openh264/codec/encoder/core/src/mv_pred.cpp
Martin Storsjö 89f2c8bf7c Remove a duplicate median function
The decoder used WelsMedian while the encoder used WELS_MEDIAN.
The former has two different implementations, WELS_MEDIAN was
identical to the disabled version of WelsMedian.

Settle on using the same implementation for both decoder and
encoder - whichever version of the implementations is faster
should be used for both.
2014-01-24 15:57:11 +02:00

363 lines
14 KiB
C++

/*!
* \copy
* Copyright (c) 2009-2013, Cisco Systems
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
*
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
* COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*
*
* \file mv_pred.c
*
* \brief Get MV predictor and update motion vector of mb cache
*
* \date 05/22/2009 Created
*
*************************************************************************************
*/
#include "mv_pred.h"
#include "ls_defines.h"
namespace WelsSVCEnc {
//basic pMv prediction unit for pMv width (4, 2, 1)
void PredMv (const SMVComponentUnit* kpMvComp, int8_t iPartIdx, int8_t iPartW, int32_t iRef, SMVUnitXY* sMvp) {
const uint8_t kuiLeftIdx = g_kuiCache30ScanIdx[iPartIdx] - 1;
const uint8_t kuiTopIdx = g_kuiCache30ScanIdx[iPartIdx] - 6;
int32_t iMatchRef;
int32_t iLeftRef = kpMvComp->iRefIndexCache[kuiLeftIdx];
int32_t iTopRef = kpMvComp->iRefIndexCache[ kuiTopIdx];
int32_t iRightTopRef = kpMvComp->iRefIndexCache[kuiTopIdx + iPartW];
int32_t iDiagonalRef;
SMVUnitXY sMvA (kpMvComp->sMotionVectorCache[kuiLeftIdx]);
SMVUnitXY sMvB (kpMvComp->sMotionVectorCache[kuiTopIdx]);
SMVUnitXY sMvC;
if (REF_NOT_AVAIL == iRightTopRef) {
iDiagonalRef = kpMvComp->iRefIndexCache[ kuiTopIdx - 1];// left_top;
sMvC = kpMvComp->sMotionVectorCache[kuiTopIdx - 1];
} else {
iDiagonalRef = iRightTopRef;// right_top;
sMvC = kpMvComp->sMotionVectorCache[kuiTopIdx + iPartW];
}
if ((REF_NOT_AVAIL == iTopRef) && (REF_NOT_AVAIL == iDiagonalRef) && iLeftRef != REF_NOT_AVAIL) {
*sMvp = sMvA;
return;
}
// b2[diag] b1[top] b0[left] is available!
iMatchRef = (iRef == iLeftRef) << MB_LEFT_BIT;
iMatchRef |= (iRef == iTopRef) << MB_TOP_BIT;
iMatchRef |= (iRef == iDiagonalRef) << MB_TOPRIGHT_BIT;
switch (iMatchRef) {
case LEFT_MB_POS:// A
*sMvp = sMvA;
break;
case TOP_MB_POS:// B
*sMvp = sMvB;
break;
case TOPRIGHT_MB_POS:// C or D
*sMvp = sMvC;
break;
default:
sMvp->iMvX = WelsMedian (sMvA.iMvX, sMvB.iMvX, sMvC.iMvX);
sMvp->iMvY = WelsMedian (sMvA.iMvY, sMvB.iMvY, sMvC.iMvY);
break;
}
}
void PredInter8x16Mv (SMbCache* pMbCache, int32_t iPartIdx, int8_t iRef, SMVUnitXY* sMvp) {
const SMVComponentUnit* kpMvComp = &pMbCache->sMvComponents;
