Context navigation

TComRdCost.cpp @ 473

Visit:

Last change on this file since 473 was 446, checked in by tech, 12 years ago
Added missing parts.
Property svn:eol-style set to `native`
File size: 74.6 KB

Line
1	/* The copyright in this software is being made available under the BSD
2	* License, included below. This software may be subject to other third party
3	* and contributor rights, including patent rights, and no such rights are
4	* granted under this license.
5	*
6	* Copyright (c) 2010-2013, ITU/ISO/IEC
7	* All rights reserved.
8	*
9	* Redistribution and use in source and binary forms, with or without
10	* modification, are permitted provided that the following conditions are met:
11	*
12	* * Redistributions of source code must retain the above copyright notice,
13	* this list of conditions and the following disclaimer.
14	* * Redistributions in binary form must reproduce the above copyright notice,
15	* this list of conditions and the following disclaimer in the documentation
16	* and/or other materials provided with the distribution.
17	* * Neither the name of the ITU/ISO/IEC nor the names of its contributors may
18	* be used to endorse or promote products derived from this software without
19	* specific prior written permission.
20	*
21	* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
22	* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24	* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
25	* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26	* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27	* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
28	* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29	* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30	* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
31	* THE POSSIBILITY OF SUCH DAMAGE.
32	*/
33
34	/** \file TComRdCost.cpp
35	\brief RD cost computation class
36	*/
37
38	#include <math.h>
39	#include <assert.h>
40	#include "TComRom.h"
41	#include "TComRdCost.h"
42	#if H_3D
43	#include "TComDataCU.h"
44	#endif
45
46	//! \ingroup TLibCommon
47	//! \{
48
49	#if H_3D_VSO
50	// SAIT_VSO_EST_A0033
51	Double TComRdCost::m_dDisparityCoeff = 1.0;
52	#endif
53
54	TComRdCost::TComRdCost()
55	{
56	init();
57	}
58
59	TComRdCost::~TComRdCost()
60	{
61	#if !FIX203
62	xUninit();
63	#endif
64	}
65
66	// Calculate RD functions
67	#if H_3D_VSO
68	Double TComRdCost::calcRdCost( UInt uiBits, Dist uiDistortion, Bool bFlag, DFunc eDFunc )
69	#else
70	Double TComRdCost::calcRdCost( UInt uiBits, UInt uiDistortion, Bool bFlag, DFunc eDFunc )
71	#endif
72	{
73	Double dRdCost = 0.0;
74	Double dLambda = 0.0;
75
76	switch ( eDFunc )
77	{
78	case DF_SSE:
79	assert(0);
80	break;
81	case DF_SAD:
82	dLambda = (Double)m_uiLambdaMotionSAD;
83	break;
84	case DF_DEFAULT:
85	dLambda = m_dLambda;
86	break;
87	case DF_SSE_FRAME:
88	dLambda = m_dFrameLambda;
89	break;
90	default:
91	assert (0);
92	break;
93	}
94
95	if (bFlag)
96	{
97	// Intra8x8, Intra4x4 Block only...
98	#if SEQUENCE_LEVEL_LOSSLESS
99	dRdCost = (Double)(uiBits);
100	#else
101	dRdCost = (((Double)uiDistortion) + ((Double)uiBits * dLambda));
102	#endif
103	}
104	else
105	{
106	if (eDFunc == DF_SAD)
107	{
108	dRdCost = ((Double)uiDistortion + (Double)((Int)(uiBits * dLambda+.5)>>16));
109	dRdCost = (Double)(UInt)floor(dRdCost);
110	}
111	else
112	{
113	#if SEQUENCE_LEVEL_LOSSLESS
114	dRdCost = (Double)(uiBits);
115	#else
116	dRdCost = ((Double)uiDistortion + (Double)((Int)(uiBits * dLambda+.5)));
117	dRdCost = (Double)(UInt)floor(dRdCost);
118	#endif
119	}
120	}
121
122	return dRdCost;
123	}
124
125	#if H_3D_VSO
126	Double TComRdCost::calcRdCost64( UInt64 uiBits, Dist64 uiDistortion, Bool bFlag, DFunc eDFunc )
127	#else
128	Double TComRdCost::calcRdCost64( UInt64 uiBits, UInt64 uiDistortion, Bool bFlag, DFunc eDFunc )
129	#endif
130	{
131	Double dRdCost = 0.0;
132	Double dLambda = 0.0;
133
134	switch ( eDFunc )
135	{
136	case DF_SSE:
137	assert(0);
138	break;
139	case DF_SAD:
140	dLambda = (Double)m_uiLambdaMotionSAD;
141	break;
142	case DF_DEFAULT:
143	dLambda = m_dLambda;
144	break;
145	case DF_SSE_FRAME:
146	dLambda = m_dFrameLambda;
147	break;
148	default:
149	assert (0);
150	break;
151	}
152
153	if (bFlag)
154	{
155	// Intra8x8, Intra4x4 Block only...
156	#if SEQUENCE_LEVEL_LOSSLESS
157	dRdCost = (Double)(uiBits);
158	#else
159	dRdCost = (((Double)(Int64)uiDistortion) + ((Double)(Int64)uiBits * dLambda));
160	#endif
161	}
162	else
163	{
164	if (eDFunc == DF_SAD)
165	{
166	dRdCost = ((Double)(Int64)uiDistortion + (Double)((Int)((Int64)uiBits * dLambda+.5)>>16));
167	dRdCost = (Double)(UInt)floor(dRdCost);
168	}
169	else
170	{
171	#if SEQUENCE_LEVEL_LOSSLESS
172	dRdCost = (Double)(uiBits);
173	#else
174	dRdCost = ((Double)(Int64)uiDistortion + (Double)((Int)((Int64)uiBits * dLambda+.5)));
175	dRdCost = (Double)(UInt)floor(dRdCost);
176	#endif
177	}
178	}
179
180	return dRdCost;
181	}
182
183	Void TComRdCost::setLambda( Double dLambda )
184	{
185	m_dLambda = dLambda;
186	m_sqrtLambda = sqrt(m_dLambda);
187	m_uiLambdaMotionSAD = (UInt)floor(65536.0 * m_sqrtLambda);
188	m_uiLambdaMotionSSE = (UInt)floor(65536.0 * m_dLambda );
189	}
190
191
192	// Initalize Function Pointer by [eDFunc]
193	Void TComRdCost::init()
194	{
195	m_afpDistortFunc[0] = NULL; // for DF_DEFAULT
196
197	m_afpDistortFunc[1] = TComRdCost::xGetSSE;
198	m_afpDistortFunc[2] = TComRdCost::xGetSSE4;
199	m_afpDistortFunc[3] = TComRdCost::xGetSSE8;
200	m_afpDistortFunc[4] = TComRdCost::xGetSSE16;
201	m_afpDistortFunc[5] = TComRdCost::xGetSSE32;
202	m_afpDistortFunc[6] = TComRdCost::xGetSSE64;
203	m_afpDistortFunc[7] = TComRdCost::xGetSSE16N;
204
205	m_afpDistortFunc[8] = TComRdCost::xGetSAD;
206	m_afpDistortFunc[9] = TComRdCost::xGetSAD4;
207	m_afpDistortFunc[10] = TComRdCost::xGetSAD8;
208	m_afpDistortFunc[11] = TComRdCost::xGetSAD16;
209	m_afpDistortFunc[12] = TComRdCost::xGetSAD32;
210	m_afpDistortFunc[13] = TComRdCost::xGetSAD64;
211	m_afpDistortFunc[14] = TComRdCost::xGetSAD16N;
212
213	m_afpDistortFunc[15] = TComRdCost::xGetSAD;
214	m_afpDistortFunc[16] = TComRdCost::xGetSAD4;
215	m_afpDistortFunc[17] = TComRdCost::xGetSAD8;
216	m_afpDistortFunc[18] = TComRdCost::xGetSAD16;
217	m_afpDistortFunc[19] = TComRdCost::xGetSAD32;
218	m_afpDistortFunc[20] = TComRdCost::xGetSAD64;
219	m_afpDistortFunc[21] = TComRdCost::xGetSAD16N;
220
221	#if AMP_SAD
222	m_afpDistortFunc[43] = TComRdCost::xGetSAD12;
223	m_afpDistortFunc[44] = TComRdCost::xGetSAD24;
224	m_afpDistortFunc[45] = TComRdCost::xGetSAD48;
225
226	m_afpDistortFunc[46] = TComRdCost::xGetSAD12;
227	m_afpDistortFunc[47] = TComRdCost::xGetSAD24;
228	m_afpDistortFunc[48] = TComRdCost::xGetSAD48;
229	#endif
230	m_afpDistortFunc[22] = TComRdCost::xGetHADs;
231	m_afpDistortFunc[23] = TComRdCost::xGetHADs;
232	m_afpDistortFunc[24] = TComRdCost::xGetHADs;
233	m_afpDistortFunc[25] = TComRdCost::xGetHADs;
234	m_afpDistortFunc[26] = TComRdCost::xGetHADs;
235	m_afpDistortFunc[27] = TComRdCost::xGetHADs;
236	m_afpDistortFunc[28] = TComRdCost::xGetHADs;
237
238	#if H_3D_VSO
239	// SAIT_VSO_EST_A0033
240	m_afpDistortFunc[29] = TComRdCost::xGetVSD;
241	m_afpDistortFunc[30] = TComRdCost::xGetVSD4;
242	m_afpDistortFunc[31] = TComRdCost::xGetVSD8;
243	m_afpDistortFunc[32] = TComRdCost::xGetVSD16;
244	m_afpDistortFunc[33] = TComRdCost::xGetVSD32;
245	m_afpDistortFunc[34] = TComRdCost::xGetVSD64;
246	m_afpDistortFunc[35] = TComRdCost::xGetVSD16N;
247	#endif
248	#if !FIX203
249	m_puiComponentCostOriginP = NULL;
250	m_puiComponentCost = NULL;
251	m_puiVerCost = NULL;
252	m_puiHorCost = NULL;
253	#endif
254	m_uiCost = 0;
255	m_iCostScale = 0;
256	#if !FIX203
257	m_iSearchLimit = 0xdeaddead;
258	#endif
259	#if H_3D_VSO
260	m_bUseVSO = false;
261	m_uiVSOMode = 0;
262	m_fpDistortFuncVSO = NULL;
263	m_pcRenModel = NULL;
264
265	// SAIT_VSO_EST_A0033
266	m_bUseEstimatedVSD = false;
267	#endif
268	}
269
270	#if !FIX203
271	Void TComRdCost::initRateDistortionModel( Int iSubPelSearchLimit )
272	{
273	// make it larger
274	iSubPelSearchLimit += 4;
275	iSubPelSearchLimit *= 8;
276
277	if( m_iSearchLimit != iSubPelSearchLimit )
278	{
279	xUninit();
280
281	m_iSearchLimit = iSubPelSearchLimit;
282
283	m_puiComponentCostOriginP = new UInt[ 4 * iSubPelSearchLimit ];
284	iSubPelSearchLimit *= 2;
285
286	m_puiComponentCost = m_puiComponentCostOriginP + iSubPelSearchLimit;
287
288	for( Int n = -iSubPelSearchLimit; n < iSubPelSearchLimit; n++)
289	{
290	m_puiComponentCost[n] = xGetComponentBits( n );
291	}
292	}
293	}
294
295	Void TComRdCost::xUninit()
296	{
297	if( NULL != m_puiComponentCostOriginP )
298	{
299	delete [] m_puiComponentCostOriginP;
300	m_puiComponentCostOriginP = NULL;
301	}
302	}
303	#endif
304
305	UInt TComRdCost::xGetComponentBits( Int iVal )
306	{
307	UInt uiLength = 1;
308	UInt uiTemp = ( iVal <= 0) ? (-iVal<<1)+1: (iVal<<1);
309
310	assert ( uiTemp );
311
312	while ( 1 != uiTemp )
313	{
314	uiTemp >>= 1;
315	uiLength += 2;
316	}
317
318	return uiLength;
319	}
320
321	Void TComRdCost::setDistParam( UInt uiBlkWidth, UInt uiBlkHeight, DFunc eDFunc, DistParam& rcDistParam )
322	{
323	// set Block Width / Height
324	rcDistParam.iCols = uiBlkWidth;
325	rcDistParam.iRows = uiBlkHeight;
326	rcDistParam.DistFunc = m_afpDistortFunc[eDFunc + g_aucConvertToBit[ rcDistParam.iCols ] + 1 ];
327
328	// initialize
329	rcDistParam.iSubShift = 0;
330	}
331
332	// Setting the Distortion Parameter for Inter (ME)
333	Void TComRdCost::setDistParam( TComPattern* pcPatternKey, Pel* piRefY, Int iRefStride, DistParam& rcDistParam )
334	{
335	// set Original & Curr Pointer / Stride
336	rcDistParam.pOrg = pcPatternKey->getROIY();
337	rcDistParam.pCur = piRefY;
338
339	rcDistParam.iStrideOrg = pcPatternKey->getPatternLStride();
340	rcDistParam.iStrideCur = iRefStride;
341
342	// set Block Width / Height
343	rcDistParam.iCols = pcPatternKey->getROIYWidth();
344	rcDistParam.iRows = pcPatternKey->getROIYHeight();
345	rcDistParam.DistFunc = m_afpDistortFunc[DF_SAD + g_aucConvertToBit[ rcDistParam.iCols ] + 1 ];
