[x265] [PATCH] no-rdo: implemented topskip algorithm
sumalatha at multicorewareinc.com
sumalatha at multicorewareinc.com
Tue Nov 19 13:05:36 CET 2013
# HG changeset patch
# User Sumalatha Polureddy
# Date 1384862724 -19800
# Node ID 9b1574807a0884ed68b9fe112586d10745240a81
# Parent 1449a1a2041f264869f8073b2358a75e470324b1
no-rdo: implemented topskip algorithm
It is basically setting the starting depth from which the partitioning should
happen for particular CU. The starting depth for particular CU in present frame
is selected form the previous frame's colocated CU minimum depth.
the performance, bitrate increase and psnr comparison are given below
CLI: x265.exe input.y4m -o abc.hevc -r recon.y4m --rd 1 --ref 1
a. Early exit OFF and Top Skip OFF
b. Early exit OFF and Top Skip ON
BasketballDrive_1920x1080_50
Timetaken to encode: 681/639s
bitrate: 3650/3657 kb/s
PSNR: 36.7/36.703
perf improvement: 6.16% (compared to TopSkip OFF and ON)
perf improvement: 0.19% (compared to TopSkip OFF and ON)
Cactus_1920x1080_50
Timetaken to encode: 530/492s
bitrate: 2787/2795 kb/s
PSNR: 35.527/35.529
perf improvement: 7.16% (compared to TopSkip OFF and ON)
perf improvement: 0.28% (compared to TopSkip OFF and ON)
Kimono1_1920x1080_24
Timetaken to encode: 278/264s
bitrate: 1243/1246 kb/s
PSNR: 38.16/38.16
perf improvement: 5.03% (compared to TopSkip OFF and ON)
perf improvement: 0.24% (compared to TopSkip OFF and ON)
FourPeople_1280x720_60
Timetaken to encode: 173/163s
bitrate: 486/492 kb/s
PSNR: 39.097/39.094
perf improvement: 5.78% (compared to TopSkip OFF and ON)
perf improvement: 1.2% (compared to TopSkip OFF and ON)
PartyScene_832x480_50
Timetaken to encode: 134/119s
bitrate: 1652/1661 kb/s
PSNR: 31.374/31.377
perf improvement: 11.16% (compared to TopSkip OFF and ON)
perf improvement: 0.544% (compared to TopSkip OFF and ON)
big_buck_bunny_360p24
Timetaken to encode: 1772/1477s
bitrate: 174/175 kb/s
PSNR: 37.798/37.797
perf improvement: 16.6% (compared to TopSkip OFF and ON)
perf improvement: 0.5% (compared to TopSkip OFF and ON)
diff -r 1449a1a2041f -r 9b1574807a08 source/Lib/TLibEncoder/TEncCu.cpp
--- a/source/Lib/TLibEncoder/TEncCu.cpp Tue Nov 19 14:49:12 2013 +0530
+++ b/source/Lib/TLibEncoder/TEncCu.cpp Tue Nov 19 17:35:24 2013 +0530
@@ -367,7 +367,7 @@
/* At the start of analysis, the best CU is a null pointer
On return, it points to the CU encode with best chosen mode*/
- xCompressInterCU(outBestCU, m_tempCU[0], cu, 0, 0);
+ xCompressInterCU(outBestCU, m_tempCU[0], cu, 0, 0, 4);
}
else
xCompressCU(m_bestCU[0], m_tempCU[0], 0);
diff -r 1449a1a2041f -r 9b1574807a08 source/Lib/TLibEncoder/TEncCu.h
--- a/source/Lib/TLibEncoder/TEncCu.h Tue Nov 19 14:49:12 2013 +0530
+++ b/source/Lib/TLibEncoder/TEncCu.h Tue Nov 19 17:35:24 2013 +0530