if (0 == iPartIdx) {
const int8_t kiLeftRef = kpMvComp->iRefIndexCache[6];
if (iRef == kiLeftRef) {
*sMvp = kpMvComp->sMotionVectorCache[6];
return;
}
} else { // 1 == iPartIdx
int8_t iDiagonalRef = kpMvComp->iRefIndexCache[5]; //top-right
int8_t iIndex = 5;
if (REF_NOT_AVAIL == iDiagonalRef) {
iDiagonalRef = kpMvComp->iRefIndexCache[2]; //top-left for 8*8 block(iIndex 1)
iIndex = 2;
}
if (iRef == iDiagonalRef) {
*sMvp = kpMvComp->sMotionVectorCache[iIndex];
return;
}
}
PredMv (kpMvComp, iPartIdx, 2, iRef, sMvp);
}
void PredInter16x8Mv (SMbCache* pMbCache, int32_t iPartIdx, int8_t iRef, SMVUnitXY* sMvp) {
const SMVComponentUnit* kpMvComp = &pMbCache->sMvComponents;
if (0 == iPartIdx) {
const int8_t kiTopRef = kpMvComp->iRefIndexCache[1];
if (iRef == kiTopRef) {
*sMvp = kpMvComp->sMotionVectorCache[1];
return;
}
} else { // 8 == iPartIdx
const int8_t kiLeftRef = kpMvComp->iRefIndexCache[18];
if (iRef == kiLeftRef) {
*sMvp = kpMvComp->sMotionVectorCache[18];
return;
}
}
PredMv (kpMvComp, iPartIdx, 4, iRef, sMvp);
}
void PredSkipMv (SMbCache* pMbCache, SMVUnitXY* sMvp) {
const SMVComponentUnit* kpMvComp = &pMbCache->sMvComponents;
const int8_t kiLeftRef = kpMvComp->iRefIndexCache[6]; //A
const int8_t kiTopRef = kpMvComp->iRefIndexCache[1]; //B
if (REF_NOT_AVAIL == kiLeftRef || REF_NOT_AVAIL == kiTopRef ||
(0 == kiLeftRef && 0 == * (int32_t*) (&kpMvComp->sMotionVectorCache[6])) ||
(0 == kiTopRef && 0 == * (int32_t*) (&kpMvComp->sMotionVectorCache[1]))) {
ST32 (sMvp, 0);
return;
}
PredMv (kpMvComp, 0, 4, 0, sMvp);
}
//update pMv and uiRefIndex cache for current MB, only for P_16*16 (SKIP inclusive)
void UpdateP16x16MotionInfo (SMbCache* pMbCache, SMB* pCurMb, const int8_t kiRef, SMVUnitXY* pMv) {
// optimized 11/25/2011
SMVComponentUnit* pMvComp = &pMbCache->sMvComponents;
const uint32_t kuiMv32 = LD32 (pMv);
const uint64_t kuiMv64 = BUTTERFLY4x8 (kuiMv32);
uint64_t uiMvBuf[8] = { kuiMv64, kuiMv64, kuiMv64, kuiMv64, kuiMv64, kuiMv64, kuiMv64, kuiMv64 };
const uint16_t kuiRef16 = BUTTERFLY1x2 (kiRef);
const uint32_t kuiRef32 = BUTTERFLY2x4 (kuiRef16);
ST32 (pCurMb->pRefIndex, kuiRef32);
// update pMv range from 0~15
memcpy (pCurMb->sMv, uiMvBuf, sizeof (uiMvBuf)); // confirmed_safe_unsafe_usage
/*
* blocks 0: 7~10, 1: 13~16, 2: 19~22, 3: 25~28
*/
pMvComp->iRefIndexCache[7] = kiRef;
ST16 (&pMvComp->iRefIndexCache[8], kuiRef16);
pMvComp->iRefIndexCache[10] = kiRef;
pMvComp->iRefIndexCache[13] = kiRef;
ST16 (&pMvComp->iRefIndexCache[14], kuiRef16);
pMvComp->iRefIndexCache[16] = kiRef;
pMvComp->iRefIndexCache[19] = kiRef;
ST16 (&pMvComp->iRefIndexCache[20], kuiRef16);