346
347	#if AMP_SAD
348	if (rcDistParam.iCols == 12)
349	{
350	rcDistParam.DistFunc = m_afpDistortFunc[43 ];
351	}
352	else if (rcDistParam.iCols == 24)
353	{
354	rcDistParam.DistFunc = m_afpDistortFunc[44 ];
355	}
356	else if (rcDistParam.iCols == 48)
357	{
358	rcDistParam.DistFunc = m_afpDistortFunc[45 ];
359	}
360	#endif
361
362	// initialize
363	rcDistParam.iSubShift = 0;
364	}
365
366	// Setting the Distortion Parameter for Inter (subpel ME with step)
367	#if NS_HAD
368	Void TComRdCost::setDistParam( TComPattern* pcPatternKey, Pel* piRefY, Int iRefStride, Int iStep, DistParam& rcDistParam, Bool bHADME, Bool bUseNSHAD )
369	#else
370	Void TComRdCost::setDistParam( TComPattern* pcPatternKey, Pel* piRefY, Int iRefStride, Int iStep, DistParam& rcDistParam, Bool bHADME )
371	#endif
372	{
373	// set Original & Curr Pointer / Stride
374	rcDistParam.pOrg = pcPatternKey->getROIY();
375	rcDistParam.pCur = piRefY;
376
377	rcDistParam.iStrideOrg = pcPatternKey->getPatternLStride();
378	rcDistParam.iStrideCur = iRefStride * iStep;
379
380	// set Step for interpolated buffer
381	rcDistParam.iStep = iStep;
382
383	// set Block Width / Height
384	rcDistParam.iCols = pcPatternKey->getROIYWidth();
385	rcDistParam.iRows = pcPatternKey->getROIYHeight();
386	#if NS_HAD
387	rcDistParam.bUseNSHAD = bUseNSHAD;
388	#endif
389
390	// set distortion function
391	if ( !bHADME )
392	{
393	rcDistParam.DistFunc = m_afpDistortFunc[DF_SADS + g_aucConvertToBit[ rcDistParam.iCols ] + 1 ];
394	#if AMP_SAD
395	if (rcDistParam.iCols == 12)
396	{
397	rcDistParam.DistFunc = m_afpDistortFunc[46 ];
398	}
399	else if (rcDistParam.iCols == 24)
400	{
401	rcDistParam.DistFunc = m_afpDistortFunc[47 ];
402	}
403	else if (rcDistParam.iCols == 48)
404	{
405	rcDistParam.DistFunc = m_afpDistortFunc[48 ];
406	}
407	#endif
408	}
409	else
410	{
411	rcDistParam.DistFunc = m_afpDistortFunc[DF_HADS + g_aucConvertToBit[ rcDistParam.iCols ] + 1 ];
412	}
413
414	// initialize
415	rcDistParam.iSubShift = 0;
416	}
417
418	Void
419	#if NS_HAD
420	TComRdCost::setDistParam( DistParam& rcDP, Pel* p1, Int iStride1, Pel* p2, Int iStride2, Int iWidth, Int iHeight, Bool bHadamard, Bool bUseNSHAD )
421	#else
422	TComRdCost::setDistParam( DistParam& rcDP, Int bitDepth, Pel* p1, Int iStride1, Pel* p2, Int iStride2, Int iWidth, Int iHeight, Bool bHadamard )
423	#endif
424	{
425	rcDP.pOrg = p1;
426	rcDP.pCur = p2;
427	rcDP.iStrideOrg = iStride1;
428	rcDP.iStrideCur = iStride2;
429	rcDP.iCols = iWidth;
430	rcDP.iRows = iHeight;
431	rcDP.iStep = 1;
432	rcDP.iSubShift = 0;
433	rcDP.bitDepth = bitDepth;
434	rcDP.DistFunc = m_afpDistortFunc[ ( bHadamard ? DF_HADS : DF_SADS ) + g_aucConvertToBit[ iWidth ] + 1 ];
435	#if NS_HAD
436	rcDP.bUseNSHAD = bUseNSHAD;
437	#endif
438	}
439
440	UInt TComRdCost::calcHAD(Int bitDepth, Pel* pi0, Int iStride0, Pel* pi1, Int iStride1, Int iWidth, Int iHeight )
441	{
442	UInt uiSum = 0;
443	Int x, y;
444
445	if ( ( (iWidth % 8) == 0 ) && ( (iHeight % 8) == 0 ) )
446	{
447	for ( y=0; y<iHeight; y+= 8 )
448	{
449	for ( x=0; x<iWidth; x+= 8 )
450	{
451	uiSum += xCalcHADs8x8( &pi0[x], &pi1[x], iStride0, iStride1, 1 );
452	}
453	pi0 += iStride0*8;
454	pi1 += iStride1*8;
455	}
456	}
457	else if ( ( (iWidth % 4) == 0 ) && ( (iHeight % 4) == 0 ) )
458	{
459	for ( y=0; y<iHeight; y+= 4 )
460	{
461	for ( x=0; x<iWidth; x+= 4 )
462	{
463	uiSum += xCalcHADs4x4( &pi0[x], &pi1[x], iStride0, iStride1, 1 );
464	}
465	pi0 += iStride0*4;
466	pi1 += iStride1*4;
467	}
468	}
469	else
470	{
471	for ( y=0; y<iHeight; y+= 2 )
472	{
473	for ( x=0; x<iWidth; x+= 2 )
474	{
475	uiSum += xCalcHADs8x8( &pi0[x], &pi1[x], iStride0, iStride1, 1 );
476	}
477	pi0 += iStride0*2;
478	pi1 += iStride1*2;
479	}
480	}
481
482	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(bitDepth-8);
483
484	}
485
486	#if WEIGHTED_CHROMA_DISTORTION
487	UInt TComRdCost::getDistPart(Int bitDepth, Pel* piCur, Int iCurStride, Pel* piOrg, Int iOrgStride, UInt uiBlkWidth, UInt uiBlkHeight, TextType eText, DFunc eDFunc)
488	#else
489	UInt TComRdCost::getDistPart(Int bitDepth, Pel* piCur, Int iCurStride, Pel* piOrg, Int iOrgStride, UInt uiBlkWidth, UInt uiBlkHeight, DFunc eDFunc )
490	#endif
491	{
492	DistParam cDtParam;
493	setDistParam( uiBlkWidth, uiBlkHeight, eDFunc, cDtParam );
494	cDtParam.pOrg = piOrg;
495	cDtParam.pCur = piCur;
496	cDtParam.iStrideOrg = iOrgStride;
497	cDtParam.iStrideCur = iCurStride;
498	cDtParam.iStep = 1;
499
500	cDtParam.bApplyWeight = false;
501	cDtParam.uiComp = 255; // just for assert: to be sure it was set before use, since only values 0,1 or 2 are allowed.
502	cDtParam.bitDepth = bitDepth;
503
504	#if WEIGHTED_CHROMA_DISTORTION
505	if (eText == TEXT_CHROMA_U)
506	{
507	return ((Int) (m_cbDistortionWeight * cDtParam.DistFunc( &cDtParam )));
508	}
509	else if (eText == TEXT_CHROMA_V)
510	{
511	return ((Int) (m_crDistortionWeight * cDtParam.DistFunc( &cDtParam )));
512	}
513	else
514	{
515	return cDtParam.DistFunc( &cDtParam );
516	}
517	#else
518	return cDtParam.DistFunc( &cDtParam );
519	#endif
520	}
521	#if H_3D_VSO
522	// SAIT_VSO_EST_A0033
523	UInt TComRdCost::getDistPartVSD( TComDataCU* pcCU, UInt uiPartOffset, Pel* piCur, Int iCurStride, Pel* piOrg, Int iOrgStride, UInt uiBlkWidth, UInt uiBlkHeight, Bool bHAD, DFunc eDFunc )
524	{
525	AOT( ( m_dDisparityCoeff <= 0 ) \|\| ( m_dDisparityCoeff > 10 ) );
526
527	Pel* piVirRec = m_pcVideoRecPicYuv->getLumaAddr(pcCU->getAddr(),pcCU->getZorderIdxInCU()+uiPartOffset);
528	Pel* piVirOrg = m_pcDepthPicYuv ->getLumaAddr(pcCU->getAddr(),pcCU->getZorderIdxInCU()+uiPartOffset);
529	Int iVirStride = m_pcVideoRecPicYuv->getStride();
530
531	DistParam cDtParam;
532	setDistParam( uiBlkWidth, uiBlkHeight, eDFunc, cDtParam );
533	cDtParam.pOrg = piOrg;
534	cDtParam.pCur = piCur;
535	cDtParam.pVirRec = piVirRec;
536	cDtParam.pVirOrg = piVirOrg;
537	cDtParam.iStrideVir = iVirStride;
538	cDtParam.iStrideOrg = iOrgStride;
539	cDtParam.iStrideCur = iCurStride;
540	cDtParam.iStep = 1;
541
542	cDtParam.bApplyWeight = false;
543	cDtParam.uiComp = 255; // just for assert: to be sure it was set before use, since only values 0,1 or 2 are allowed.
544
545	Dist dist = cDtParam.DistFunc( &cDtParam );
546
547	if ( m_bUseWVSO )
548	{
549	Int iDWeight = m_iDWeight * m_iDWeight;
550	Int iVSOWeight = m_iVSDWeight * m_iVSDWeight;
551	Dist distDepth;
552
553	if ( !bHAD )
554	{
555	distDepth = (Dist) getDistPart( g_bitDepthY, piCur, iCurStride, piOrg, iOrgStride, uiBlkWidth, uiBlkHeight);
556	}
557	else
558	{
559	distDepth = (Dist) calcHAD( g_bitDepthY, piCur, iCurStride, piOrg, iOrgStride, uiBlkWidth, uiBlkHeight);
560	}
561
562	dist = (Dist) (iDWeight * distDepth + iVSOWeight * dist ) / ( iDWeight + iVSOWeight);
563	}
564	return dist;
565
566	}
567	#endif
568
569	#if RATE_CONTROL_LAMBDA_DOMAIN
570	UInt TComRdCost::getSADPart ( Int bitDepth, Pel* pelCur, Int curStride, Pel* pelOrg, Int orgStride, UInt width, UInt height )
571	{
572	UInt SAD = 0;
573	Int shift = DISTORTION_PRECISION_ADJUSTMENT(bitDepth-8);
574	for ( Int i=0; i<height; i++ )
575	{
576	for( Int j=0; j<width; j++ )
577	{
578	SAD += abs((pelCur[j] - pelOrg[j])) >> shift;
579	}
580	pelCur = pelCur + curStride;
581	pelOrg = pelOrg + orgStride;
582	}
583	return SAD;
584	}
585	#endif
586
587	// ====================================================================================================================
588	// Distortion functions
589	// ====================================================================================================================
590
591	// --------------------------------------------------------------------------------------------------------------------
592	// SAD
593	// --------------------------------------------------------------------------------------------------------------------
594
595	UInt TComRdCost::xGetSAD( DistParam* pcDtParam )
596	{
597	if ( pcDtParam->bApplyWeight )
598	{
599	return xGetSADw( pcDtParam );
600	}
601	Pel* piOrg = pcDtParam->pOrg;
602	Pel* piCur = pcDtParam->pCur;
603	Int iRows = pcDtParam->iRows;
604	Int iCols = pcDtParam->iCols;
605	Int iStrideCur = pcDtParam->iStrideCur;
606	Int iStrideOrg = pcDtParam->iStrideOrg;
607
608	UInt uiSum = 0;
609
610	for( ; iRows != 0; iRows-- )
611	{
612	for (Int n = 0; n < iCols; n++ )
613	{
614	uiSum += abs( piOrg[n] - piCur[n] );
615	}
616	piOrg += iStrideOrg;
617	piCur += iStrideCur;
618	}
619
620	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
621	}
622
623	UInt TComRdCost::xGetSAD4( DistParam* pcDtParam )
624	{
625	if ( pcDtParam->bApplyWeight )
626	{
627	return xGetSADw( pcDtParam );
628	}
629	Pel* piOrg = pcDtParam->pOrg;
630	Pel* piCur = pcDtParam->pCur;
631	Int iRows = pcDtParam->iRows;
632	Int iSubShift = pcDtParam->iSubShift;
633	Int iSubStep = ( 1 << iSubShift );
634	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
635	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
636
637	UInt uiSum = 0;
638
639	for( ; iRows != 0; iRows-=iSubStep )
640	{
641	uiSum += abs( piOrg[0] - piCur[0] );
642	uiSum += abs( piOrg[1] - piCur[1] );
643	uiSum += abs( piOrg[2] - piCur[2] );
644	uiSum += abs( piOrg[3] - piCur[3] );
645
646	piOrg += iStrideOrg;
647	piCur += iStrideCur;
648	}
649
650	uiSum <<= iSubShift;
651	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
652	}
653
654	UInt TComRdCost::xGetSAD8( DistParam* pcDtParam )
655	{
656	if ( pcDtParam->bApplyWeight )
657	{
658	return xGetSADw( pcDtParam );
659	}
660	Pel* piOrg = pcDtParam->pOrg;
661	Pel* piCur = pcDtParam->pCur;
662	Int iRows = pcDtParam->iRows;
663	Int iSubShift = pcDtParam->iSubShift;
664	Int iSubStep = ( 1 << iSubShift );
665	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
666	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
667
668	UInt uiSum = 0;
669
670	for( ; iRows != 0; iRows-=iSubStep )
671	{
672	uiSum += abs( piOrg[0] - piCur[0] );
673	uiSum += abs( piOrg[1] - piCur[1] );
674	uiSum += abs( piOrg[2] - piCur[2] );
675	uiSum += abs( piOrg[3] - piCur[3] );