@@ -161,7 +161,7 @@
void finishCU(TComDataCU* cu, uint32_t absPartIdx, uint32_t depth);
void xCompressCU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, uint32_t depth, PartSize parentSize = SIZE_NONE);
void xCompressIntraCU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, uint32_t depth);
- void xCompressInterCU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, TComDataCU*& cu, uint32_t depth, uint32_t partitionIndex);
+ void xCompressInterCU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, TComDataCU*& cu, uint32_t depth, uint32_t partitionIndex, UChar minDepth);
void xEncodeCU(TComDataCU* cu, uint32_t absPartIdx, uint32_t depth);
int xComputeQP(TComDataCU* cu);
void xCheckBestMode(TComDataCU*& outBestCU, TComDataCU*& outTempCU, uint32_t depth);
diff -r 1449a1a2041f -r 9b1574807a08 source/encoder/compress.cpp
--- a/source/encoder/compress.cpp Tue Nov 19 14:49:12 2013 +0530
+++ b/source/encoder/compress.cpp Tue Nov 19 17:35:24 2013 +0530
@@ -27,6 +27,7 @@
/* Lambda Partition Select adjusts the threshold value for Early Exit in No-RDO flow */
#define LAMBDA_PARTITION_SELECT 0.9
#define EARLY_EXIT 1
+#define TOPSKIP 1
using namespace x265;
@@ -347,7 +348,7 @@
x265_emms();
}
-void TEncCu::xCompressInterCU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, TComDataCU*& cu, uint32_t depth, uint32_t PartitionIndex)
+void TEncCu::xCompressInterCU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, TComDataCU*& cu, uint32_t depth, uint32_t PartitionIndex, UChar minDepth)
{
m_log->cntTotalCu[depth]++;
m_abortFlag = false;
@@ -378,91 +379,172 @@
// We need to split, so don't try these modes.
TComYuv* tempYuv = NULL;
+#if TOPSKIP
+ TComDataCU* colocated0 = outTempCU->getCUColocated(REF_PIC_LIST_0);
+ TComDataCU* colocated1 = outTempCU->getCUColocated(REF_PIC_LIST_1);
+ char currentQP = outTempCU->getQP(0);
+ char previousQP = colocated0->getQP(0);
+ UChar delta = 0, minDepth0 = 4, minDepth1 = 4;
+ if(depth == 0)
+ {
+ double sum0 = 0;
+ double sum1 = 0;
+ double avgDepth0 = 0;
+ double avgDepth1 = 0;
+ double avgDepth = 0;
+ for (uint32_t i = 0; i < outTempCU->getTotalNumPart(); i = i+4)
+ {
+ if (colocated0 && colocated0->getDepth(i) < minDepth0)
+ minDepth0 = colocated0->getDepth(i);
+ if (colocated1 && colocated1->getDepth(i) < minDepth1)
+ minDepth1 = colocated1->getDepth(i);
+ if (colocated0)
+ sum0 += (colocated0->getDepth(i) * 4);
+ if (colocated1)
+ sum1 += (colocated1->getDepth(i) * 4);
+ }
+ avgDepth0 = sum0 / outTempCU->getTotalNumPart();
+ avgDepth1 = sum1 / outTempCU->getTotalNumPart();
+ avgDepth = (avgDepth0 + avgDepth1) / 2;
- if (!bSliceEnd && bInsidePicture)
+ if (minDepth1 < minDepth0)
+ minDepth = minDepth1;
+ else
+ minDepth = minDepth0;