pMvComp->iRefIndexCache[22] = kiRef;
pMvComp->iRefIndexCache[25] = kiRef;
ST16 (&pMvComp->iRefIndexCache[26], kuiRef16);
pMvComp->iRefIndexCache[28] = kiRef;
/*
* blocks 0: 7~10, 1: 13~16, 2: 19~22, 3: 25~28
*/
pMvComp->sMotionVectorCache[7] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[8], kuiMv64);
pMvComp->sMotionVectorCache[10] = *pMv;
pMvComp->sMotionVectorCache[13] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[14], kuiMv64);
pMvComp->sMotionVectorCache[16] = *pMv;
pMvComp->sMotionVectorCache[19] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[20], kuiMv64);
pMvComp->sMotionVectorCache[22] = *pMv;
pMvComp->sMotionVectorCache[25] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[26], kuiMv64);
pMvComp->sMotionVectorCache[28] = *pMv;
}
//update uiRefIndex and pMv of both SMB and Mb_cache, only for P16x8
void UpdateP16x8MotionInfo (SMbCache* pMbCache, SMB* pCurMb, const int32_t kiPartIdx, const int8_t kiRef,
SMVUnitXY* pMv) {
// optimized 11/25/2011
SMVComponentUnit* pMvComp = &pMbCache->sMvComponents;
const uint32_t kuiMv32 = LD32 (pMv);
const uint64_t kuiMv64 = BUTTERFLY4x8 (kuiMv32);
uint64_t uiMvBuf[4] = { kuiMv64, kuiMv64, kuiMv64, kuiMv64 };
const int16_t kiScan4Idx = g_kuiMbCountScan4Idx[kiPartIdx];
const int16_t kiCacheIdx = g_kuiCache30ScanIdx[kiPartIdx];
const int16_t kiCacheIdx1 = 1 + kiCacheIdx;
const int16_t kiCacheIdx3 = 3 + kiCacheIdx;
const int16_t kiCacheIdx6 = 6 + kiCacheIdx;
const int16_t kiCacheIdx7 = 7 + kiCacheIdx;
const int16_t kiCacheIdx9 = 9 + kiCacheIdx;
const uint16_t kuiRef16 = BUTTERFLY1x2 (kiRef);
ST16 (&pCurMb->pRefIndex[ (kiPartIdx >> 2)], kuiRef16);
memcpy (&pCurMb->sMv[kiScan4Idx], uiMvBuf, sizeof (uiMvBuf)); // confirmed_safe_unsafe_usage
/*
* blocks 0: g_kuiCache30ScanIdx[iPartIdx]~g_kuiCache30ScanIdx[iPartIdx]+3, 1: g_kuiCache30ScanIdx[iPartIdx]+6~g_kuiCache30ScanIdx[iPartIdx]+9
*/
pMvComp->iRefIndexCache[kiCacheIdx] = kiRef;
ST16 (&pMvComp->iRefIndexCache[kiCacheIdx1], kuiRef16);
pMvComp->iRefIndexCache[kiCacheIdx3] = kiRef;
pMvComp->iRefIndexCache[kiCacheIdx6] = kiRef;
ST16 (&pMvComp->iRefIndexCache[kiCacheIdx7], kuiRef16);
pMvComp->iRefIndexCache[kiCacheIdx9] = kiRef;
/*
* blocks 0: g_kuiCache30ScanIdx[iPartIdx]~g_kuiCache30ScanIdx[iPartIdx]+3, 1: g_kuiCache30ScanIdx[iPartIdx]+6~g_kuiCache30ScanIdx[iPartIdx]+9
*/
pMvComp->sMotionVectorCache[kiCacheIdx] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[kiCacheIdx1], kuiMv64);
pMvComp->sMotionVectorCache[kiCacheIdx3] = *pMv;
pMvComp->sMotionVectorCache[kiCacheIdx6] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[kiCacheIdx7], kuiMv64);
pMvComp->sMotionVectorCache[kiCacheIdx9] = *pMv;