676	uiSum += abs( piOrg[4] - piCur[4] );
677	uiSum += abs( piOrg[5] - piCur[5] );
678	uiSum += abs( piOrg[6] - piCur[6] );
679	uiSum += abs( piOrg[7] - piCur[7] );
680
681	piOrg += iStrideOrg;
682	piCur += iStrideCur;
683	}
684
685	uiSum <<= iSubShift;
686	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
687	}
688
689	UInt TComRdCost::xGetSAD16( DistParam* pcDtParam )
690	{
691	if ( pcDtParam->bApplyWeight )
692	{
693	return xGetSADw( pcDtParam );
694	}
695	Pel* piOrg = pcDtParam->pOrg;
696	Pel* piCur = pcDtParam->pCur;
697	Int iRows = pcDtParam->iRows;
698	Int iSubShift = pcDtParam->iSubShift;
699	Int iSubStep = ( 1 << iSubShift );
700	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
701	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
702
703	UInt uiSum = 0;
704
705	for( ; iRows != 0; iRows-=iSubStep )
706	{
707	uiSum += abs( piOrg[0] - piCur[0] );
708	uiSum += abs( piOrg[1] - piCur[1] );
709	uiSum += abs( piOrg[2] - piCur[2] );
710	uiSum += abs( piOrg[3] - piCur[3] );
711	uiSum += abs( piOrg[4] - piCur[4] );
712	uiSum += abs( piOrg[5] - piCur[5] );
713	uiSum += abs( piOrg[6] - piCur[6] );
714	uiSum += abs( piOrg[7] - piCur[7] );
715	uiSum += abs( piOrg[8] - piCur[8] );
716	uiSum += abs( piOrg[9] - piCur[9] );
717	uiSum += abs( piOrg[10] - piCur[10] );
718	uiSum += abs( piOrg[11] - piCur[11] );
719	uiSum += abs( piOrg[12] - piCur[12] );
720	uiSum += abs( piOrg[13] - piCur[13] );
721	uiSum += abs( piOrg[14] - piCur[14] );
722	uiSum += abs( piOrg[15] - piCur[15] );
723
724	piOrg += iStrideOrg;
725	piCur += iStrideCur;
726	}
727
728	uiSum <<= iSubShift;
729	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
730	}
731
732	#if AMP_SAD
733	UInt TComRdCost::xGetSAD12( DistParam* pcDtParam )
734	{
735	if ( pcDtParam->bApplyWeight )
736	{
737	return xGetSADw( pcDtParam );
738	}
739	Pel* piOrg = pcDtParam->pOrg;
740	Pel* piCur = pcDtParam->pCur;
741	Int iRows = pcDtParam->iRows;
742	Int iSubShift = pcDtParam->iSubShift;
743	Int iSubStep = ( 1 << iSubShift );
744	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
745	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
746
747	UInt uiSum = 0;
748
749	for( ; iRows != 0; iRows-=iSubStep )
750	{
751	uiSum += abs( piOrg[0] - piCur[0] );
752	uiSum += abs( piOrg[1] - piCur[1] );
753	uiSum += abs( piOrg[2] - piCur[2] );
754	uiSum += abs( piOrg[3] - piCur[3] );
755	uiSum += abs( piOrg[4] - piCur[4] );
756	uiSum += abs( piOrg[5] - piCur[5] );
757	uiSum += abs( piOrg[6] - piCur[6] );
758	uiSum += abs( piOrg[7] - piCur[7] );
759	uiSum += abs( piOrg[8] - piCur[8] );
760	uiSum += abs( piOrg[9] - piCur[9] );
761	uiSum += abs( piOrg[10] - piCur[10] );
762	uiSum += abs( piOrg[11] - piCur[11] );
763
764	piOrg += iStrideOrg;
765	piCur += iStrideCur;
766	}
767
768	uiSum <<= iSubShift;
769	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
770	}
771	#endif
772
773	UInt TComRdCost::xGetSAD16N( DistParam* pcDtParam )
774	{
775	Pel* piOrg = pcDtParam->pOrg;
776	Pel* piCur = pcDtParam->pCur;
777	Int iRows = pcDtParam->iRows;
778	Int iCols = pcDtParam->iCols;
779	Int iSubShift = pcDtParam->iSubShift;
780	Int iSubStep = ( 1 << iSubShift );
781	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
782	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
783
784	UInt uiSum = 0;
785
786	for( ; iRows != 0; iRows-=iSubStep )
787	{
788	for (Int n = 0; n < iCols; n+=16 )
789	{
790	uiSum += abs( piOrg[n+ 0] - piCur[n+ 0] );
791	uiSum += abs( piOrg[n+ 1] - piCur[n+ 1] );
792	uiSum += abs( piOrg[n+ 2] - piCur[n+ 2] );
793	uiSum += abs( piOrg[n+ 3] - piCur[n+ 3] );
794	uiSum += abs( piOrg[n+ 4] - piCur[n+ 4] );
795	uiSum += abs( piOrg[n+ 5] - piCur[n+ 5] );
796	uiSum += abs( piOrg[n+ 6] - piCur[n+ 6] );
797	uiSum += abs( piOrg[n+ 7] - piCur[n+ 7] );
798	uiSum += abs( piOrg[n+ 8] - piCur[n+ 8] );
799	uiSum += abs( piOrg[n+ 9] - piCur[n+ 9] );
800	uiSum += abs( piOrg[n+10] - piCur[n+10] );
801	uiSum += abs( piOrg[n+11] - piCur[n+11] );
802	uiSum += abs( piOrg[n+12] - piCur[n+12] );
803	uiSum += abs( piOrg[n+13] - piCur[n+13] );
804	uiSum += abs( piOrg[n+14] - piCur[n+14] );
805	uiSum += abs( piOrg[n+15] - piCur[n+15] );
806	}
807	piOrg += iStrideOrg;
808	piCur += iStrideCur;
809	}
810
811	uiSum <<= iSubShift;
812	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
813	}
814
815	UInt TComRdCost::xGetSAD32( DistParam* pcDtParam )
816	{
817	if ( pcDtParam->bApplyWeight )
818	{
819	return xGetSADw( pcDtParam );
820	}
821	Pel* piOrg = pcDtParam->pOrg;
822	Pel* piCur = pcDtParam->pCur;
823	Int iRows = pcDtParam->iRows;
824	Int iSubShift = pcDtParam->iSubShift;
825	Int iSubStep = ( 1 << iSubShift );
826	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
827	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
828
829	UInt uiSum = 0;
830
831	for( ; iRows != 0; iRows-=iSubStep )
832	{
833	uiSum += abs( piOrg[0] - piCur[0] );
834	uiSum += abs( piOrg[1] - piCur[1] );
835	uiSum += abs( piOrg[2] - piCur[2] );
836	uiSum += abs( piOrg[3] - piCur[3] );
837	uiSum += abs( piOrg[4] - piCur[4] );
838	uiSum += abs( piOrg[5] - piCur[5] );
839	uiSum += abs( piOrg[6] - piCur[6] );
840	uiSum += abs( piOrg[7] - piCur[7] );
841	uiSum += abs( piOrg[8] - piCur[8] );
842	uiSum += abs( piOrg[9] - piCur[9] );
843	uiSum += abs( piOrg[10] - piCur[10] );
844	uiSum += abs( piOrg[11] - piCur[11] );
845	uiSum += abs( piOrg[12] - piCur[12] );
846	uiSum += abs( piOrg[13] - piCur[13] );
847	uiSum += abs( piOrg[14] - piCur[14] );
848	uiSum += abs( piOrg[15] - piCur[15] );
849	uiSum += abs( piOrg[16] - piCur[16] );
850	uiSum += abs( piOrg[17] - piCur[17] );
851	uiSum += abs( piOrg[18] - piCur[18] );
852	uiSum += abs( piOrg[19] - piCur[19] );
853	uiSum += abs( piOrg[20] - piCur[20] );
854	uiSum += abs( piOrg[21] - piCur[21] );
855	uiSum += abs( piOrg[22] - piCur[22] );
856	uiSum += abs( piOrg[23] - piCur[23] );
857	uiSum += abs( piOrg[24] - piCur[24] );
858	uiSum += abs( piOrg[25] - piCur[25] );
859	uiSum += abs( piOrg[26] - piCur[26] );
860	uiSum += abs( piOrg[27] - piCur[27] );
861	uiSum += abs( piOrg[28] - piCur[28] );
862	uiSum += abs( piOrg[29] - piCur[29] );
863	uiSum += abs( piOrg[30] - piCur[30] );
864	uiSum += abs( piOrg[31] - piCur[31] );
865
866	piOrg += iStrideOrg;
867	piCur += iStrideCur;
868	}
869
870	uiSum <<= iSubShift;
871	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
872	}
873
874	#if AMP_SAD
875	UInt TComRdCost::xGetSAD24( DistParam* pcDtParam )
876	{
877	if ( pcDtParam->bApplyWeight )
878	{
879	return xGetSADw( pcDtParam );
880	}
881	Pel* piOrg = pcDtParam->pOrg;
882	Pel* piCur = pcDtParam->pCur;
883	Int iRows = pcDtParam->iRows;
884	Int iSubShift = pcDtParam->iSubShift;
885	Int iSubStep = ( 1 << iSubShift );
886	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
887	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
888
889	UInt uiSum = 0;
890
891	for( ; iRows != 0; iRows-=iSubStep )
892	{
893	uiSum += abs( piOrg[0] - piCur[0] );
894	uiSum += abs( piOrg[1] - piCur[1] );
895	uiSum += abs( piOrg[2] - piCur[2] );
896	uiSum += abs( piOrg[3] - piCur[3] );
897	uiSum += abs( piOrg[4] - piCur[4] );
898	uiSum += abs( piOrg[5] - piCur[5] );
899	uiSum += abs( piOrg[6] - piCur[6] );
900	uiSum += abs( piOrg[7] - piCur[7] );
901	uiSum += abs( piOrg[8] - piCur[8] );
902	uiSum += abs( piOrg[9] - piCur[9] );
903	uiSum += abs( piOrg[10] - piCur[10] );
904	uiSum += abs( piOrg[11] - piCur[11] );
905	uiSum += abs( piOrg[12] - piCur[12] );
906	uiSum += abs( piOrg[13] - piCur[13] );
907	uiSum += abs( piOrg[14] - piCur[14] );
908	uiSum += abs( piOrg[15] - piCur[15] );
909	uiSum += abs( piOrg[16] - piCur[16] );
910	uiSum += abs( piOrg[17] - piCur[17] );
911	uiSum += abs( piOrg[18] - piCur[18] );
912	uiSum += abs( piOrg[19] - piCur[19] );
913	uiSum += abs( piOrg[20] - piCur[20] );
914	uiSum += abs( piOrg[21] - piCur[21] );
915	uiSum += abs( piOrg[22] - piCur[22] );
916	uiSum += abs( piOrg[23] - piCur[23] );
917
918	piOrg += iStrideOrg;
919	piCur += iStrideCur;
920	}
921
922	uiSum <<= iSubShift;
923	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
924	}
925
926	#endif
927
928	UInt TComRdCost::xGetSAD64( DistParam* pcDtParam )
929	{
930	if ( pcDtParam->bApplyWeight )
931	{
932	return xGetSADw( pcDtParam );
933	}
934	Pel* piOrg = pcDtParam->pOrg;
935	Pel* piCur = pcDtParam->pCur;
936	Int iRows = pcDtParam->iRows;
937	Int iSubShift = pcDtParam->iSubShift;
938	Int iSubStep = ( 1 << iSubShift );
939	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
940	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
941
942	UInt uiSum = 0;
943
944	for( ; iRows != 0; iRows-=iSubStep )
945	{
946	uiSum += abs( piOrg[0] - piCur[0] );
947	uiSum += abs( piOrg[1] - piCur[1] );
948	uiSum += abs( piOrg[2] - piCur[2] );
949	uiSum += abs( piOrg[3] - piCur[3] );
950	uiSum += abs( piOrg[4] - piCur[4] );
951	uiSum += abs( piOrg[5] - piCur[5] );
952	uiSum += abs( piOrg[6] - piCur[6] );
953	uiSum += abs( piOrg[7] - piCur[7] );
954	uiSum += abs( piOrg[8] - piCur[8] );
955	uiSum += abs( piOrg[9] - piCur[9] );
956	uiSum += abs( piOrg[10] - piCur[10] );
957	uiSum += abs( piOrg[11] - piCur[11] );
958	uiSum += abs( piOrg[12] - piCur[12] );
959	uiSum += abs( piOrg[13] - piCur[13] );
960	uiSum += abs( piOrg[14] - piCur[14] );
961	uiSum += abs( piOrg[15] - piCur[15] );
962	uiSum += abs( piOrg[16] - piCur[16] );
963	uiSum += abs( piOrg[17] - piCur[17] );
964	uiSum += abs( piOrg[18] - piCur[18] );
965	uiSum += abs( piOrg[19] - piCur[19] );
966	uiSum += abs( piOrg[20] - piCur[20] );
967	uiSum += abs( piOrg[21] - piCur[21] );
968	uiSum += abs( piOrg[22] - piCur[22] );
969	uiSum += abs( piOrg[23] - piCur[23] );
970	uiSum += abs( piOrg[24] - piCur[24] );
971	uiSum += abs( piOrg[25] - piCur[25] );
972	uiSum += abs( piOrg[26] - piCur[26] );
973	uiSum += abs( piOrg[27] - piCur[27] );
974	uiSum += abs( piOrg[28] - piCur[28] );
975	uiSum += abs( piOrg[29] - piCur[29] );
976	uiSum += abs( piOrg[30] - piCur[30] );
977	uiSum += abs( piOrg[31] - piCur[31] );
978	uiSum += abs( piOrg[32] - piCur[32] );
979	uiSum += abs( piOrg[33] - piCur[33] );
