCacheIdx], LD64 (pMV));
1053	1098	ST64 (pMotionVector[LIST_0][iCacheIdx + 6], LD64 (pMV));
@@ -1056,16 +1101,16 @@ void FillTemporalDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const in · 2 unmodified lines
1056	1101	ST64 (pMvdCache[LIST_0][iCacheIdx], 0);
1057	1102	ST64 (pMvdCache[LIST_0][iCacheIdx + 6], 0);
1058	1103	}
1059		− if (!pCurLayer->iColocIntra[g_kuiScan4[iIdx8]]) {
	1104	+ if (!pCurDqLayer->iColocIntra[g_kuiScan4[iIdx8]]) {
1060	1105	pMvDirect[LIST_1][0] = pMvDirect[LIST_0][0] - mv[0];
1061	1106	pMvDirect[LIST_1][1] = pMvDirect[LIST_0][1] - mv[1];
1062	1107	}
1063	1108	ST32 (pMV, LD32 (pMvDirect[LIST_1]));
1064	1109	ST32 ((pMV + 2), LD32 (pMvDirect[LIST_1]));
1065		− ST64 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx], LD64 (pMV));
1066		− ST64 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx + 4], LD64 (pMV));
1067		− ST64 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
1068		− ST64 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx + 4], 0);
	1110	+ ST64 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx], LD64 (pMV));
	1111	+ ST64 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx + 4], LD64 (pMV));
	1112	+ ST64 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
	1113	+ ST64 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx + 4], 0);
1069	1114	if (pMotionVector != NULL) {
1070	1115	ST64 (pMotionVector[LIST_1][iCacheIdx], LD64 (pMV));
1071	1116	ST64 (pMotionVector[LIST_1][iCacheIdx + 6], LD64 (pMV));
@@ -1075,24 +1120,24 @@ void FillTemporalDirect8x8Mv (PDqLayer pCurLayer, const int16_t& iIdx8, const in · 3 unmodified lines
1075	1120	ST64 (pMvdCache[LIST_1][iCacheIdx + 6], 0);
1076	1121	}
1077	1122	} else { //SUB_4x4
1078		− if (!pCurLayer->iColocIntra[iColocIdx]) {
	1123	+ if (!pCurDqLayer->iColocIntra[iColocIdx]) {
1079	1124	pMvDirect[LIST_0][0] = (pSlice->iMvScale[LIST_0][iRef[LIST_0]] * mv[0] + 128) >> 8;
1080	1125	pMvDirect[LIST_0][1] = (pSlice->iMvScale[LIST_0][iRef[LIST_0]] * mv[1] + 128) >> 8;
1081	1126	}
1082		− ST32 (pCurLayer->pMv[LIST_0][iMbXy][iScan4Idx], LD32 (pMvDirect[LIST_0]));
1083		− ST32 (pCurLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
	1127	+ ST32 (pCurDqLayer->pDec->pMv[LIST_0][iMbXy][iScan4Idx], LD32 (pMvDirect[LIST_0]));
	1128	+ ST32 (pCurDqLayer->pMvd[LIST_0][iMbXy][iScan4Idx], 0);
1084	1129	if (pMotionVector != NULL) {
1085	1130	ST32 (pMotionVector[LIST_0][iCacheIdx], LD32 (pMvDirect[LIST_0]));
1086	1131	}
1087	1132	if (pMvdCache != NULL) {
1088	1133	ST32 (pMvdCache[LIST_0][iCacheIdx], 0);
1089	1134	}
1090		− if (!pCurLayer->iColocIntra[iColocIdx]) {
	1135	+ if (!pCurDqLayer->iColocIntra[iColocIdx]) {
1091	1136	pMvDirect[LIST_1][0] = pMvDirect[LIST_0][0] - mv[0];
1092	1137	pMvDirect[LIST_1][1] = pMvDirect[LIST_0][1] - mv[1];
1093	1138	}
1094		− ST32 (pCurLayer->pMv[LIST_1][iMbXy][iScan4Idx], LD32 (pMvDirect[LIST_1]));
1095		− ST32 (pCurLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
	1139	+ ST32 (pCurDqLayer->pDec->pMv[LIST_1][iMbXy][iScan4Idx], LD32 (pMvDirect[LIST_1]));
	1140	+ ST32 (pCurDqLayer->pMvd[LIST_1][iMbXy][iScan4Idx], 0);
1096	1141	if (pMotionVector != NULL) {
1097	1142	ST32 (pMotionVector[LIST_1][iCacheIdx], LD32 (pMvDirect[LIST_1]));
1098	1143	}
@@ -1122,8 +1167,9 @@ int8_t MapColToList0 (PWelsDecoderContext& pCtx, const int8_t& colocRefIndexL0, · 23 unmodified lines
1122	1167	void Update8x8RefIdx (PDqLayer& pCurDqLayer, const int16_t& iPartIdx, const int32_t& listIdx, const int8_t& iRef) {
1123	1168	int32_t iMbXy = pCurDqLayer->iMbXyIndex;
1124	1169	const uint8_t iScan4Idx = g_kuiScan4[iPartIdx];
1125		− pCurDqLayer->pRefIndex[listIdx][iMbXy][iScan4Idx] = pCurDqLayer->pRefIndex[listIdx][iMbXy][iScan4Idx + 1] =
1126		− pCurDqLayer->pRefIndex[listIdx][iMbXy][iScan4Idx + 4] = pCurDqLayer->pRefIndex[listIdx][iMbXy][iScan4Idx + 5] = iRef;
	1170	+ pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][iScan4Idx] = pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][iScan4Idx + 1] =
	1171	+ pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][iScan4Idx + 4] = pCurDqLayer->pDec->pRefIndex[listIdx][iMbXy][iScan4Idx +
	1172	+ 5] = iRef;
1127	1173
1128	1174	}
1129	1175	} // namespace WelsDec

re-structure SWelsDecoderContext and others and prepare for the support of multi-threading support.

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code