}
//update uiRefIndex and pMv of both SMB and Mb_cache, only for P8x16
void update_P8x16_motion_info (SMbCache* pMbCache, SMB* pCurMb, const int32_t kiPartIdx, const int8_t kiRef,
SMVUnitXY* pMv) {
// optimized 11/25/2011
SMVComponentUnit* pMvComp = &pMbCache->sMvComponents;
const uint32_t kuiMv32 = LD32 (pMv);
const uint64_t kuiMv64 = BUTTERFLY4x8 (kuiMv32);
const int16_t kiScan4Idx = g_kuiMbCountScan4Idx[kiPartIdx];
const int16_t kiCacheIdx = g_kuiCache30ScanIdx[kiPartIdx];
const int16_t kiCacheIdx1 = 1 + kiCacheIdx;
const int16_t kiCacheIdx3 = 3 + kiCacheIdx;
const int16_t kiCacheIdx12 = 12 + kiCacheIdx;
const int16_t kiCacheIdx13 = 13 + kiCacheIdx;
const int16_t kiCacheIdx15 = 15 + kiCacheIdx;
const int16_t kiBlkIdx = kiPartIdx >> 2;
const uint16_t kuiRef16 = BUTTERFLY1x2 (kiRef);
pCurMb->pRefIndex[kiBlkIdx] = kiRef;
pCurMb->pRefIndex[2 + kiBlkIdx] = kiRef;
ST64 (&pCurMb->sMv[kiScan4Idx], kuiMv64);
ST64 (&pCurMb->sMv[4 + kiScan4Idx], kuiMv64);
ST64 (&pCurMb->sMv[8 + kiScan4Idx], kuiMv64);
ST64 (&pCurMb->sMv[12 + kiScan4Idx], kuiMv64);
/*
* blocks 0: g_kuiCache30ScanIdx[iPartIdx]~g_kuiCache30ScanIdx[iPartIdx]+3, 1: g_kuiCache30ScanIdx[iPartIdx]+6~g_kuiCache30ScanIdx[iPartIdx]+9
*/
pMvComp->iRefIndexCache[kiCacheIdx] = kiRef;
ST16 (&pMvComp->iRefIndexCache[kiCacheIdx1], kuiRef16);
pMvComp->iRefIndexCache[kiCacheIdx3] = kiRef;
pMvComp->iRefIndexCache[kiCacheIdx12] = kiRef;
ST16 (&pMvComp->iRefIndexCache[kiCacheIdx13], kuiRef16);
pMvComp->iRefIndexCache[kiCacheIdx15] = kiRef;
/*
* blocks 0: g_kuiCache30ScanIdx[iPartIdx]~g_kuiCache30ScanIdx[iPartIdx]+3, 1: g_kuiCache30ScanIdx[iPartIdx]+6~g_kuiCache30ScanIdx[iPartIdx]+9
*/
pMvComp->sMotionVectorCache[kiCacheIdx] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[kiCacheIdx1], kuiMv64);
pMvComp->sMotionVectorCache[kiCacheIdx3] = *pMv;
pMvComp->sMotionVectorCache[kiCacheIdx12] = *pMv;
ST64 (&pMvComp->sMotionVectorCache[kiCacheIdx13], kuiMv64);
pMvComp->sMotionVectorCache[kiCacheIdx15] = *pMv;
}
//update uiRefIndex and pMv of both SMB and Mb_cache, only for P8x8
void UpdateP8x8MotionInfo (SMbCache* pMbCache, SMB* pCurMb, const int32_t kiPartIdx, const int8_t kiRef,
SMVUnitXY* pMv) {
SMVComponentUnit* pMvComp = &pMbCache->sMvComponents;
const uint32_t kuiMv32 = LD32 (pMv);
const uint64_t kuiMv64 = BUTTERFLY4x8 (kuiMv32);
const int16_t kiScan4Idx = g_kuiMbCountScan4Idx[kiPartIdx];
const int16_t kiCacheIdx = g_kuiCache30ScanIdx[kiPartIdx];
const int16_t kiCacheIdx1 = 1 + kiCacheIdx;
const int16_t kiCacheIdx6 = 6 + kiCacheIdx;
const int16_t kiCacheIdx7 = 7 + kiCacheIdx;
//mb
ST64 (&pCurMb->sMv[ kiScan4Idx], kuiMv64);