980	uiSum += abs( piOrg[34] - piCur[34] );
981	uiSum += abs( piOrg[35] - piCur[35] );
982	uiSum += abs( piOrg[36] - piCur[36] );
983	uiSum += abs( piOrg[37] - piCur[37] );
984	uiSum += abs( piOrg[38] - piCur[38] );
985	uiSum += abs( piOrg[39] - piCur[39] );
986	uiSum += abs( piOrg[40] - piCur[40] );
987	uiSum += abs( piOrg[41] - piCur[41] );
988	uiSum += abs( piOrg[42] - piCur[42] );
989	uiSum += abs( piOrg[43] - piCur[43] );
990	uiSum += abs( piOrg[44] - piCur[44] );
991	uiSum += abs( piOrg[45] - piCur[45] );
992	uiSum += abs( piOrg[46] - piCur[46] );
993	uiSum += abs( piOrg[47] - piCur[47] );
994	uiSum += abs( piOrg[48] - piCur[48] );
995	uiSum += abs( piOrg[49] - piCur[49] );
996	uiSum += abs( piOrg[50] - piCur[50] );
997	uiSum += abs( piOrg[51] - piCur[51] );
998	uiSum += abs( piOrg[52] - piCur[52] );
999	uiSum += abs( piOrg[53] - piCur[53] );
1000	uiSum += abs( piOrg[54] - piCur[54] );
1001	uiSum += abs( piOrg[55] - piCur[55] );
1002	uiSum += abs( piOrg[56] - piCur[56] );
1003	uiSum += abs( piOrg[57] - piCur[57] );
1004	uiSum += abs( piOrg[58] - piCur[58] );
1005	uiSum += abs( piOrg[59] - piCur[59] );
1006	uiSum += abs( piOrg[60] - piCur[60] );
1007	uiSum += abs( piOrg[61] - piCur[61] );
1008	uiSum += abs( piOrg[62] - piCur[62] );
1009	uiSum += abs( piOrg[63] - piCur[63] );
1010
1011	piOrg += iStrideOrg;
1012	piCur += iStrideCur;
1013	}
1014
1015	uiSum <<= iSubShift;
1016	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
1017	}
1018
1019	#if AMP_SAD
1020	UInt TComRdCost::xGetSAD48( DistParam* pcDtParam )
1021	{
1022	if ( pcDtParam->bApplyWeight )
1023	{
1024	return xGetSADw( pcDtParam );
1025	}
1026	Pel* piOrg = pcDtParam->pOrg;
1027	Pel* piCur = pcDtParam->pCur;
1028	Int iRows = pcDtParam->iRows;
1029	Int iSubShift = pcDtParam->iSubShift;
1030	Int iSubStep = ( 1 << iSubShift );
1031	Int iStrideCur = pcDtParam->iStrideCur*iSubStep;
1032	Int iStrideOrg = pcDtParam->iStrideOrg*iSubStep;
1033
1034	UInt uiSum = 0;
1035
1036	for( ; iRows != 0; iRows-=iSubStep )
1037	{
1038	uiSum += abs( piOrg[0] - piCur[0] );
1039	uiSum += abs( piOrg[1] - piCur[1] );
1040	uiSum += abs( piOrg[2] - piCur[2] );
1041	uiSum += abs( piOrg[3] - piCur[3] );
1042	uiSum += abs( piOrg[4] - piCur[4] );
1043	uiSum += abs( piOrg[5] - piCur[5] );
1044	uiSum += abs( piOrg[6] - piCur[6] );
1045	uiSum += abs( piOrg[7] - piCur[7] );
1046	uiSum += abs( piOrg[8] - piCur[8] );
1047	uiSum += abs( piOrg[9] - piCur[9] );
1048	uiSum += abs( piOrg[10] - piCur[10] );
1049	uiSum += abs( piOrg[11] - piCur[11] );
1050	uiSum += abs( piOrg[12] - piCur[12] );
1051	uiSum += abs( piOrg[13] - piCur[13] );
1052	uiSum += abs( piOrg[14] - piCur[14] );
1053	uiSum += abs( piOrg[15] - piCur[15] );
1054	uiSum += abs( piOrg[16] - piCur[16] );
1055	uiSum += abs( piOrg[17] - piCur[17] );
1056	uiSum += abs( piOrg[18] - piCur[18] );
1057	uiSum += abs( piOrg[19] - piCur[19] );
1058	uiSum += abs( piOrg[20] - piCur[20] );
1059	uiSum += abs( piOrg[21] - piCur[21] );
1060	uiSum += abs( piOrg[22] - piCur[22] );
1061	uiSum += abs( piOrg[23] - piCur[23] );
1062	uiSum += abs( piOrg[24] - piCur[24] );
1063	uiSum += abs( piOrg[25] - piCur[25] );
1064	uiSum += abs( piOrg[26] - piCur[26] );
1065	uiSum += abs( piOrg[27] - piCur[27] );
1066	uiSum += abs( piOrg[28] - piCur[28] );
1067	uiSum += abs( piOrg[29] - piCur[29] );
1068	uiSum += abs( piOrg[30] - piCur[30] );
1069	uiSum += abs( piOrg[31] - piCur[31] );
1070	uiSum += abs( piOrg[32] - piCur[32] );
1071	uiSum += abs( piOrg[33] - piCur[33] );
1072	uiSum += abs( piOrg[34] - piCur[34] );
1073	uiSum += abs( piOrg[35] - piCur[35] );
1074	uiSum += abs( piOrg[36] - piCur[36] );
1075	uiSum += abs( piOrg[37] - piCur[37] );
1076	uiSum += abs( piOrg[38] - piCur[38] );
1077	uiSum += abs( piOrg[39] - piCur[39] );
1078	uiSum += abs( piOrg[40] - piCur[40] );
1079	uiSum += abs( piOrg[41] - piCur[41] );
1080	uiSum += abs( piOrg[42] - piCur[42] );
1081	uiSum += abs( piOrg[43] - piCur[43] );
1082	uiSum += abs( piOrg[44] - piCur[44] );
1083	uiSum += abs( piOrg[45] - piCur[45] );
1084	uiSum += abs( piOrg[46] - piCur[46] );
1085	uiSum += abs( piOrg[47] - piCur[47] );
1086
1087	piOrg += iStrideOrg;
1088	piCur += iStrideCur;
1089	}
1090
1091	uiSum <<= iSubShift;
1092	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
1093	}
1094	#endif
1095
1096	// --------------------------------------------------------------------------------------------------------------------
1097	// SSE
1098	// --------------------------------------------------------------------------------------------------------------------
1099
1100	UInt TComRdCost::xGetSSE( DistParam* pcDtParam )
1101	{
1102	if ( pcDtParam->bApplyWeight )
1103	{
1104	return xGetSSEw( pcDtParam );
1105	}
1106	Pel* piOrg = pcDtParam->pOrg;
1107	Pel* piCur = pcDtParam->pCur;
1108	Int iRows = pcDtParam->iRows;
1109	Int iCols = pcDtParam->iCols;
1110	Int iStrideOrg = pcDtParam->iStrideOrg;
1111	Int iStrideCur = pcDtParam->iStrideCur;
1112
1113	UInt uiSum = 0;
1114	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT((pcDtParam->bitDepth-8) << 1);
1115
1116	Int iTemp;
1117
1118	for( ; iRows != 0; iRows-- )
1119	{
1120	for (Int n = 0; n < iCols; n++ )
1121	{
1122	iTemp = piOrg[n ] - piCur[n ];
1123	uiSum += ( iTemp * iTemp ) >> uiShift;
1124	}
1125	piOrg += iStrideOrg;
1126	piCur += iStrideCur;
1127	}
1128
1129	return ( uiSum );
1130	}
1131
1132	UInt TComRdCost::xGetSSE4( DistParam* pcDtParam )
1133	{
1134	if ( pcDtParam->bApplyWeight )
1135	{
1136	assert( pcDtParam->iCols == 4 );
1137	return xGetSSEw( pcDtParam );
1138	}
1139	Pel* piOrg = pcDtParam->pOrg;
1140	Pel* piCur = pcDtParam->pCur;
1141	Int iRows = pcDtParam->iRows;
1142	Int iStrideOrg = pcDtParam->iStrideOrg;
1143	Int iStrideCur = pcDtParam->iStrideCur;
1144
1145	UInt uiSum = 0;
1146	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT((pcDtParam->bitDepth-8) << 1);
1147
1148	Int iTemp;
1149
1150	for( ; iRows != 0; iRows-- )
1151	{
1152
1153	iTemp = piOrg[0] - piCur[0]; uiSum += ( iTemp * iTemp ) >> uiShift;
1154	iTemp = piOrg[1] - piCur[1]; uiSum += ( iTemp * iTemp ) >> uiShift;
1155	iTemp = piOrg[2] - piCur[2]; uiSum += ( iTemp * iTemp ) >> uiShift;
1156	iTemp = piOrg[3] - piCur[3]; uiSum += ( iTemp * iTemp ) >> uiShift;
1157
1158	piOrg += iStrideOrg;
1159	piCur += iStrideCur;
1160	}
1161
1162	return ( uiSum );
1163	}
1164
1165	UInt TComRdCost::xGetSSE8( DistParam* pcDtParam )
1166	{
1167	if ( pcDtParam->bApplyWeight )
1168	{
1169	assert( pcDtParam->iCols == 8 );
1170	return xGetSSEw( pcDtParam );
1171	}
1172	Pel* piOrg = pcDtParam->pOrg;
1173	Pel* piCur = pcDtParam->pCur;
1174	Int iRows = pcDtParam->iRows;
1175	Int iStrideOrg = pcDtParam->iStrideOrg;
1176	Int iStrideCur = pcDtParam->iStrideCur;
1177
1178	UInt uiSum = 0;
1179	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT((pcDtParam->bitDepth-8) << 1);
1180
1181	Int iTemp;
1182
1183	for( ; iRows != 0; iRows-- )
1184	{
1185	iTemp = piOrg[0] - piCur[0]; uiSum += ( iTemp * iTemp ) >> uiShift;
1186	iTemp = piOrg[1] - piCur[1]; uiSum += ( iTemp * iTemp ) >> uiShift;
1187	iTemp = piOrg[2] - piCur[2]; uiSum += ( iTemp * iTemp ) >> uiShift;
1188	iTemp = piOrg[3] - piCur[3]; uiSum += ( iTemp * iTemp ) >> uiShift;
1189	iTemp = piOrg[4] - piCur[4]; uiSum += ( iTemp * iTemp ) >> uiShift;
1190	iTemp = piOrg[5] - piCur[5]; uiSum += ( iTemp * iTemp ) >> uiShift;
1191	iTemp = piOrg[6] - piCur[6]; uiSum += ( iTemp * iTemp ) >> uiShift;
1192	iTemp = piOrg[7] - piCur[7]; uiSum += ( iTemp * iTemp ) >> uiShift;
1193
1194	piOrg += iStrideOrg;
1195	piCur += iStrideCur;
1196	}
1197
1198	return ( uiSum );
1199	}
1200
1201	UInt TComRdCost::xGetSSE16( DistParam* pcDtParam )
1202	{
1203	if ( pcDtParam->bApplyWeight )
1204	{
1205	assert( pcDtParam->iCols == 16 );
1206	return xGetSSEw( pcDtParam );
1207	}
1208	Pel* piOrg = pcDtParam->pOrg;
1209	Pel* piCur = pcDtParam->pCur;
1210	Int iRows = pcDtParam->iRows;
1211	Int iStrideOrg = pcDtParam->iStrideOrg;
1212	Int iStrideCur = pcDtParam->iStrideCur;
1213
1214	UInt uiSum = 0;
1215	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT((pcDtParam->bitDepth-8) << 1);
1216
1217	Int iTemp;
1218
1219	for( ; iRows != 0; iRows-- )
1220	{
1221
1222	iTemp = piOrg[ 0] - piCur[ 0]; uiSum += ( iTemp * iTemp ) >> uiShift;
1223	iTemp = piOrg[ 1] - piCur[ 1]; uiSum += ( iTemp * iTemp ) >> uiShift;
1224	iTemp = piOrg[ 2] - piCur[ 2]; uiSum += ( iTemp * iTemp ) >> uiShift;
1225	iTemp = piOrg[ 3] - piCur[ 3]; uiSum += ( iTemp * iTemp ) >> uiShift;
1226	iTemp = piOrg[ 4] - piCur[ 4]; uiSum += ( iTemp * iTemp ) >> uiShift;
1227	iTemp = piOrg[ 5] - piCur[ 5]; uiSum += ( iTemp * iTemp ) >> uiShift;
1228	iTemp = piOrg[ 6] - piCur[ 6]; uiSum += ( iTemp * iTemp ) >> uiShift;
1229	iTemp = piOrg[ 7] - piCur[ 7]; uiSum += ( iTemp * iTemp ) >> uiShift;
1230	iTemp = piOrg[ 8] - piCur[ 8]; uiSum += ( iTemp * iTemp ) >> uiShift;
1231	iTemp = piOrg[ 9] - piCur[ 9]; uiSum += ( iTemp * iTemp ) >> uiShift;
1232	iTemp = piOrg[10] - piCur[10]; uiSum += ( iTemp * iTemp ) >> uiShift;
1233	iTemp = piOrg[11] - piCur[11]; uiSum += ( iTemp * iTemp ) >> uiShift;
1234	iTemp = piOrg[12] - piCur[12]; uiSum += ( iTemp * iTemp ) >> uiShift;
1235	iTemp = piOrg[13] - piCur[13]; uiSum += ( iTemp * iTemp ) >> uiShift;
1236	iTemp = piOrg[14] - piCur[14]; uiSum += ( iTemp * iTemp ) >> uiShift;
1237	iTemp = piOrg[15] - piCur[15]; uiSum += ( iTemp * iTemp ) >> uiShift;
1238
1239	piOrg += iStrideOrg;
1240	piCur += iStrideCur;
1241	}
1242
1243	return ( uiSum );
1244	}
1245
1246	UInt TComRdCost::xGetSSE16N( DistParam* pcDtParam )
1247	{
1248	if ( pcDtParam->bApplyWeight )
1249	{
1250	return xGetSSEw( pcDtParam );
1251	}
1252	Pel* piOrg = pcDtParam->pOrg;
1253	Pel* piCur = pcDtParam->pCur;
1254	Int iRows = pcDtParam->iRows;
1255	Int iCols = pcDtParam->iCols;
1256	Int iStrideOrg = pcDtParam->iStrideOrg;
1257	Int iStrideCur = pcDtParam->iStrideCur;
1258
1259	UInt uiSum = 0;
1260	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT((pcDtParam->bitDepth-8) << 1);
1261	Int iTemp;
1262
1263	for( ; iRows != 0; iRows-- )