+
+ if (((currentQP - previousQP) < 0) || (((currentQP - previousQP) >= 0) && ((avgDepth - minDepth) > 0.5)))
+ delta = 0;
+ else
+ delta = 1;
+ if (minDepth > 0)
+ minDepth = minDepth - delta;
+ }
+#endif
+#if TOPSKIP
+ if (!(depth < minDepth)) //topskip
+#endif
{
- // variables for fast encoder decision
- bTrySplit = true;
+ if (!bSliceEnd && bInsidePicture)
+ {
+ // variables for fast encoder decision
+ bTrySplit = true;
- /* Initialise all Mode-CUs based on parentCU */
- if (depth == 0)
- {
- m_interCU_2Nx2N[depth]->initCU(cu->getPic(), cu->getAddr());
- m_interCU_Nx2N[depth]->initCU(cu->getPic(), cu->getAddr());
- m_interCU_2NxN[depth]->initCU(cu->getPic(), cu->getAddr());
- m_intraInInterCU[depth]->initCU(cu->getPic(), cu->getAddr());
- m_mergeCU[depth]->initCU(cu->getPic(), cu->getAddr());
- m_bestMergeCU[depth]->initCU(cu->getPic(), cu->getAddr());
- }
- else
- {
- m_interCU_2Nx2N[depth]->initSubCU(cu, PartitionIndex, depth, qp);
- m_interCU_2NxN[depth]->initSubCU(cu, PartitionIndex, depth, qp);
- m_interCU_Nx2N[depth]->initSubCU(cu, PartitionIndex, depth, qp);
- m_intraInInterCU[depth]->initSubCU(cu, PartitionIndex, depth, qp);
- m_mergeCU[depth]->initSubCU(cu, PartitionIndex, depth, qp);
- m_bestMergeCU[depth]->initSubCU(cu, PartitionIndex, depth, qp);
- }
-
- /* Compute Merge Cost */
- bool earlyDetectionSkip = false;
- xComputeCostMerge2Nx2N(m_bestMergeCU[depth], m_mergeCU[depth], &earlyDetectionSkip, m_modePredYuv[3][depth], m_bestMergeRecoYuv[depth]);
-
- if (!earlyDetectionSkip)
- {
- /*Compute 2Nx2N mode costs*/
+ /* Initialise all Mode-CUs based on parentCU */
+ if (depth == 0)
{
- xComputeCostInter(m_interCU_2Nx2N[depth], m_modePredYuv[0][depth], SIZE_2Nx2N);
- /*Choose best mode; initialise outBestCU to 2Nx2N*/
- outBestCU = m_interCU_2Nx2N[depth];
- tempYuv = m_modePredYuv[0][depth];
- m_modePredYuv[0][depth] = m_bestPredYuv[depth];
- m_bestPredYuv[depth] = tempYuv;
+ m_interCU_2Nx2N[depth]->initCU(cu->getPic(), cu->getAddr());
+ m_interCU_Nx2N[depth]->initCU(cu->getPic(), cu->getAddr());
+ m_interCU_2NxN[depth]->initCU(cu->getPic(), cu->getAddr());
+ m_intraInInterCU[depth]->initCU(cu->getPic(), cu->getAddr());
+ m_mergeCU[depth]->initCU(cu->getPic(), cu->getAddr());
+ m_bestMergeCU[depth]->initCU(cu->getPic(), cu->getAddr());
+ }
+ else
+ {
+ m_interCU_2Nx2N[depth]->initSubCU(cu, PartitionIndex, depth, qp);
+ m_interCU_2NxN[depth]->initSubCU(cu, PartitionIndex, depth, qp);
+ m_interCU_Nx2N[depth]->initSubCU(cu, PartitionIndex, depth, qp);
+ m_intraInInterCU[depth]->initSubCU(cu, PartitionIndex, depth, qp);
+ m_mergeCU[depth]->initSubCU(cu, PartitionIndex, depth, qp);
+ m_bestMergeCU[depth]->initSubCU(cu, PartitionIndex, depth, qp);
}
- bTrySplitDQP = bTrySplit;
+ /* Compute Merge Cost */
+ bool earlyDetectionSkip = false;
+ xComputeCostMerge2Nx2N(m_bestMergeCU[depth], m_mergeCU[depth], &earlyDetectionSkip, m_modePredYuv[3][depth], m_bestMergeRecoYuv[depth]);
- if ((int)depth <= m_addSADDepth)
+ if (!earlyDetectionSkip)
{
- m_LCUPredictionSAD += m_temporalSAD;
- m_addSADDepth = depth;
+ /*Compute 2Nx2N mode costs*/
+ {
+ xComputeCostInter(m_interCU_2Nx2N[depth], m_modePredYuv[0][depth], SIZE_2Nx2N);
+ /*Choose best mode; initialise outBestCU to 2Nx2N*/
+ outBestCU = m_interCU_2Nx2N[depth];
+ tempYuv = m_modePredYuv[0][depth];
+ m_modePredYuv[0][depth] = m_bestPredYuv[depth];
+ m_bestPredYuv[depth] = tempYuv;
+ }
+
+ bTrySplitDQP = bTrySplit;
+
+ if ((int)depth <= m_addSADDepth)
+ {
+ m_LCUPredictionSAD += m_temporalSAD;
+ m_addSADDepth = depth;
+ }
+
+ /*Compute Rect costs*/
+ if (m_cfg->param.bEnableRectInter)
+ {
+ xComputeCostInter(m_interCU_Nx2N[depth], m_modePredYuv[1][depth], SIZE_Nx2N);
+ xComputeCostInter(m_interCU_2NxN[depth], m_modePredYuv[2][depth], SIZE_2NxN);
+ }
+
+ if (m_interCU_Nx2N[depth]->m_totalCost < outBestCU->m_totalCost)
+ {
+ outBestCU = m_interCU_Nx2N[depth];
+
+ tempYuv = m_modePredYuv[1][depth];
+ m_modePredYuv[1][depth] = m_bestPredYuv[depth];
+ m_bestPredYuv[depth] = tempYuv;
+ }
+ if (m_interCU_2NxN[depth]->m_totalCost < outBestCU->m_totalCost)
+ {
+ outBestCU = m_interCU_2NxN[depth];
+
+ tempYuv = m_modePredYuv[2][depth];
+ m_modePredYuv[2][depth] = m_bestPredYuv[depth];
+ m_bestPredYuv[depth] = tempYuv;
+ }
+ //calculate the motion compensation for chroma for the best mode selected
+ int numPart = outBestCU->getNumPartInter();
+ for (int partIdx = 0; partIdx < numPart; partIdx++)
+ {
+ m_search->motionCompensation(outBestCU, m_bestPredYuv[depth], REF_PIC_LIST_X, partIdx, false, true);
+ }
+
+ m_search->estimateRDInterCU(outBestCU, m_origYuv[depth], m_bestPredYuv[depth], m_tmpResiYuv[depth],
+ m_bestResiYuv[depth], m_bestRecoYuv[depth], false);
+
+
+ if (m_bestMergeCU[depth]->m_totalCost < outBestCU->m_totalCost)
+ {
+ outBestCU = m_bestMergeCU[depth];
+ tempYuv = m_modePredYuv[3][depth];
+ m_modePredYuv[3][depth] = m_bestPredYuv[depth];
+ m_bestPredYuv[depth] = tempYuv;
+
+ tempYuv = m_bestRecoYuv[depth];
+ m_bestRecoYuv[depth] = m_bestMergeRecoYuv[depth];
+ m_bestMergeRecoYuv[depth] = tempYuv;
+ }
+
+ /* Check for Intra in inter frames only if its a P-slice*/
+ if (outBestCU->getSlice()->getSliceType() == P_SLICE)
+ {
+ /*compute intra cost */
+ if (outBestCU->getCbf(0, TEXT_LUMA) != 0 ||
+ outBestCU->getCbf(0, TEXT_CHROMA_U) != 0 ||
+ outBestCU->getCbf(0, TEXT_CHROMA_V) != 0)
+ {
+ xComputeCostIntraInInter(m_intraInInterCU[depth], SIZE_2Nx2N);
+ xEncodeIntraInInter(m_intraInInterCU[depth], m_origYuv[depth], m_modePredYuv[5][depth], m_tmpResiYuv[depth], m_tmpRecoYuv[depth]);
+
+ if (m_intraInInterCU[depth]->m_totalCost < outBestCU->m_totalCost)
+ {
+ outBestCU = m_intraInInterCU[depth];
+ tempYuv = m_modePredYuv[5][depth];
+ m_modePredYuv[5][depth] = m_bestPredYuv[depth];
+ m_bestPredYuv[depth] = tempYuv;
+
+ TComYuv* tmpPic = m_bestRecoYuv[depth];
+ m_bestRecoYuv[depth] = m_tmpRecoYuv[depth];
+ m_tmpRecoYuv[depth] = tmpPic;
+ }
+ }
+ }
}
-
- /*Compute Rect costs*/
- if (m_cfg->param.bEnableRectInter)
- {
- xComputeCostInter(m_interCU_Nx2N[depth], m_modePredYuv[1][depth], SIZE_Nx2N);
- xComputeCostInter(m_interCU_2NxN[depth], m_modePredYuv[2][depth], SIZE_2NxN);
- }
-
- if (m_interCU_Nx2N[depth]->m_totalCost < outBestCU->m_totalCost)
- {
- outBestCU = m_interCU_Nx2N[depth];
-
- tempYuv = m_modePredYuv[1][depth];
- m_modePredYuv[1][depth] = m_bestPredYuv[depth];
- m_bestPredYuv[depth] = tempYuv;
- }
- if (m_interCU_2NxN[depth]->m_totalCost < outBestCU->m_totalCost)
- {
- outBestCU = m_interCU_2NxN[depth];
-
- tempYuv = m_modePredYuv[2][depth];
- m_modePredYuv[2][depth] = m_bestPredYuv[depth];
- m_bestPredYuv[depth] = tempYuv;
- }
- //calculate the motion compensation for chroma for the best mode selected
- int numPart = outBestCU->getNumPartInter();
- for (int partIdx = 0; partIdx < numPart; partIdx++)
- {
- m_search->motionCompensation(outBestCU, m_bestPredYuv[depth], REF_PIC_LIST_X, partIdx, false, true);
- }
-
- m_search->estimateRDInterCU(outBestCU, m_origYuv[depth], m_bestPredYuv[depth], m_tmpResiYuv[depth],
- m_bestResiYuv[depth], m_bestRecoYuv[depth], false);
-
-
- if (m_bestMergeCU[depth]->m_totalCost < outBestCU->m_totalCost)
+ else
{
outBestCU = m_bestMergeCU[depth];
tempYuv = m_modePredYuv[3][depth];
@@ -474,64 +556,36 @@
m_bestMergeRecoYuv[depth] = tempYuv;
}
- /* Check for Intra in inter frames only if its a P-slice*/
- if (outBestCU->getSlice()->getSliceType() == P_SLICE)
+ /* Disable recursive analysis for whole CUs temporarily */
+ if ((outBestCU != 0) && (outBestCU->isSkipped(0)))
{
- /*compute intra cost */
- if (outBestCU->getCbf(0, TEXT_LUMA) != 0 ||
- outBestCU->getCbf(0, TEXT_CHROMA_U) != 0 ||
- outBestCU->getCbf(0, TEXT_CHROMA_V) != 0)
- {
- xComputeCostIntraInInter(m_intraInInterCU[depth], SIZE_2Nx2N);
- xEncodeIntraInInter(m_intraInInterCU[depth], m_origYuv[depth], m_modePredYuv[5][depth], m_tmpResiYuv[depth], m_tmpRecoYuv[depth]);
+ m_log->cntSkipCu[depth]++;
+ bSubBranch = false;
+ }
+ else
+ {
+ bSubBranch = true;
+ }
- if (m_intraInInterCU[depth]->m_totalCost < outBestCU->m_totalCost)
- {
- outBestCU = m_intraInInterCU[depth];
- tempYuv = m_modePredYuv[5][depth];
- m_modePredYuv[5][depth] = m_bestPredYuv[depth];
- m_bestPredYuv[depth] = tempYuv;
-
- TComYuv* tmpPic = m_bestRecoYuv[depth];
- m_bestRecoYuv[depth] = m_tmpRecoYuv[depth];
- m_tmpRecoYuv[depth] = tmpPic;
- }
- }
- }
+ m_entropyCoder->resetBits();
+ m_entropyCoder->encodeSplitFlag(outBestCU, 0, depth, true);
+ outBestCU->m_totalBits += m_entropyCoder->getNumberOfWrittenBits(); // split bits
+ outBestCU->m_totalCost = m_rdCost->calcRdCost(outBestCU->m_totalDistortion, outBestCU->m_totalBits);
}
- else
+ else if (!(bSliceEnd && bInsidePicture))
{
- outBestCU = m_bestMergeCU[depth];
- tempYuv = m_modePredYuv[3][depth];
- m_modePredYuv[3][depth] = m_bestPredYuv[depth];
- m_bestPredYuv[depth] = tempYuv;
-
- tempYuv = m_bestRecoYuv[depth];
- m_bestRecoYuv[depth] = m_bestMergeRecoYuv[depth];
- m_bestMergeRecoYuv[depth] = tempYuv;
+ bBoundary = true;
+ m_addSADDepth++;
}
-
- /* Disable recursive analysis for whole CUs temporarily */
- if ((outBestCU != 0) && (outBestCU->isSkipped(0)))
- {
- m_log->cntSkipCu[depth]++;
- bSubBranch = false;
- }
- else
- {
- bSubBranch = true;
- }
-
- m_entropyCoder->resetBits();
- m_entropyCoder->encodeSplitFlag(outBestCU, 0, depth, true);
- outBestCU->m_totalBits += m_entropyCoder->getNumberOfWrittenBits(); // split bits
- outBestCU->m_totalCost = m_rdCost->calcRdCost(outBestCU->m_totalDistortion, outBestCU->m_totalBits);
}
- else if (!(bSliceEnd && bInsidePicture))
+#if CU_STAT_LOGFILE
+ if (outBestCU)
{
- bBoundary = true;
- m_addSADDepth++;
+ fprintf(fp1, "Inter 2Nx2N_Merge : %d , Intra : %d", m_bestMergeCU[depth]->m_totalCost, m_intraInInterCU[depth]->m_totalCost);
+ if (outBestCU != m_bestMergeCU[depth] && outBestCU != m_intraInInterCU[depth])
+ fprintf(fp1, " , Best Part Mode chosen :%d, Pred Mode : %d", outBestCU->getPartitionSize(0), outBestCU->getPredictionMode(0));
}
+#endif
// further split
if (bSubBranch && bTrySplitDQP && depth < g_maxCUDepth - g_addCUDepth)
@@ -539,7 +593,11 @@
#if EARLY_EXIT // turn ON this to enable early exit
// early exit when the RD cost of best mode at depth n is less than the sum of avgerage of RD cost of the neighbour
// CU's(above, aboveleft, aboveright, left, colocated) and avg cost of that CU at depth "n" with weightage for each quantity
+#if TOPSKIP
+ if (outBestCU != 0 && !(depth < minDepth))//topskip
+#else
if (outBestCU != 0)
+#endif
{
uint64_t totalCostNeigh = 0, totalCostCU = 0, totalCountCU = 0;
double avgCost = 0;
@@ -615,7 +673,7 @@
{
m_rdSbacCoders[nextDepth][CI_CURR_BEST]->load(m_rdSbacCoders[nextDepth][CI_NEXT_BEST]);
}
- xCompressInterCU(subBestPartCU, subTempPartCU, outTempCU, nextDepth, nextDepth_partIndex);
+ xCompressInterCU(subBestPartCU, subTempPartCU, outTempCU, nextDepth, nextDepth_partIndex, minDepth);
#if EARLY_EXIT
if (subBestPartCU->getPredictionMode(0) != MODE_INTRA)
{
More information about the x265-devel
mailing list