ST64 (&pCurMb->sMv[4 + kiScan4Idx], kuiMv64);
//cache
pMvComp->iRefIndexCache[kiCacheIdx ] =
pMvComp->iRefIndexCache[kiCacheIdx1] =
pMvComp->iRefIndexCache[kiCacheIdx6] =
pMvComp->iRefIndexCache[kiCacheIdx7] = kiRef;
pMvComp->sMotionVectorCache[kiCacheIdx ] =
pMvComp->sMotionVectorCache[kiCacheIdx1] =
pMvComp->sMotionVectorCache[kiCacheIdx6] =
pMvComp->sMotionVectorCache[kiCacheIdx7] = *pMv;
}
//=========================update motion info(MV and ref_idx) into Mb_cache==========================
//update pMv and uiRefIndex cache only for Mb_cache, only for P_16*16 (SKIP inclusive)
//update uiRefIndex and pMv of only Mb_cache, only for P16x8
void UpdateP16x8Motion2Cache (SMbCache* pMbCache, int32_t iPartIdx, int8_t iRef, SMVUnitXY* pMv) {
SMVComponentUnit* pMvComp = &pMbCache->sMvComponents;
int32_t i;
for (i = 0; i < 2; i++, iPartIdx += 4) {
//cache
const uint8_t kuiCacheIdx = g_kuiCache30ScanIdx[iPartIdx];
pMvComp->iRefIndexCache[ kuiCacheIdx] =
pMvComp->iRefIndexCache[1 + kuiCacheIdx] =
pMvComp->iRefIndexCache[6 + kuiCacheIdx] =
pMvComp->iRefIndexCache[7 + kuiCacheIdx] = iRef;
pMvComp->sMotionVectorCache[ kuiCacheIdx] =
pMvComp->sMotionVectorCache[1 + kuiCacheIdx] =
pMvComp->sMotionVectorCache[6 + kuiCacheIdx] =
pMvComp->sMotionVectorCache[7 + kuiCacheIdx] = *pMv;
}
}
//update uiRefIndex and pMv of only Mb_cache, only for P8x16
void UpdateP8x16Motion2Cache (SMbCache* pMbCache, int32_t iPartIdx, int8_t iRef, SMVUnitXY* pMv) {
SMVComponentUnit* pMvComp = &pMbCache->sMvComponents;
int32_t i;
for (i = 0; i < 2; i++, iPartIdx += 8) {
//cache
const uint8_t kuiCacheIdx = g_kuiCache30ScanIdx[iPartIdx];
pMvComp->iRefIndexCache[ kuiCacheIdx] =
pMvComp->iRefIndexCache[1 + kuiCacheIdx] =
pMvComp->iRefIndexCache[6 + kuiCacheIdx] =
pMvComp->iRefIndexCache[7 + kuiCacheIdx] = iRef;
pMvComp->sMotionVectorCache[ kuiCacheIdx] =
pMvComp->sMotionVectorCache[1 + kuiCacheIdx] =
pMvComp->sMotionVectorCache[6 + kuiCacheIdx] =
pMvComp->sMotionVectorCache[7 + kuiCacheIdx] = *pMv;
}
}
//update uiRefIndex and pMv of only Mb_cache, only for P8x8
void UpdateP8x8Motion2Cache (SMbCache* pMbCache, int32_t iPartIdx, int8_t pRef, SMVUnitXY* pMv) {
SMVComponentUnit* pMvComp = &pMbCache->sMvComponents;
const uint8_t kuiCacheIdx = g_kuiCache30ScanIdx[iPartIdx];
pMvComp->iRefIndexCache[ kuiCacheIdx] =
pMvComp->iRefIndexCache[1 + kuiCacheIdx] =
pMvComp->iRefIndexCache[6 + kuiCacheIdx] =
pMvComp->iRefIndexCache[7 + kuiCacheIdx] = pRef;
pMvComp->sMotionVectorCache[ kuiCacheIdx] =
pMvComp->sMotionVectorCache[1 + kuiCacheIdx] =
pMvComp->sMotionVectorCache[6 + kuiCacheIdx] =
pMvComp->sMotionVectorCache[7 + kuiCacheIdx] = *pMv;
}
} // namespace WelsSVCEnc