1264	{
1265	for (Int n = 0; n < iCols; n+=16 )
1266	{
1267
1268	iTemp = piOrg[n+ 0] - piCur[n+ 0]; uiSum += ( iTemp * iTemp ) >> uiShift;
1269	iTemp = piOrg[n+ 1] - piCur[n+ 1]; uiSum += ( iTemp * iTemp ) >> uiShift;
1270	iTemp = piOrg[n+ 2] - piCur[n+ 2]; uiSum += ( iTemp * iTemp ) >> uiShift;
1271	iTemp = piOrg[n+ 3] - piCur[n+ 3]; uiSum += ( iTemp * iTemp ) >> uiShift;
1272	iTemp = piOrg[n+ 4] - piCur[n+ 4]; uiSum += ( iTemp * iTemp ) >> uiShift;
1273	iTemp = piOrg[n+ 5] - piCur[n+ 5]; uiSum += ( iTemp * iTemp ) >> uiShift;
1274	iTemp = piOrg[n+ 6] - piCur[n+ 6]; uiSum += ( iTemp * iTemp ) >> uiShift;
1275	iTemp = piOrg[n+ 7] - piCur[n+ 7]; uiSum += ( iTemp * iTemp ) >> uiShift;
1276	iTemp = piOrg[n+ 8] - piCur[n+ 8]; uiSum += ( iTemp * iTemp ) >> uiShift;
1277	iTemp = piOrg[n+ 9] - piCur[n+ 9]; uiSum += ( iTemp * iTemp ) >> uiShift;
1278	iTemp = piOrg[n+10] - piCur[n+10]; uiSum += ( iTemp * iTemp ) >> uiShift;
1279	iTemp = piOrg[n+11] - piCur[n+11]; uiSum += ( iTemp * iTemp ) >> uiShift;
1280	iTemp = piOrg[n+12] - piCur[n+12]; uiSum += ( iTemp * iTemp ) >> uiShift;
1281	iTemp = piOrg[n+13] - piCur[n+13]; uiSum += ( iTemp * iTemp ) >> uiShift;
1282	iTemp = piOrg[n+14] - piCur[n+14]; uiSum += ( iTemp * iTemp ) >> uiShift;
1283	iTemp = piOrg[n+15] - piCur[n+15]; uiSum += ( iTemp * iTemp ) >> uiShift;
1284
1285	}
1286	piOrg += iStrideOrg;
1287	piCur += iStrideCur;
1288	}
1289
1290	return ( uiSum );
1291	}
1292
1293	UInt TComRdCost::xGetSSE32( DistParam* pcDtParam )
1294	{
1295	if ( pcDtParam->bApplyWeight )
1296	{
1297	assert( pcDtParam->iCols == 32 );
1298	return xGetSSEw( pcDtParam );
1299	}
1300	Pel* piOrg = pcDtParam->pOrg;
1301	Pel* piCur = pcDtParam->pCur;
1302	Int iRows = pcDtParam->iRows;
1303	Int iStrideOrg = pcDtParam->iStrideOrg;
1304	Int iStrideCur = pcDtParam->iStrideCur;
1305
1306	UInt uiSum = 0;
1307	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT((pcDtParam->bitDepth-8) << 1);
1308	Int iTemp;
1309
1310	for( ; iRows != 0; iRows-- )
1311	{
1312
1313	iTemp = piOrg[ 0] - piCur[ 0]; uiSum += ( iTemp * iTemp ) >> uiShift;
1314	iTemp = piOrg[ 1] - piCur[ 1]; uiSum += ( iTemp * iTemp ) >> uiShift;
1315	iTemp = piOrg[ 2] - piCur[ 2]; uiSum += ( iTemp * iTemp ) >> uiShift;
1316	iTemp = piOrg[ 3] - piCur[ 3]; uiSum += ( iTemp * iTemp ) >> uiShift;
1317	iTemp = piOrg[ 4] - piCur[ 4]; uiSum += ( iTemp * iTemp ) >> uiShift;
1318	iTemp = piOrg[ 5] - piCur[ 5]; uiSum += ( iTemp * iTemp ) >> uiShift;
1319	iTemp = piOrg[ 6] - piCur[ 6]; uiSum += ( iTemp * iTemp ) >> uiShift;
1320	iTemp = piOrg[ 7] - piCur[ 7]; uiSum += ( iTemp * iTemp ) >> uiShift;
1321	iTemp = piOrg[ 8] - piCur[ 8]; uiSum += ( iTemp * iTemp ) >> uiShift;
1322	iTemp = piOrg[ 9] - piCur[ 9]; uiSum += ( iTemp * iTemp ) >> uiShift;
1323	iTemp = piOrg[10] - piCur[10]; uiSum += ( iTemp * iTemp ) >> uiShift;
1324	iTemp = piOrg[11] - piCur[11]; uiSum += ( iTemp * iTemp ) >> uiShift;
1325	iTemp = piOrg[12] - piCur[12]; uiSum += ( iTemp * iTemp ) >> uiShift;
1326	iTemp = piOrg[13] - piCur[13]; uiSum += ( iTemp * iTemp ) >> uiShift;
1327	iTemp = piOrg[14] - piCur[14]; uiSum += ( iTemp * iTemp ) >> uiShift;
1328	iTemp = piOrg[15] - piCur[15]; uiSum += ( iTemp * iTemp ) >> uiShift;
1329	iTemp = piOrg[16] - piCur[16]; uiSum += ( iTemp * iTemp ) >> uiShift;
1330	iTemp = piOrg[17] - piCur[17]; uiSum += ( iTemp * iTemp ) >> uiShift;
1331	iTemp = piOrg[18] - piCur[18]; uiSum += ( iTemp * iTemp ) >> uiShift;
1332	iTemp = piOrg[19] - piCur[19]; uiSum += ( iTemp * iTemp ) >> uiShift;
1333	iTemp = piOrg[20] - piCur[20]; uiSum += ( iTemp * iTemp ) >> uiShift;
1334	iTemp = piOrg[21] - piCur[21]; uiSum += ( iTemp * iTemp ) >> uiShift;
1335	iTemp = piOrg[22] - piCur[22]; uiSum += ( iTemp * iTemp ) >> uiShift;
1336	iTemp = piOrg[23] - piCur[23]; uiSum += ( iTemp * iTemp ) >> uiShift;
1337	iTemp = piOrg[24] - piCur[24]; uiSum += ( iTemp * iTemp ) >> uiShift;
1338	iTemp = piOrg[25] - piCur[25]; uiSum += ( iTemp * iTemp ) >> uiShift;
1339	iTemp = piOrg[26] - piCur[26]; uiSum += ( iTemp * iTemp ) >> uiShift;
1340	iTemp = piOrg[27] - piCur[27]; uiSum += ( iTemp * iTemp ) >> uiShift;
1341	iTemp = piOrg[28] - piCur[28]; uiSum += ( iTemp * iTemp ) >> uiShift;
1342	iTemp = piOrg[29] - piCur[29]; uiSum += ( iTemp * iTemp ) >> uiShift;
1343	iTemp = piOrg[30] - piCur[30]; uiSum += ( iTemp * iTemp ) >> uiShift;
1344	iTemp = piOrg[31] - piCur[31]; uiSum += ( iTemp * iTemp ) >> uiShift;
1345
1346	piOrg += iStrideOrg;
1347	piCur += iStrideCur;
1348	}
1349
1350	return ( uiSum );
1351	}
1352
1353	UInt TComRdCost::xGetSSE64( DistParam* pcDtParam )
1354	{
1355	if ( pcDtParam->bApplyWeight )
1356	{
1357	assert( pcDtParam->iCols == 64 );
1358	return xGetSSEw( pcDtParam );
1359	}
1360	Pel* piOrg = pcDtParam->pOrg;
1361	Pel* piCur = pcDtParam->pCur;
1362	Int iRows = pcDtParam->iRows;
1363	Int iStrideOrg = pcDtParam->iStrideOrg;
1364	Int iStrideCur = pcDtParam->iStrideCur;
1365
1366	UInt uiSum = 0;
1367	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT((pcDtParam->bitDepth-8) << 1);
1368	Int iTemp;
1369
1370	for( ; iRows != 0; iRows-- )
1371	{
1372	iTemp = piOrg[ 0] - piCur[ 0]; uiSum += ( iTemp * iTemp ) >> uiShift;
1373	iTemp = piOrg[ 1] - piCur[ 1]; uiSum += ( iTemp * iTemp ) >> uiShift;
1374	iTemp = piOrg[ 2] - piCur[ 2]; uiSum += ( iTemp * iTemp ) >> uiShift;
1375	iTemp = piOrg[ 3] - piCur[ 3]; uiSum += ( iTemp * iTemp ) >> uiShift;
1376	iTemp = piOrg[ 4] - piCur[ 4]; uiSum += ( iTemp * iTemp ) >> uiShift;
1377	iTemp = piOrg[ 5] - piCur[ 5]; uiSum += ( iTemp * iTemp ) >> uiShift;
1378	iTemp = piOrg[ 6] - piCur[ 6]; uiSum += ( iTemp * iTemp ) >> uiShift;
1379	iTemp = piOrg[ 7] - piCur[ 7]; uiSum += ( iTemp * iTemp ) >> uiShift;
1380	iTemp = piOrg[ 8] - piCur[ 8]; uiSum += ( iTemp * iTemp ) >> uiShift;
1381	iTemp = piOrg[ 9] - piCur[ 9]; uiSum += ( iTemp * iTemp ) >> uiShift;
1382	iTemp = piOrg[10] - piCur[10]; uiSum += ( iTemp * iTemp ) >> uiShift;
1383	iTemp = piOrg[11] - piCur[11]; uiSum += ( iTemp * iTemp ) >> uiShift;
1384	iTemp = piOrg[12] - piCur[12]; uiSum += ( iTemp * iTemp ) >> uiShift;
1385	iTemp = piOrg[13] - piCur[13]; uiSum += ( iTemp * iTemp ) >> uiShift;
1386	iTemp = piOrg[14] - piCur[14]; uiSum += ( iTemp * iTemp ) >> uiShift;
1387	iTemp = piOrg[15] - piCur[15]; uiSum += ( iTemp * iTemp ) >> uiShift;
1388	iTemp = piOrg[16] - piCur[16]; uiSum += ( iTemp * iTemp ) >> uiShift;
1389	iTemp = piOrg[17] - piCur[17]; uiSum += ( iTemp * iTemp ) >> uiShift;
1390	iTemp = piOrg[18] - piCur[18]; uiSum += ( iTemp * iTemp ) >> uiShift;
1391	iTemp = piOrg[19] - piCur[19]; uiSum += ( iTemp * iTemp ) >> uiShift;
1392	iTemp = piOrg[20] - piCur[20]; uiSum += ( iTemp * iTemp ) >> uiShift;
1393	iTemp = piOrg[21] - piCur[21]; uiSum += ( iTemp * iTemp ) >> uiShift;
1394	iTemp = piOrg[22] - piCur[22]; uiSum += ( iTemp * iTemp ) >> uiShift;
1395	iTemp = piOrg[23] - piCur[23]; uiSum += ( iTemp * iTemp ) >> uiShift;
1396	iTemp = piOrg[24] - piCur[24]; uiSum += ( iTemp * iTemp ) >> uiShift;
1397	iTemp = piOrg[25] - piCur[25]; uiSum += ( iTemp * iTemp ) >> uiShift;
1398	iTemp = piOrg[26] - piCur[26]; uiSum += ( iTemp * iTemp ) >> uiShift;
1399	iTemp = piOrg[27] - piCur[27]; uiSum += ( iTemp * iTemp ) >> uiShift;
1400	iTemp = piOrg[28] - piCur[28]; uiSum += ( iTemp * iTemp ) >> uiShift;
1401	iTemp = piOrg[29] - piCur[29]; uiSum += ( iTemp * iTemp ) >> uiShift;
1402	iTemp = piOrg[30] - piCur[30]; uiSum += ( iTemp * iTemp ) >> uiShift;
1403	iTemp = piOrg[31] - piCur[31]; uiSum += ( iTemp * iTemp ) >> uiShift;
1404	iTemp = piOrg[32] - piCur[32]; uiSum += ( iTemp * iTemp ) >> uiShift;
1405	iTemp = piOrg[33] - piCur[33]; uiSum += ( iTemp * iTemp ) >> uiShift;
1406	iTemp = piOrg[34] - piCur[34]; uiSum += ( iTemp * iTemp ) >> uiShift;
1407	iTemp = piOrg[35] - piCur[35]; uiSum += ( iTemp * iTemp ) >> uiShift;
1408	iTemp = piOrg[36] - piCur[36]; uiSum += ( iTemp * iTemp ) >> uiShift;
1409	iTemp = piOrg[37] - piCur[37]; uiSum += ( iTemp * iTemp ) >> uiShift;
1410	iTemp = piOrg[38] - piCur[38]; uiSum += ( iTemp * iTemp ) >> uiShift;
1411	iTemp = piOrg[39] - piCur[39]; uiSum += ( iTemp * iTemp ) >> uiShift;
1412	iTemp = piOrg[40] - piCur[40]; uiSum += ( iTemp * iTemp ) >> uiShift;
1413	iTemp = piOrg[41] - piCur[41]; uiSum += ( iTemp * iTemp ) >> uiShift;
1414	iTemp = piOrg[42] - piCur[42]; uiSum += ( iTemp * iTemp ) >> uiShift;
1415	iTemp = piOrg[43] - piCur[43]; uiSum += ( iTemp * iTemp ) >> uiShift;
1416	iTemp = piOrg[44] - piCur[44]; uiSum += ( iTemp * iTemp ) >> uiShift;
1417	iTemp = piOrg[45] - piCur[45]; uiSum += ( iTemp * iTemp ) >> uiShift;
1418	iTemp = piOrg[46] - piCur[46]; uiSum += ( iTemp * iTemp ) >> uiShift;
1419	iTemp = piOrg[47] - piCur[47]; uiSum += ( iTemp * iTemp ) >> uiShift;
1420	iTemp = piOrg[48] - piCur[48]; uiSum += ( iTemp * iTemp ) >> uiShift;
1421	iTemp = piOrg[49] - piCur[49]; uiSum += ( iTemp * iTemp ) >> uiShift;
1422	iTemp = piOrg[50] - piCur[50]; uiSum += ( iTemp * iTemp ) >> uiShift;
1423	iTemp = piOrg[51] - piCur[51]; uiSum += ( iTemp * iTemp ) >> uiShift;
1424	iTemp = piOrg[52] - piCur[52]; uiSum += ( iTemp * iTemp ) >> uiShift;
1425	iTemp = piOrg[53] - piCur[53]; uiSum += ( iTemp * iTemp ) >> uiShift;
1426	iTemp = piOrg[54] - piCur[54]; uiSum += ( iTemp * iTemp ) >> uiShift;
1427	iTemp = piOrg[55] - piCur[55]; uiSum += ( iTemp * iTemp ) >> uiShift;
1428	iTemp = piOrg[56] - piCur[56]; uiSum += ( iTemp * iTemp ) >> uiShift;
1429	iTemp = piOrg[57] - piCur[57]; uiSum += ( iTemp * iTemp ) >> uiShift;
1430	iTemp = piOrg[58] - piCur[58]; uiSum += ( iTemp * iTemp ) >> uiShift;
1431	iTemp = piOrg[59] - piCur[59]; uiSum += ( iTemp * iTemp ) >> uiShift;
1432	iTemp = piOrg[60] - piCur[60]; uiSum += ( iTemp * iTemp ) >> uiShift;
1433	iTemp = piOrg[61] - piCur[61]; uiSum += ( iTemp * iTemp ) >> uiShift;
1434	iTemp = piOrg[62] - piCur[62]; uiSum += ( iTemp * iTemp ) >> uiShift;
1435	iTemp = piOrg[63] - piCur[63]; uiSum += ( iTemp * iTemp ) >> uiShift;
1436
1437	piOrg += iStrideOrg;
1438	piCur += iStrideCur;
1439	}
1440
1441	return ( uiSum );
1442	}
1443	#if H_3D_VSO
1444	//SAIT_VSO_EST_A0033
1445	UInt TComRdCost::getVSDEstimate( Int dDM, Pel* pOrg, Int iOrgStride, Pel* pVirRec, Pel* pVirOrg, Int iVirStride, Int x, Int y )
1446	{
1447	Double dD;
1448	Int iTemp;
1449
1450	dD = ( (Double) ( dDM >> DISTORTION_PRECISION_ADJUSTMENT( g_bitDepthY - 8 ) ) ) * m_dDisparityCoeff;
1451
1452	Double dTemp = ( 0.5 * fabs(dD) * ( abs( (Int) pVirRec[ x+yiVirStride ] - (Int) pVirRec[ x-1+yiVirStride ] ) + abs( (Int) pVirRec[ x+yiVirStride ] - (Int) pVirRec[ x+1+yiVirStride ] ) ) );
1453	iTemp = (Int) (((dTemp) < 0)? (Int)((dTemp) - 0.5) : (Int)((dTemp) + 0.5));
1454
1455	return (UInt) ( (iTemp*iTemp)>>1 );
1456	}
1457
1458	UInt TComRdCost::xGetVSD( DistParam* pcDtParam )
1459	{
1460	Pel* piOrg = pcDtParam->pOrg;
1461	Pel* piCur = pcDtParam->pCur;
1462	Pel* piVirRec = pcDtParam->pVirRec;
1463	Pel* piVirOrg = pcDtParam->pVirOrg;
1464	Int iRows = pcDtParam->iRows;
1465	Int iCols = pcDtParam->iCols;
1466	Int iStrideOrg = pcDtParam->iStrideOrg;
1467	Int iStrideCur = pcDtParam->iStrideCur;
1468	Int iStrideVir = pcDtParam->iStrideVir;
1469
1470	UInt uiSum = 0;
1471	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8)<<1;
1472
1473	Int dDM;
1474
1475	for ( Int y = 0 ; y < iRows ; y++ )
1476	{
1477	for (Int x = 0; x < iCols; x++ )
1478	{
1479	dDM = (Int) ( piOrg[x ] - piCur[x ] );
1480	uiSum += getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, x, y ) >> uiShift;
1481	}
1482	piOrg += iStrideOrg;
1483	piCur += iStrideCur;
1484	}
1485
1486	return ( uiSum );
1487	}
1488
1489	UInt TComRdCost::xGetVSD4( DistParam* pcDtParam )
1490	{
1491	Pel* piOrg = pcDtParam->pOrg;
1492	Pel* piCur = pcDtParam->pCur;
1493	Pel* piVirRec = pcDtParam->pVirRec;
1494	Pel* piVirOrg = pcDtParam->pVirOrg;
1495	Int iRows = pcDtParam->iRows;
1496	Int iStrideOrg = pcDtParam->iStrideOrg;
1497	Int iStrideCur = pcDtParam->iStrideCur;
1498	Int iStrideVir = pcDtParam->iStrideVir;
1499
1500	UInt uiSum = 0;
1501	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8)<<1;
1502
1503	Int dDM;
1504
1505	for ( Int y = 0 ; y < iRows ; y++ )
1506	{
1507	dDM = (Int) ( piOrg[0] - piCur[0] ); uiSum += ( getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, 0, y ) ) >> uiShift;
1508	dDM = (Int) ( piOrg[1] - piCur[1] ); uiSum += ( getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, 1, y ) ) >> uiShift;
1509	dDM = (Int) ( piOrg[2] - piCur[2] ); uiSum += ( getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, 2, y ) ) >> uiShift;
1510	dDM = (Int) ( piOrg[3] - piCur[3] ); uiSum += ( getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, 3, y ) ) >> uiShift;
1511
1512	piOrg += iStrideOrg;
1513	piCur += iStrideCur;
1514	}
1515
1516	return ( uiSum );
1517	}
1518
1519	UInt TComRdCost::xGetVSD8( DistParam* pcDtParam )
1520	{
1521	Pel* piOrg = pcDtParam->pOrg;
1522	Pel* piCur = pcDtParam->pCur;
1523	Pel* piVirRec = pcDtParam->pVirRec;
1524	Pel* piVirOrg = pcDtParam->pVirOrg;
1525	Int iRows = pcDtParam->iRows;
1526	Int iStrideOrg = pcDtParam->iStrideOrg;
1527	Int iStrideCur = pcDtParam->iStrideCur;
1528	Int iStrideVir = pcDtParam->iStrideVir;
1529
1530	UInt uiSum = 0;
1531	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8)<<1;
1532
1533	Int dDM;
1534
1535	for ( Int y = 0 ; y < iRows ; y++ )
1536	{
1537	for (Int x = 0; x < 8; x++ )
1538	{
1539	dDM = (Int) ( piOrg[x] - piCur[x] );
1540	uiSum += getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, x, y ) >> uiShift;
1541	}
1542	piOrg += iStrideOrg;
1543	piCur += iStrideCur;
1544	}
1545
1546	return ( uiSum );
1547	}
1548
1549	UInt TComRdCost::xGetVSD16( DistParam* pcDtParam )
1550	{
1551	Pel* piOrg = pcDtParam->pOrg;
1552	Pel* piCur = pcDtParam->pCur;
1553	Pel* piVirRec = pcDtParam->pVirRec;
1554	Pel* piVirOrg = pcDtParam->pVirOrg;
1555	Int iRows = pcDtParam->iRows;
1556	Int iStrideOrg = pcDtParam->iStrideOrg;
1557	Int iStrideCur = pcDtParam->iStrideCur;
1558	Int iStrideVir = pcDtParam->iStrideVir;
1559
1560	UInt uiSum = 0;
1561	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8)<<1;
1562
1563	Int dDM;
1564
1565	for ( Int y = 0 ; y < iRows ; y++ )
1566	{
1567	for (Int x = 0; x < 16; x++ )
1568	{
1569	dDM = (Int) ( piOrg[x] - piCur[x] );
1570	uiSum += getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, x, y ) >> uiShift;
1571	}
1572	piOrg += iStrideOrg;
1573	piCur += iStrideCur;
1574	}
1575
1576	return ( uiSum );
1577	}
1578
1579	UInt TComRdCost::xGetVSD16N( DistParam* pcDtParam )
1580	{
1581	Pel* piOrg = pcDtParam->pOrg;
1582	Pel* piCur = pcDtParam->pCur;
1583	Pel* piVirRec = pcDtParam->pVirRec;
1584	Pel* piVirOrg = pcDtParam->pVirOrg;
1585	Int iRows = pcDtParam->iRows;
1586	Int iCols = pcDtParam->iCols;
1587	Int iStrideOrg = pcDtParam->iStrideOrg;
1588	Int iStrideCur = pcDtParam->iStrideCur;
1589	Int iStrideVir = pcDtParam->iStrideVir;
1590
1591	UInt uiSum = 0;
1592	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8)<<1;
1593
1594	Int dDM;
1595
1596	for ( Int y = 0 ; y < iRows ; y++ )
1597	{
1598	for (Int x = 0; x < iCols; x+=16 )
1599	{
1600	for ( Int k = 0 ; k < 16 ; k++ )
1601	{
1602	dDM = (Int) ( piOrg[x+k] - piCur[x+k] );
1603	uiSum += getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, x+k, y ) >> uiShift;
1604	}
1605	}
1606	piOrg += iStrideOrg;
1607	piCur += iStrideCur;
1608	}
1609
1610	return ( uiSum );
1611	}
1612
1613	UInt TComRdCost::xGetVSD32( DistParam* pcDtParam )
1614	{
1615	Pel* piOrg = pcDtParam->pOrg;
1616	Pel* piCur = pcDtParam->pCur;
1617	Pel* piVirRec = pcDtParam->pVirRec;
1618	Pel* piVirOrg = pcDtParam->pVirOrg;
1619	Int iRows = pcDtParam->iRows;
1620	Int iStrideOrg = pcDtParam->iStrideOrg;
1621	Int iStrideCur = pcDtParam->iStrideCur;
1622	Int iStrideVir = pcDtParam->iStrideVir;
1623
1624	UInt uiSum = 0;
1625	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8)<<1;
1626
1627	Int dDM;
1628
1629	for ( Int y = 0 ; y < iRows ; y++ )
1630	{
1631	for (Int x = 0; x < 32 ; x++ )
1632	{
1633	dDM = (Int) ( piOrg[x] - piCur[x] );
1634	uiSum += getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, x, y ) >> uiShift;
1635	}
1636	piOrg += iStrideOrg;
1637	piCur += iStrideCur;
1638	}
1639
1640	return ( uiSum );
1641	}
1642
1643	UInt TComRdCost::xGetVSD64( DistParam* pcDtParam )
1644	{
1645	Pel* piOrg = pcDtParam->pOrg;
1646	Pel* piCur = pcDtParam->pCur;
1647	Pel* piVirRec = pcDtParam->pVirRec;
1648	Pel* piVirOrg = pcDtParam->pVirOrg;
1649	Int iRows = pcDtParam->iRows;
1650	Int iStrideOrg = pcDtParam->iStrideOrg;
1651	Int iStrideCur = pcDtParam->iStrideCur;
1652	Int iStrideVir = pcDtParam->iStrideVir;
1653
1654	UInt uiSum = 0;
1655	UInt uiShift = DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8)<<1;
1656
1657	Int dDM;
1658
1659	for ( Int y = 0 ; y < iRows ; y++ )
1660	{
1661	for (Int x = 0; x < 64; x++ )
1662	{
1663	dDM = (Int) ( piOrg[x] - piCur[x] );
1664	uiSum += getVSDEstimate( dDM, piOrg, iStrideOrg, piVirRec, piVirOrg, iStrideVir, x, y ) >> uiShift;
1665	}
1666	piOrg += iStrideOrg;
1667	piCur += iStrideCur;
1668	}
1669
1670	return ( uiSum );
1671	}
1672
1673	#endif
1674
1675	// --------------------------------------------------------------------------------------------------------------------
1676	// HADAMARD with step (used in fractional search)
1677	// --------------------------------------------------------------------------------------------------------------------
1678
1679	UInt TComRdCost::xCalcHADs2x2( Pel piOrg, Pel piCur, Int iStrideOrg, Int iStrideCur, Int iStep )
1680	{
1681	Int satd = 0, diff[4], m[4];
1682	assert( iStep == 1 );
1683	diff[0] = piOrg[0 ] - piCur[0];
1684	diff[1] = piOrg[1 ] - piCur[1];
1685	diff[2] = piOrg[iStrideOrg ] - piCur[0 + iStrideCur];
1686	diff[3] = piOrg[iStrideOrg + 1] - piCur[1 + iStrideCur];
1687	m[0] = diff[0] + diff[2];
1688	m[1] = diff[1] + diff[3];
1689	m[2] = diff[0] - diff[2];
1690	m[3] = diff[1] - diff[3];
1691
1692	satd += abs(m[0] + m[1]);
1693	satd += abs(m[0] - m[1]);
1694	satd += abs(m[2] + m[3]);
1695	satd += abs(m[2] - m[3]);
1696
1697	return satd;
1698	}
1699
1700	UInt TComRdCost::xCalcHADs4x4( Pel piOrg, Pel piCur, Int iStrideOrg, Int iStrideCur, Int iStep )
1701	{
1702	Int k, satd = 0, diff[16], m[16], d[16];
1703
1704	assert( iStep == 1 );
1705	for( k = 0; k < 16; k+=4 )
1706	{
1707	diff[k+0] = piOrg[0] - piCur[0];
1708	diff[k+1] = piOrg[1] - piCur[1];
1709	diff[k+2] = piOrg[2] - piCur[2];
1710	diff[k+3] = piOrg[3] - piCur[3];
1711
1712	piCur += iStrideCur;
1713	piOrg += iStrideOrg;
1714	}
1715
1716	/===== hadamard transform =====/
1717	m[ 0] = diff[ 0] + diff[12];
1718	m[ 1] = diff[ 1] + diff[13];
1719	m[ 2] = diff[ 2] + diff[14];
1720	m[ 3] = diff[ 3] + diff[15];
1721	m[ 4] = diff[ 4] + diff[ 8];
1722	m[ 5] = diff[ 5] + diff[ 9];
1723	m[ 6] = diff[ 6] + diff[10];
1724	m[ 7] = diff[ 7] + diff[11];
1725	m[ 8] = diff[ 4] - diff[ 8];
1726	m[ 9] = diff[ 5] - diff[ 9];
1727	m[10] = diff[ 6] - diff[10];
1728	m[11] = diff[ 7] - diff[11];
1729	m[12] = diff[ 0] - diff[12];
1730	m[13] = diff[ 1] - diff[13];
1731	m[14] = diff[ 2] - diff[14];
1732	m[15] = diff[ 3] - diff[15];
1733
1734	d[ 0] = m[ 0] + m[ 4];
1735	d[ 1] = m[ 1] + m[ 5];
1736	d[ 2] = m[ 2] + m[ 6];
1737	d[ 3] = m[ 3] + m[ 7];
1738	d[ 4] = m[ 8] + m[12];
1739	d[ 5] = m[ 9] + m[13];
1740	d[ 6] = m[10] + m[14];
1741	d[ 7] = m[11] + m[15];
1742	d[ 8] = m[ 0] - m[ 4];
1743	d[ 9] = m[ 1] - m[ 5];
1744	d[10] = m[ 2] - m[ 6];
1745	d[11] = m[ 3] - m[ 7];
1746	d[12] = m[12] - m[ 8];
1747	d[13] = m[13] - m[ 9];
1748	d[14] = m[14] - m[10];
1749	d[15] = m[15] - m[11];
1750
1751	m[ 0] = d[ 0] + d[ 3];
1752	m[ 1] = d[ 1] + d[ 2];
1753	m[ 2] = d[ 1] - d[ 2];
1754	m[ 3] = d[ 0] - d[ 3];
1755	m[ 4] = d[ 4] + d[ 7];
1756	m[ 5] = d[ 5] + d[ 6];
1757	m[ 6] = d[ 5] - d[ 6];
1758	m[ 7] = d[ 4] - d[ 7];
1759	m[ 8] = d[ 8] + d[11];
1760	m[ 9] = d[ 9] + d[10];
1761	m[10] = d[ 9] - d[10];
1762	m[11] = d[ 8] - d[11];
1763	m[12] = d[12] + d[15];
1764	m[13] = d[13] + d[14];
1765	m[14] = d[13] - d[14];
1766	m[15] = d[12] - d[15];
1767
1768	d[ 0] = m[ 0] + m[ 1];
1769	d[ 1] = m[ 0] - m[ 1];
1770	d[ 2] = m[ 2] + m[ 3];
1771	d[ 3] = m[ 3] - m[ 2];
1772	d[ 4] = m[ 4] + m[ 5];
1773	d[ 5] = m[ 4] - m[ 5];
1774	d[ 6] = m[ 6] + m[ 7];
1775	d[ 7] = m[ 7] - m[ 6];
1776	d[ 8] = m[ 8] + m[ 9];
1777	d[ 9] = m[ 8] - m[ 9];
1778	d[10] = m[10] + m[11];
1779	d[11] = m[11] - m[10];
1780	d[12] = m[12] + m[13];
1781	d[13] = m[12] - m[13];
1782	d[14] = m[14] + m[15];
1783	d[15] = m[15] - m[14];
1784
1785	for (k=0; k<16; ++k)
1786	{
1787	satd += abs(d[k]);
1788	}
1789	satd = ((satd+1)>>1);
1790
1791	return satd;
1792	}
1793
1794	UInt TComRdCost::xCalcHADs8x8( Pel piOrg, Pel piCur, Int iStrideOrg, Int iStrideCur, Int iStep )
1795	{
1796	Int k, i, j, jj, sad=0;
1797	Int diff[64], m1[8][8], m2[8][8], m3[8][8];
1798	assert( iStep == 1 );
1799	for( k = 0; k < 64; k += 8 )
1800	{
1801	diff[k+0] = piOrg[0] - piCur[0];
1802	diff[k+1] = piOrg[1] - piCur[1];
1803	diff[k+2] = piOrg[2] - piCur[2];
1804	diff[k+3] = piOrg[3] - piCur[3];
1805	diff[k+4] = piOrg[4] - piCur[4];
1806	diff[k+5] = piOrg[5] - piCur[5];
1807	diff[k+6] = piOrg[6] - piCur[6];
1808	diff[k+7] = piOrg[7] - piCur[7];
1809
1810	piCur += iStrideCur;
1811	piOrg += iStrideOrg;
1812	}
1813
1814	//horizontal
1815	for (j=0; j < 8; j++)
1816	{
1817	jj = j << 3;
1818	m2[j][0] = diff[jj ] + diff[jj+4];
1819	m2[j][1] = diff[jj+1] + diff[jj+5];
1820	m2[j][2] = diff[jj+2] + diff[jj+6];
1821	m2[j][3] = diff[jj+3] + diff[jj+7];
1822	m2[j][4] = diff[jj ] - diff[jj+4];
1823	m2[j][5] = diff[jj+1] - diff[jj+5];
1824	m2[j][6] = diff[jj+2] - diff[jj+6];
1825	m2[j][7] = diff[jj+3] - diff[jj+7];
1826
1827	m1[j][0] = m2[j][0] + m2[j][2];
1828	m1[j][1] = m2[j][1] + m2[j][3];
1829	m1[j][2] = m2[j][0] - m2[j][2];
1830	m1[j][3] = m2[j][1] - m2[j][3];
1831	m1[j][4] = m2[j][4] + m2[j][6];
1832	m1[j][5] = m2[j][5] + m2[j][7];
1833	m1[j][6] = m2[j][4] - m2[j][6];
1834	m1[j][7] = m2[j][5] - m2[j][7];
1835
1836	m2[j][0] = m1[j][0] + m1[j][1];
1837	m2[j][1] = m1[j][0] - m1[j][1];
1838	m2[j][2] = m1[j][2] + m1[j][3];
1839	m2[j][3] = m1[j][2] - m1[j][3];
1840	m2[j][4] = m1[j][4] + m1[j][5];
1841	m2[j][5] = m1[j][4] - m1[j][5];
1842	m2[j][6] = m1[j][6] + m1[j][7];
1843	m2[j][7] = m1[j][6] - m1[j][7];
1844	}
1845
1846	//vertical
1847	for (i=0; i < 8; i++)
1848	{
1849	m3[0][i] = m2[0][i] + m2[4][i];
1850	m3[1][i] = m2[1][i] + m2[5][i];
1851	m3[2][i] = m2[2][i] + m2[6][i];
1852	m3[3][i] = m2[3][i] + m2[7][i];
1853	m3[4][i] = m2[0][i] - m2[4][i];
1854	m3[5][i] = m2[1][i] - m2[5][i];
1855	m3[6][i] = m2[2][i] - m2[6][i];
1856	m3[7][i] = m2[3][i] - m2[7][i];
1857
1858	m1[0][i] = m3[0][i] + m3[2][i];
1859	m1[1][i] = m3[1][i] + m3[3][i];
1860	m1[2][i] = m3[0][i] - m3[2][i];
1861	m1[3][i] = m3[1][i] - m3[3][i];
1862	m1[4][i] = m3[4][i] + m3[6][i];
1863	m1[5][i] = m3[5][i] + m3[7][i];
1864	m1[6][i] = m3[4][i] - m3[6][i];
1865	m1[7][i] = m3[5][i] - m3[7][i];
1866
1867	m2[0][i] = m1[0][i] + m1[1][i];
1868	m2[1][i] = m1[0][i] - m1[1][i];
1869	m2[2][i] = m1[2][i] + m1[3][i];
1870	m2[3][i] = m1[2][i] - m1[3][i];
1871	m2[4][i] = m1[4][i] + m1[5][i];
1872	m2[5][i] = m1[4][i] - m1[5][i];
1873	m2[6][i] = m1[6][i] + m1[7][i];
1874	m2[7][i] = m1[6][i] - m1[7][i];
1875	}
1876
1877	for (i = 0; i < 8; i++)
1878	{
1879	for (j = 0; j < 8; j++)
1880	{
1881	sad += abs(m2[i][j]);
1882	}
1883	}
1884
1885	sad=((sad+2)>>2);
1886
1887	return sad;
1888	}
1889
1890	#if NS_HAD
1891	UInt TComRdCost::xCalcHADs16x4( Pel piOrg, Pel piCur, Int iStrideOrg, Int iStrideCur, Int iStep )
1892	{
1893	Int k, i, j, jj, sad=0;
1894	Int diff[64], m1[4][16], m2[4][16];
1895	assert( iStep == 1 );
1896	for( k = 0; k < 64; k += 16 )
1897	{
1898	diff[k+0] = piOrg[0] - piCur[0];
1899	diff[k+1] = piOrg[1] - piCur[1];
1900	diff[k+2] = piOrg[2] - piCur[2];
1901	diff[k+3] = piOrg[3] - piCur[3];
1902	diff[k+4] = piOrg[4] - piCur[4];
1903	diff[k+5] = piOrg[5] - piCur[5];
1904	diff[k+6] = piOrg[6] - piCur[6];
1905	diff[k+7] = piOrg[7] - piCur[7];
1906
1907	diff[k+8] = piOrg[8] - piCur[8] ;
1908	diff[k+9] = piOrg[9] - piCur[9] ;
1909	diff[k+10] = piOrg[10] - piCur[10];
1910	diff[k+11] = piOrg[11] - piCur[11];
1911	diff[k+12] = piOrg[12] - piCur[12];
1912	diff[k+13] = piOrg[13] - piCur[13];
1913	diff[k+14] = piOrg[14] - piCur[14];
1914	diff[k+15] = piOrg[15] - piCur[15];
1915
1916	piCur += iStrideCur;
1917	piOrg += iStrideOrg;
1918	}
1919
1920	//horizontal
1921	for (j=0; j < 4; j++)
1922	{
1923	jj = j << 4;
1924
1925	m2[j][0] = diff[jj ] + diff[jj+8];
1926	m2[j][1] = diff[jj+1] + diff[jj+9];
1927	m2[j][2] = diff[jj+2] + diff[jj+10];
1928	m2[j][3] = diff[jj+3] + diff[jj+11];
1929	m2[j][4] = diff[jj+4] + diff[jj+12];
1930	m2[j][5] = diff[jj+5] + diff[jj+13];
1931	m2[j][6] = diff[jj+6] + diff[jj+14];
1932	m2[j][7] = diff[jj+7] + diff[jj+15];
1933	m2[j][8] = diff[jj ] - diff[jj+8];
1934	m2[j][9] = diff[jj+1] - diff[jj+9];
1935	m2[j][10] = diff[jj+2] - diff[jj+10];
1936	m2[j][11] = diff[jj+3] - diff[jj+11];
1937	m2[j][12] = diff[jj+4] - diff[jj+12];
1938	m2[j][13] = diff[jj+5] - diff[jj+13];
1939	m2[j][14] = diff[jj+6] - diff[jj+14];
1940	m2[j][15] = diff[jj+7] - diff[jj+15];
1941
1942	m1[j][0] = m2[j][0] + m2[j][4];
1943	m1[j][1] = m2[j][1] + m2[j][5];
1944	m1[j][2] = m2[j][2] + m2[j][6];
1945	m1[j][3] = m2[j][3] + m2[j][7];
1946	m1[j][4] = m2[j][0] - m2[j][4];
1947	m1[j][5] = m2[j][1] - m2[j][5];
1948	m1[j][6] = m2[j][2] - m2[j][6];
1949	m1[j][7] = m2[j][3] - m2[j][7];
1950	m1[j][8] = m2[j][8] + m2[j][12];
1951	m1[j][9] = m2[j][9] + m2[j][13];
1952	m1[j][10] = m2[j][10] + m2[j][14];
1953	m1[j][11] = m2[j][11] + m2[j][15];
1954	m1[j][12] = m2[j][8] - m2[j][12];
1955	m1[j][13] = m2[j][9] - m2[j][13];
1956	m1[j][14] = m2[j][10] - m2[j][14];
1957	m1[j][15] = m2[j][11] - m2[j][15];
1958
1959	m2[j][0] = m1[j][0] + m1[j][2];
1960	m2[j][1] = m1[j][1] + m1[j][3];
1961	m2[j][2] = m1[j][0] - m1[j][2];
1962	m2[j][3] = m1[j][1] - m1[j][3];
1963	m2[j][4] = m1[j][4] + m1[j][6];
1964	m2[j][5] = m1[j][5] + m1[j][7];
1965	m2[j][6] = m1[j][4] - m1[j][6];
1966	m2[j][7] = m1[j][5] - m1[j][7];
1967	m2[j][8] = m1[j][8] + m1[j][10];
1968	m2[j][9] = m1[j][9] + m1[j][11];
1969	m2[j][10] = m1[j][8] - m1[j][10];
1970	m2[j][11] = m1[j][9] - m1[j][11];
1971	m2[j][12] = m1[j][12] + m1[j][14];
1972	m2[j][13] = m1[j][13] + m1[j][15];
1973	m2[j][14] = m1[j][12] - m1[j][14];
1974	m2[j][15] = m1[j][13] - m1[j][15];
1975
1976	m1[j][0] = m2[j][0] + m2[j][1];
1977	m1[j][1] = m2[j][0] - m2[j][1];
1978	m1[j][2] = m2[j][2] + m2[j][3];
1979	m1[j][3] = m2[j][2] - m2[j][3];
1980	m1[j][4] = m2[j][4] + m2[j][5];
1981	m1[j][5] = m2[j][4] - m2[j][5];
1982	m1[j][6] = m2[j][6] + m2[j][7];
1983	m1[j][7] = m2[j][6] - m2[j][7];
1984	m1[j][8] = m2[j][8] + m2[j][9];
1985	m1[j][9] = m2[j][8] - m2[j][9];
1986	m1[j][10] = m2[j][10] + m2[j][11];
1987	m1[j][11] = m2[j][10] - m2[j][11];
1988	m1[j][12] = m2[j][12] + m2[j][13];
1989	m1[j][13] = m2[j][12] - m2[j][13];
1990	m1[j][14] = m2[j][14] + m2[j][15];
1991	m1[j][15] = m2[j][14] - m2[j][15];
1992	}
1993
1994	//vertical
1995	for (i=0; i < 16; i++)
1996	{
1997	m2[0][i] = m1[0][i] + m1[2][i];
1998	m2[1][i] = m1[1][i] + m1[3][i];
1999	m2[2][i] = m1[0][i] - m1[2][i];
2000	m2[3][i] = m1[1][i] - m1[3][i];
2001
2002	m1[0][i] = m2[0][i] + m2[1][i];
2003	m1[1][i] = m2[0][i] - m2[1][i];
2004	m1[2][i] = m2[2][i] + m2[3][i];
2005	m1[3][i] = m2[2][i] - m2[3][i];
2006	}
2007
2008	for (i = 0; i < 4; i++)
2009	{
2010	for (j = 0; j < 16; j++)
2011	{
2012	sad += abs(m1[i][j]);
2013	}
2014	}
2015
2016	sad=((sad+2)>>2);
2017
2018	return sad;
2019	}
2020
2021	UInt TComRdCost::xCalcHADs4x16( Pel piOrg, Pel piCur, Int iStrideOrg, Int iStrideCur, Int iStep )
2022	{
2023	Int k, i, j, jj, sad=0;
2024	Int diff[64], m1[16][4], m2[16][4], m3[16][4];
2025	assert( iStep == 1 );
2026	for( k = 0; k < 64; k += 4 )
2027	{
2028	diff[k+0] = piOrg[0] - piCur[0];
2029	diff[k+1] = piOrg[1] - piCur[1];
2030	diff[k+2] = piOrg[2] - piCur[2];
2031	diff[k+3] = piOrg[3] - piCur[3];
2032
2033	piCur += iStrideCur;
2034	piOrg += iStrideOrg;
2035	}
2036
2037	//horizontal
2038	for (j=0; j < 16; j++)
2039	{
2040	jj = j << 2;
2041	m2[j][0] = diff[jj ] + diff[jj+2];
2042	m2[j][1] = diff[jj+1] + diff[jj+3];
2043	m2[j][2] = diff[jj ] - diff[jj+2];
2044	m2[j][3] = diff[jj+1] - diff[jj+3];
2045
2046	m1[j][0] = m2[j][0] + m2[j][1];
2047	m1[j][1] = m2[j][0] - m2[j][1];
2048	m1[j][2] = m2[j][2] + m2[j][3];
2049	m1[j][3] = m2[j][2] - m2[j][3];
2050	}
2051
2052	//vertical
2053	for (i=0; i < 4; i++)
2054	{
2055	m2[0][i] = m1[0][i] + m1[8][i];
2056	m2[1][i] = m1[1][i] + m1[9][i];
2057	m2[2][i] = m1[2][i] + m1[10][i];
2058	m2[3][i] = m1[3][i] + m1[11][i];
2059	m2[4][i] = m1[4][i] + m1[12][i];
2060	m2[5][i] = m1[5][i] + m1[13][i];
2061	m2[6][i] = m1[6][i] + m1[14][i];
2062	m2[7][i] = m1[7][i] + m1[15][i];
2063	m2[8][i] = m1[0][i] - m1[8][i];
2064	m2[9][i] = m1[1][i] - m1[9][i];
2065	m2[10][i] = m1[2][i] - m1[10][i];
2066	m2[11][i] = m1[3][i] - m1[11][i];
2067	m2[12][i] = m1[4][i] - m1[12][i];
2068	m2[13][i] = m1[5][i] - m1[13][i];
2069	m2[14][i] = m1[6][i] - m1[14][i];
2070	m2[15][i] = m1[7][i] - m1[15][i];
2071
2072	m3[0][i] = m2[0][i] + m2[4][i];
2073	m3[1][i] = m2[1][i] + m2[5][i];
2074	m3[2][i] = m2[2][i] + m2[6][i];
2075	m3[3][i] = m2[3][i] + m2[7][i];
2076	m3[4][i] = m2[0][i] - m2[4][i];
2077	m3[5][i] = m2[1][i] - m2[5][i];
2078	m3[6][i] = m2[2][i] - m2[6][i];
2079	m3[7][i] = m2[3][i] - m2[7][i];
2080	m3[8][i] = m2[8][i] + m2[12][i];
2081	m3[9][i] = m2[9][i] + m2[13][i];
2082	m3[10][i] = m2[10][i] + m2[14][i];
2083	m3[11][i] = m2[11][i] + m2[15][i];
2084	m3[12][i] = m2[8][i] - m2[12][i];
2085	m3[13][i] = m2[9][i] - m2[13][i];
2086	m3[14][i] = m2[10][i] - m2[14][i];
2087	m3[15][i] = m2[11][i] - m2[15][i];
2088
2089	m1[0][i] = m3[0][i] + m3[2][i];
2090	m1[1][i] = m3[1][i] + m3[3][i];
2091	m1[2][i] = m3[0][i] - m3[2][i];
2092	m1[3][i] = m3[1][i] - m3[3][i];
2093	m1[4][i] = m3[4][i] + m3[6][i];
2094	m1[5][i] = m3[5][i] + m3[7][i];
2095	m1[6][i] = m3[4][i] - m3[6][i];
2096	m1[7][i] = m3[5][i] - m3[7][i];
2097	m1[8][i] = m3[8][i] + m3[10][i];
2098	m1[9][i] = m3[9][i] + m3[11][i];
2099	m1[10][i] = m3[8][i] - m3[10][i];
2100	m1[11][i] = m3[9][i] - m3[11][i];
2101	m1[12][i] = m3[12][i] + m3[14][i];
2102	m1[13][i] = m3[13][i] + m3[15][i];
2103	m1[14][i] = m3[12][i] - m3[14][i];
2104	m1[15][i] = m3[13][i] - m3[15][i];
2105
2106	m2[0][i] = m1[0][i] + m1[1][i];
2107	m2[1][i] = m1[0][i] - m1[1][i];
2108	m2[2][i] = m1[2][i] + m1[3][i];
2109	m2[3][i] = m1[2][i] - m1[3][i];
2110	m2[4][i] = m1[4][i] + m1[5][i];
2111	m2[5][i] = m1[4][i] - m1[5][i];
2112	m2[6][i] = m1[6][i] + m1[7][i];
2113	m2[7][i] = m1[6][i] - m1[7][i];
2114	m2[8][i] = m1[8][i] + m1[9][i];
2115	m2[9][i] = m1[8][i] - m1[9][i];
2116	m2[10][i] = m1[10][i] + m1[11][i];
2117	m2[11][i] = m1[10][i] - m1[11][i];
2118	m2[12][i] = m1[12][i] + m1[13][i];
2119	m2[13][i] = m1[12][i] - m1[13][i];
2120	m2[14][i] = m1[14][i] + m1[15][i];
2121	m2[15][i] = m1[14][i] - m1[15][i];
2122	}
2123
2124	for (i = 0; i < 16; i++)
2125	{
2126	for (j = 0; j < 4; j++)
2127	{
2128	sad += abs(m2[i][j]);
2129	}
2130	}
2131
2132	sad=((sad+2)>>2);
2133
2134	return sad;
2135	}
2136	#endif
2137
2138	UInt TComRdCost::xGetHADs4( DistParam* pcDtParam )
2139	{
2140	if ( pcDtParam->bApplyWeight )
2141	{
2142	return xGetHADs4w( pcDtParam );
2143	}
2144	Pel* piOrg = pcDtParam->pOrg;
2145	Pel* piCur = pcDtParam->pCur;
2146	Int iRows = pcDtParam->iRows;
2147	Int iStrideCur = pcDtParam->iStrideCur;
2148	Int iStrideOrg = pcDtParam->iStrideOrg;
2149	Int iStep = pcDtParam->iStep;
2150	Int y;
2151	Int iOffsetOrg = iStrideOrg<<2;
2152	Int iOffsetCur = iStrideCur<<2;
2153
2154	UInt uiSum = 0;
2155
2156	for ( y=0; y<iRows; y+= 4 )
2157	{
2158	uiSum += xCalcHADs4x4( piOrg, piCur, iStrideOrg, iStrideCur, iStep );
2159	piOrg += iOffsetOrg;
2160	piCur += iOffsetCur;
2161	}
2162
2163	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
2164	}
2165
2166	UInt TComRdCost::xGetHADs8( DistParam* pcDtParam )
2167	{
2168	if ( pcDtParam->bApplyWeight )
2169	{
2170	return xGetHADs8w( pcDtParam );
2171	}
2172	Pel* piOrg = pcDtParam->pOrg;
2173	Pel* piCur = pcDtParam->pCur;
2174	Int iRows = pcDtParam->iRows;
2175	Int iStrideCur = pcDtParam->iStrideCur;
2176	Int iStrideOrg = pcDtParam->iStrideOrg;
2177	Int iStep = pcDtParam->iStep;
2178	Int y;
2179
2180	UInt uiSum = 0;
2181
2182	if ( iRows == 4 )
2183	{
2184	uiSum += xCalcHADs4x4( piOrg+0, piCur , iStrideOrg, iStrideCur, iStep );
2185	uiSum += xCalcHADs4x4( piOrg+4, piCur+4*iStep, iStrideOrg, iStrideCur, iStep );
2186	}
2187	else
2188	{
2189	Int iOffsetOrg = iStrideOrg<<3;
2190	Int iOffsetCur = iStrideCur<<3;
2191	for ( y=0; y<iRows; y+= 8 )
2192	{
2193	uiSum += xCalcHADs8x8( piOrg, piCur, iStrideOrg, iStrideCur, iStep );
2194	piOrg += iOffsetOrg;
2195	piCur += iOffsetCur;
2196	}
2197	}
2198
2199	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
2200	}
2201
2202	UInt TComRdCost::xGetHADs( DistParam* pcDtParam )
2203	{
2204	if ( pcDtParam->bApplyWeight )
2205	{
2206	return xGetHADsw( pcDtParam );
2207	}
2208	Pel* piOrg = pcDtParam->pOrg;
2209	Pel* piCur = pcDtParam->pCur;
2210	Int iRows = pcDtParam->iRows;
2211	Int iCols = pcDtParam->iCols;
2212	Int iStrideCur = pcDtParam->iStrideCur;
2213	Int iStrideOrg = pcDtParam->iStrideOrg;
2214	Int iStep = pcDtParam->iStep;
2215
2216	Int x, y;
2217
2218	UInt uiSum = 0;
2219
2220	#if NS_HAD
2221	if( ( ( iRows % 8 == 0) && (iCols % 8 == 0) && ( iRows == iCols ) ) \|\| ( ( iRows % 8 == 0 ) && (iCols % 8 == 0) && !pcDtParam->bUseNSHAD ) )
2222	#else
2223	if( ( iRows % 8 == 0) && (iCols % 8 == 0) )
2224	#endif
2225	{
2226	Int iOffsetOrg = iStrideOrg<<3;
2227	Int iOffsetCur = iStrideCur<<3;
2228	for ( y=0; y<iRows; y+= 8 )
2229	{
2230	for ( x=0; x<iCols; x+= 8 )
2231	{
2232	uiSum += xCalcHADs8x8( &piOrg[x], &piCur[x*iStep], iStrideOrg, iStrideCur, iStep );
2233	}
2234	piOrg += iOffsetOrg;
2235	piCur += iOffsetCur;
2236	}
2237	}
2238	#if NS_HAD
2239	else if ( ( iCols > 8 ) && ( iCols > iRows ) && pcDtParam->bUseNSHAD )
2240	{
2241	Int iOffsetOrg = iStrideOrg<<2;
2242	Int iOffsetCur = iStrideCur<<2;
2243	for ( y=0; y<iRows; y+= 4 )
2244	{
2245	for ( x=0; x<iCols; x+= 16 )
2246	{
2247	uiSum += xCalcHADs16x4( &piOrg[x], &piCur[x*iStep], iStrideOrg, iStrideCur, iStep );
2248	}
2249	piOrg += iOffsetOrg;
2250	piCur += iOffsetCur;
2251	}
2252	}
2253	else if ( ( iRows > 8 ) && ( iCols < iRows ) && pcDtParam->bUseNSHAD )
2254	{
2255	Int iOffsetOrg = iStrideOrg<<4;
2256	Int iOffsetCur = iStrideCur<<4;
2257	for ( y=0; y<iRows; y+= 16 )
2258	{
2259	for ( x=0; x<iCols; x+= 4 )
2260	{
2261	uiSum += xCalcHADs4x16( &piOrg[x], &piCur[x*iStep], iStrideOrg, iStrideCur, iStep );
2262	}
2263	piOrg += iOffsetOrg;
2264	piCur += iOffsetCur;
2265	}
2266	}
2267	#endif
2268	else if( ( iRows % 4 == 0) && (iCols % 4 == 0) )
2269	{
2270	Int iOffsetOrg = iStrideOrg<<2;
2271	Int iOffsetCur = iStrideCur<<2;
2272
2273	for ( y=0; y<iRows; y+= 4 )
2274	{
2275	for ( x=0; x<iCols; x+= 4 )
2276	{
2277	uiSum += xCalcHADs4x4( &piOrg[x], &piCur[x*iStep], iStrideOrg, iStrideCur, iStep );
2278	}
2279	piOrg += iOffsetOrg;
2280	piCur += iOffsetCur;
2281	}
2282	}
2283	else if( ( iRows % 2 == 0) && (iCols % 2 == 0) )
2284	{
2285	Int iOffsetOrg = iStrideOrg<<1;
2286	Int iOffsetCur = iStrideCur<<1;
2287	for ( y=0; y<iRows; y+=2 )
2288	{
2289	for ( x=0; x<iCols; x+=2 )
2290	{
2291	uiSum += xCalcHADs2x2( &piOrg[x], &piCur[x*iStep], iStrideOrg, iStrideCur, iStep );
2292	}
2293	piOrg += iOffsetOrg;
2294	piCur += iOffsetCur;
2295	}
2296	}
2297	else
2298	{
2299	assert(false);
2300	}
2301
2302	return uiSum >> DISTORTION_PRECISION_ADJUSTMENT(pcDtParam->bitDepth-8);
2303	}
2304	#if H_3D_VSO
2305	Void TComRdCost::setLambdaVSO( Double dLambdaVSO )
2306	{
2307	m_dLambdaVSO = dLambdaVSO;
2308	m_dSqrtLambdaVSO = sqrt(m_dLambdaVSO);
2309	m_uiLambdaMotionSADVSO = (UInt)floor(65536.0 * m_dSqrtLambdaVSO);
2310	m_uiLambdaMotionSSEVSO = (UInt)floor(65536.0 * m_dLambdaVSO );
2311	}
2312
2313	Dist TComRdCost::xGetDistVSOMode4( Int iStartPosX, Int iStartPosY, Pel* piCur, Int iCurStride, Pel* piOrg, Int iOrgStride, UInt uiBlkWidth, UInt uiBlkHeight, Bool bSAD )
2314	{
2315	AOT(bSAD);
2316	#if H_3D_VSO_EARLY_SKIP
2317	RMDist iDist = m_pcRenModel->getDist( iStartPosX, iStartPosY, (Int) uiBlkWidth, (Int) uiBlkHeight, iCurStride, piCur, piOrg, iOrgStride);
2318	#else
2319	RMDist iDist = m_pcRenModel->getDist( iStartPosX, iStartPosY, (Int) uiBlkWidth, (Int) uiBlkHeight, iCurStride, piCur );
2320	#endif
2321
2322	RMDist iDistMin = (RMDist) RDO_DIST_MIN;
2323	iDistMin = m_bAllowNegDist ? RDO_DIST_MIN : 0;
2324
2325	iDist = std::min( iDist, (RMDist) RDO_DIST_MAX);
2326	iDist = std::max( iDist, iDistMin);
2327	return (Dist) iDist;
2328	}
2329
2330
2331	Dist TComRdCost::getDistPartVSO( TComDataCU* pcCU, UInt uiAbsPartIndex, Pel* piCur, Int iCurStride, Pel* piOrg, Int iOrgStride, UInt uiBlkWidth, UInt uiBlkHeight, Bool bHAD )
2332	{
2333	assert( m_bUseVSO );
2334	assert( this->m_fpDistortFuncVSO != 0 );
2335
2336	Int iPosX;
2337	Int iPosY;
2338
2339	pcCU->getPosInPic( uiAbsPartIndex, iPosX, iPosY );
2340
2341	Dist dist = (this->*m_fpDistortFuncVSO) ( iPosX, iPosY, piCur, iCurStride, piOrg, iOrgStride, uiBlkWidth, uiBlkHeight, bHAD );
2342
2343	if ( m_bUseWVSO )
2344	{
2345	Int iDWeight = m_iDWeight * m_iDWeight;
2346	Int iVSOWeight = m_iVSOWeight * m_iVSOWeight;
2347	Dist distDepth;
2348
2349	if ( !bHAD )
2350	{
2351	distDepth = (Dist) getDistPart( g_bitDepthY, piCur, iCurStride, piOrg, iOrgStride, uiBlkWidth, uiBlkHeight);
2352	}
2353	else
2354	{
2355	distDepth = (Dist) calcHAD( g_bitDepthY, piCur, iCurStride, piOrg, iOrgStride, uiBlkWidth, uiBlkHeight);
2356	}
2357
2358	dist = (Dist) (iDWeight * distDepth + iVSOWeight * dist ) / ( iDWeight + iVSOWeight);
2359	}
2360	return dist;
2361	};
2362
2363
2364	Void TComRdCost::setVSOMode( UInt uiIn )
2365	{
2366	m_uiVSOMode = uiIn;
2367	switch (m_uiVSOMode )
2368	{
2369	case 4:
2370	m_fpDistortFuncVSO = &TComRdCost::xGetDistVSOMode4;
2371	break;
2372	default:
2373	assert(0);
2374	break;
2375	}
2376	}
2377
2378
2379	Double TComRdCost::calcRdCostVSO( UInt uiBits, Dist uiDistortion, Bool bFlag, DFunc eDFunc )
2380	{
2381	assert( m_bUseLambdaScaleVSO );
2382
2383	Double dRdCost = 0.0;
2384	Double dLambda = 0.0;
2385
2386	switch ( eDFunc )
2387	{
2388	case DF_SSE:
2389	assert(0);
2390	break;
2391	case DF_SAD:
2392	dLambda = (Double)m_uiLambdaMotionSADVSO;
2393	break;
2394	case DF_DEFAULT:
2395	dLambda = m_dLambdaVSO;
2396	break;
2397	case DF_SSE_FRAME:
2398	dLambda = m_dFrameLambdaVSO;
2399	break;
2400	default:
2401	assert (0);
2402	break;
2403	}
2404
2405	if (bFlag)
2406	{
2407	// Intra8x8, Intra4x4 Block only...
2408	#if SEQUENCE_LEVEL_LOSSLESS
2409	dRdCost = (Double)(uiBits);
2410	#else
2411	dRdCost = (((Double)uiDistortion) + ((Double)uiBits * dLambda));
2412	#endif
2413	}
2414	else
2415	{
2416	if (eDFunc == DF_SAD)
2417	{
2418	dRdCost = ((Double)uiDistortion + (Double)((Int)(uiBits * dLambda+.5)>>16));
2419	dRdCost = (Double)(Dist)floor(dRdCost);
2420	}
2421	else
2422	{
2423	#if SEQUENCE_LEVEL_LOSSLESS
2424	dRdCost = (Double)(uiBits);
2425	#else
2426	dRdCost = ((Double)uiDistortion + (Double)((Int)(uiBits * dLambda+.5)));
2427	dRdCost = (Double)(Dist)floor(dRdCost);
2428	#endif
2429	}
2430	}
2431
2432	return dRdCost;
2433	}
2434
2435	Void TComRdCost::setRenModelData( TComDataCU* pcCU, UInt uiAbsPartIndex, Pel* piData, Int iStride, Int iBlkWidth, Int iBlkHeight )
2436	{
2437	UInt iBlkX = g_auiRasterToPelX[g_auiZscanToRaster[uiAbsPartIndex]];
2438	UInt iBlkY = g_auiRasterToPelY[g_auiZscanToRaster[uiAbsPartIndex]];
2439
2440	Int iStartPosX = iBlkX + pcCU->getCUPelX();
2441	Int iStartPosY = iBlkY + pcCU->getCUPelY();
2442
2443	m_pcRenModel->setData( iStartPosX, iStartPosY, iBlkWidth, iBlkHeight, iStride, piData );
2444	}
2445
2446	Void TComRdCost::setAllowNegDist( Bool bAllowNegDist )
2447	{
2448	m_bAllowNegDist = bAllowNegDist;
2449	}
2450	#endif
2451
2452	//! \}

Note: See TracBrowser for help on using the repository browser.

JCT-3V 3D-HEVC

Context navigation

source: 3DVCSoftware/branches/HTM-DEV-0.3-dev1/source/Lib/TLibCommon/TComRdCost.cpp @ 473

Download in other formats: