[x265-commits] [x265] rc: fixes for 2 pass + vbv to calculate frameSizePlanned ...
Aarthi at videolan.org
Aarthi at videolan.org
Wed Sep 17 10:40:42 CEST 2014
details: http://hg.videolan.org/x265/rev/7c1aba99f40d
branches:
changeset: 8071:7c1aba99f40d
user: Aarthi Thirumalai
date: Mon Sep 15 10:33:53 2014 +0530
description:
rc: fixes for 2 pass + vbv to calculate frameSizePlanned accurately.
Subject: [x265] search: save a few cycles
details: http://hg.videolan.org/x265/rev/a1fc4e9bba51
branches:
changeset: 8072:a1fc4e9bba51
user: Steve Borho <steve at borho.org>
date: Mon Sep 15 17:29:40 2014 +0200
description:
search: save a few cycles
Subject: [x265] denoiseDct: test bench code
details: http://hg.videolan.org/x265/rev/63a78516630c
branches:
changeset: 8073:63a78516630c
user: Praveen Tiwari
date: Tue Sep 16 12:20:30 2014 +0530
description:
denoiseDct: test bench code
Subject: [x265] analysis: intra picture estimation (mode and split decision)information sharing
details: http://hg.videolan.org/x265/rev/7784ad03d6d4
branches:
changeset: 8074:7784ad03d6d4
user: Gopu Govindaswamy <gopu at multicorewareinc.com>
date: Tue Sep 16 16:50:56 2014 +0530
description:
analysis: intra picture estimation (mode and split decision)information sharing
when --analysis-mode=save - the encoder runs a full encode and dump the
best split and mode decisions into x265_analysis.dat(default file name if file
name is not provided) file
when --analysis-mode=load - the encoder reads the best split and mode decisions
from x265_analysis.dat and bypass the actual split and mode decisions, and
therefore perform a much faster encode
Subject: [x265] analysis: nits
details: http://hg.videolan.org/x265/rev/b276d567d771
branches:
changeset: 8075:b276d567d771
user: Steve Borho <steve at borho.org>
date: Tue Sep 16 13:41:45 2014 +0200
description:
analysis: nits
Subject: [x265] analysis: add CU specific details to encodeCU()
details: http://hg.videolan.org/x265/rev/06bac60ee4cf
branches:
changeset: 8076:06bac60ee4cf
user: Santhoshini Sekar <santhoshini at multicorewareinc.com>
date: Tue Sep 16 11:53:32 2014 +0530
description:
analysis: add CU specific details to encodeCU()
Subject: [x265] api: do not reuse the analysisData buffer for more then one picture, set it NULL
details: http://hg.videolan.org/x265/rev/d71d363c0dbb
branches:
changeset: 8077:d71d363c0dbb
user: Gopu Govindaswamy <gopu at multicorewareinc.com>
date: Tue Sep 16 17:16:53 2014 +0530
description:
api: do not reuse the analysisData buffer for more then one picture, set it NULL
Subject: [x265] add fanout validation module to check param compatibility
details: http://hg.videolan.org/x265/rev/199e8f2e0d54
branches:
changeset: 8078:199e8f2e0d54
user: Sagar Kotecha <sagar at multicorewareinc.com>
date: Tue Sep 16 17:50:06 2014 +0530
description:
add fanout validation module to check param compatibility
diffstat:
source/Lib/TLibCommon/TComRom.cpp | 12 ++
source/Lib/TLibCommon/TComRom.h | 2 +
source/common/param.cpp | 2 +-
source/common/param.h | 2 +
source/encoder/analysis.cpp | 171 +++++++++++++++++++++++++++++++++++--
source/encoder/analysis.h | 3 +-
source/encoder/api.cpp | 8 +
source/encoder/encoder.cpp | 11 +-
source/encoder/entropy.cpp | 39 ++++----
source/encoder/entropy.h | 2 +-
source/encoder/frameencoder.cpp | 2 +
source/encoder/ratecontrol.cpp | 19 ++-
source/encoder/search.cpp | 65 ++++++++++++-
source/encoder/search.h | 1 +
source/test/mbdstharness.cpp | 58 ++++++++++++
source/test/mbdstharness.h | 8 +
source/test/testharness.h | 1 +
source/x265.cpp | 120 +++++++++++++++++++++++++-
18 files changed, 464 insertions(+), 62 deletions(-)
diffs (truncated from 861 to 300 lines):
diff -r 1de67321275e -r 199e8f2e0d54 source/Lib/TLibCommon/TComRom.cpp
--- a/source/Lib/TLibCommon/TComRom.cpp Mon Sep 15 15:00:13 2014 +0200
+++ b/source/Lib/TLibCommon/TComRom.cpp Tue Sep 16 17:50:06 2014 +0530
@@ -505,5 +505,17 @@ const uint8_t g_intraFilterFlags[35] =
0x38,
};
+/* Contains how much to increment shared depth buffer for different ctu sizes to get next best depth
+ * here, depth 0 = 64x64, depth 1 = 32x32, depth 2 = 16x16 and depth 3 = 8x8
+ * if ctu = 64, depth buffer size is 256 combination of depth values 0, 1, 2, 3
+ * if ctu = 32, depth buffer size is 64 combination of depth values 1, 2, 3
+ * if ctu = 16, depth buffer size is 16 combination of depth values 2, 3 */
+const uint32_t g_depthInc[3][4] =
+{
+ { 16, 4, 0, 0},
+ { 64, 16, 4, 1},
+ {256, 64, 16, 4}
+};
+
}
//! \}
diff -r 1de67321275e -r 199e8f2e0d54 source/Lib/TLibCommon/TComRom.h
--- a/source/Lib/TLibCommon/TComRom.h Mon Sep 15 15:00:13 2014 +0200
+++ b/source/Lib/TLibCommon/TComRom.h Tue Sep 16 17:50:06 2014 +0530
@@ -155,6 +155,8 @@ extern const uint8_t x265_exp2_lut[64];
// Intra tables
extern const uint8_t g_intraFilterFlags[35];
+extern const uint32_t g_depthInc[3][4];
+
}
#endif //ifndef X265_TCOMROM_H
diff -r 1de67321275e -r 199e8f2e0d54 source/common/param.cpp
--- a/source/common/param.cpp Mon Sep 15 15:00:13 2014 +0200
+++ b/source/common/param.cpp Tue Sep 16 17:50:06 2014 +0530
@@ -1191,7 +1191,7 @@ char *x265_param2string(x265_param *p)
{
char *buf, *s;
- buf = s = X265_MALLOC(char, 2000);
+ buf = s = X265_MALLOC(char, MAXPARAMSIZE);
if (!buf)
return NULL;
diff -r 1de67321275e -r 199e8f2e0d54 source/common/param.h
--- a/source/common/param.h Mon Sep 15 15:00:13 2014 +0200
+++ b/source/common/param.h Tue Sep 16 17:50:06 2014 +0530
@@ -38,6 +38,8 @@ bool parseLambdaFile(x265_param *param)
/* this table is kept internal to avoid confusion, since log level indices start at -1 */
static const char * const logLevelNames[] = { "none", "error", "warning", "info", "debug", "full", 0 };
+
+#define MAXPARAMSIZE 2000
}
#endif // ifndef X265_PARAM_H
diff -r 1de67321275e -r 199e8f2e0d54 source/encoder/analysis.cpp
--- a/source/encoder/analysis.cpp Mon Sep 15 15:00:13 2014 +0200
+++ b/source/encoder/analysis.cpp Tue Sep 16 17:50:06 2014 +0530
@@ -301,7 +301,6 @@ void Analysis::compressCU(TComDataCU* cu
{
if (cu->m_slice->m_pps->bUseDQP)
m_bEncodeDQP = true;
- loadCTUData(cu);
// initialize CU data
m_bestCU[0]->initCU(cu->m_pic, cu->getAddr());
@@ -311,14 +310,25 @@ void Analysis::compressCU(TComDataCU* cu
uint32_t numPartition = cu->getTotalNumPart();
if (m_bestCU[0]->m_slice->m_sliceType == I_SLICE)
{
- compressIntraCU(m_bestCU[0], m_tempCU[0], false, cu, cu->m_CULocalData);
- if (m_param->analysisMode == 1)
+ if (m_param->analysisMode == X265_ANALYSIS_LOAD && m_bestCU[0]->m_pic->m_intraData)
{
- memcpy(&m_bestCU[0]->m_pic->m_intraData->depth[cu->getAddr() * cu->m_numPartitions], m_bestCU[0]->getDepth(), sizeof(uint8_t) * cu->getTotalNumPart());
- memcpy(&m_bestCU[0]->m_pic->m_intraData->modes[cu->getAddr() * cu->m_numPartitions], m_bestCU[0]->getLumaIntraDir(), sizeof(uint8_t) * cu->getTotalNumPart());
- memcpy(&m_bestCU[0]->m_pic->m_intraData->partSizes[cu->getAddr() * cu->m_numPartitions], m_bestCU[0]->getPartitionSize(), sizeof(char) * cu->getTotalNumPart());
- m_bestCU[0]->m_pic->m_intraData->cuAddr[cu->getAddr()] = cu->getAddr();
- m_bestCU[0]->m_pic->m_intraData->poc[cu->getAddr()] = cu->m_pic->m_POC;
+ uint32_t zOrder = 0;
+ compressSharedIntraCTU(m_bestCU[0], m_tempCU[0], false, cu, cu->m_CULocalData,
+ &m_bestCU[0]->m_pic->m_intraData->depth[cu->getAddr() * cu->m_numPartitions],
+ &m_bestCU[0]->m_pic->m_intraData->partSizes[cu->getAddr() * cu->m_numPartitions],
+ &m_bestCU[0]->m_pic->m_intraData->modes[cu->getAddr() * cu->m_numPartitions], zOrder);
+ }
+ else
+ {
+ compressIntraCU(m_bestCU[0], m_tempCU[0], false, cu, cu->m_CULocalData);
+ if (m_param->analysisMode == X265_ANALYSIS_SAVE && m_bestCU[0]->m_pic->m_intraData)
+ {
+ memcpy(&m_bestCU[0]->m_pic->m_intraData->depth[cu->getAddr() * cu->m_numPartitions], m_bestCU[0]->getDepth(), sizeof(uint8_t) * cu->getTotalNumPart());
+ memcpy(&m_bestCU[0]->m_pic->m_intraData->modes[cu->getAddr() * cu->m_numPartitions], m_bestCU[0]->getLumaIntraDir(), sizeof(uint8_t) * cu->getTotalNumPart());
+ memcpy(&m_bestCU[0]->m_pic->m_intraData->partSizes[cu->getAddr() * cu->m_numPartitions], m_bestCU[0]->getPartitionSize(), sizeof(char) * cu->getTotalNumPart());
+ m_bestCU[0]->m_pic->m_intraData->cuAddr[cu->getAddr()] = cu->getAddr();
+ m_bestCU[0]->m_pic->m_intraData->poc[cu->getAddr()] = cu->m_pic->m_POC;
+ }
}
if (m_param->bLogCuStats || m_param->rc.bStatWrite)
{
@@ -424,9 +434,9 @@ void Analysis::compressIntraCU(TComDataC
if (cu_unsplit_flag)
{
m_quant.setQPforQuant(outTempCU);
- checkIntra(outBestCU, outTempCU, SIZE_2Nx2N, cu);
+ checkIntra(outBestCU, outTempCU, SIZE_2Nx2N, cu, NULL);
if (depth == g_maxCUDepth)
- checkIntra(outBestCU, outTempCU, SIZE_NxN, cu);
+ checkIntra(outBestCU, outTempCU, SIZE_NxN, cu, NULL);
else
{
m_entropyCoder->resetBits();
@@ -533,7 +543,141 @@ void Analysis::compressIntraCU(TComDataC
#endif
}
-void Analysis::checkIntra(TComDataCU*& outBestCU, TComDataCU*& outTempCU, PartSize partSize, CU *cu)
+void Analysis::compressSharedIntraCTU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, uint32_t depth, TComDataCU* cuPicsym, CU *cu, uint8_t* sharedDepth, char* sharedPartSizes, uint8_t* sharedModes, uint32_t &zOrder)
+{
+ Frame* pic = outBestCU->m_pic;
+
+ // if current depth == shared depth then skip further splitting.
+ bool bSubBranch = true;
+
+ // index to g_depthInc array to increment zOrder offset to next depth
+ int32_t ctuToDepthIndex = g_maxCUDepth - 1;
+
+ if (depth)
+ m_origYuv[0]->copyPartToYuv(m_origYuv[depth], outBestCU->getZorderIdxInCU());
+ else
+ m_origYuv[depth]->copyFromPicYuv(pic->getPicYuvOrg(), outBestCU->getAddr(), outBestCU->getZorderIdxInCU());
+
+ Slice* slice = outTempCU->m_slice;
+ int32_t cu_split_flag = !(cu->flags & CU::LEAF);
+ int32_t cu_unsplit_flag = !(cu->flags & CU::SPLIT_MANDATORY);
+
+ if (cu_unsplit_flag && ((zOrder == outBestCU->getZorderIdxInCU()) && (depth == sharedDepth[zOrder])))
+ {
+ m_quant.setQPforQuant(outTempCU);
+ checkIntra(outBestCU, outTempCU, (PartSize)sharedPartSizes[zOrder], cu, &sharedModes[zOrder]);
+
+ if (!(depth == g_maxCUDepth))
+ {
+ m_entropyCoder->resetBits();
+ m_entropyCoder->codeSplitFlag(outBestCU, 0, depth);
+ outBestCU->m_totalBits += m_entropyCoder->getNumberOfWrittenBits();
+ }
+
+ // set current best CU cost to 0 marking as best CU present in shared CU data
+ outBestCU->m_totalRDCost = 0;
+ bSubBranch = false;
+
+ // increment zOrder offset to point to next best depth in sharedDepth buffer
+ zOrder += g_depthInc[ctuToDepthIndex][sharedDepth[zOrder]];
+ }
+
+ // copy original YUV samples in lossless mode
+ if (outBestCU->isLosslessCoded(0))
+ fillOrigYUVBuffer(outBestCU, m_origYuv[depth]);
+
+ // further split
+ if (cu_split_flag && bSubBranch)
+ {
+ uint32_t nextDepth = depth + 1;
+ TComDataCU* subBestPartCU = m_bestCU[nextDepth];
+ TComDataCU* subTempPartCU = m_tempCU[nextDepth];
+ for (uint32_t partUnitIdx = 0; partUnitIdx < 4; partUnitIdx++)
+ {
+ CU *child_cu = cuPicsym->m_CULocalData + cu->childIdx + partUnitIdx;
+
+ if (child_cu->flags & CU::PRESENT)
+ {
+ int32_t qp = outTempCU->getQP(0);
+ subBestPartCU->initSubCU(outTempCU, partUnitIdx, nextDepth, qp); // clear sub partition datas or init.
+ subTempPartCU->initSubCU(outTempCU, partUnitIdx, nextDepth, qp); // clear sub partition datas or init.
+
+ if (partUnitIdx) // initialize RD with previous depth buffer
+ m_rdEntropyCoders[nextDepth][CI_CURR_BEST].load(m_rdEntropyCoders[nextDepth][CI_NEXT_BEST]);
+ else
+ m_rdEntropyCoders[nextDepth][CI_CURR_BEST].load(m_rdEntropyCoders[depth][CI_CURR_BEST]);
+
+ // set current best CU cost to 1 marking as non-best CU by default
+ subTempPartCU->m_totalRDCost = 1;
+
+ compressSharedIntraCTU(subBestPartCU, subTempPartCU, nextDepth, cuPicsym, child_cu, sharedDepth, sharedPartSizes, sharedModes, zOrder);
+ outTempCU->copyPartFrom(subBestPartCU, partUnitIdx, nextDepth); // Keep best part data to current temporary data.
+
+ if (!subBestPartCU->m_totalRDCost) // if cost is 0, CU is best CU
+ outTempCU->m_totalRDCost = 0; // set outTempCU cost to 0, so later check will use this CU as best CU
+
+ copyYuv2Tmp(subBestPartCU->getTotalNumPart() * partUnitIdx, nextDepth);
+ }
+ else
+ {
+ subBestPartCU->copyToPic(nextDepth);
+ outTempCU->copyPartFrom(subBestPartCU, partUnitIdx, nextDepth);
+
+ // increment zOrder offset to point to next best depth in sharedDepth buffer
+ zOrder += g_depthInc[ctuToDepthIndex][nextDepth];
+ }
+ }
+
+ if (cu->flags & CU::PRESENT)
+ {
+ m_entropyCoder->resetBits();
+ m_entropyCoder->codeSplitFlag(outTempCU, 0, depth);
+ outTempCU->m_totalBits += m_entropyCoder->getNumberOfWrittenBits(); // split bits
+ }
+ if (depth == slice->m_pps->maxCuDQPDepth && slice->m_pps->bUseDQP)
+ {
+ bool hasResidual = false;
+ for (uint32_t blkIdx = 0; blkIdx < outTempCU->getTotalNumPart(); blkIdx++)
+ {
+ if (outTempCU->getCbf(blkIdx, TEXT_LUMA) || outTempCU->getCbf(blkIdx, TEXT_CHROMA_U) ||
+ outTempCU->getCbf(blkIdx, TEXT_CHROMA_V))
+ {
+ hasResidual = true;
+ break;
+ }
+ }
+
+ uint32_t targetPartIdx = 0;
+ if (hasResidual)
+ {
+ bool foundNonZeroCbf = false;
+ outTempCU->setQPSubCUs(outTempCU->getRefQP(targetPartIdx), outTempCU, 0, depth, foundNonZeroCbf);
+ X265_CHECK(foundNonZeroCbf, "expected to find non-zero CBF\n");
+ }
+ else
+ outTempCU->setQPSubParts(outTempCU->getRefQP(targetPartIdx), 0, depth); // set QP to default QP
+ }
+ m_rdEntropyCoders[nextDepth][CI_NEXT_BEST].store(m_rdEntropyCoders[depth][CI_TEMP_BEST]);
+ checkBestMode(outBestCU, outTempCU, depth);
+ }
+ outBestCU->copyToPic(depth);
+ copyYuv2Pic(pic, outBestCU->getAddr(), outBestCU->getZorderIdxInCU(), depth);
+
+#if CHECKED_BUILD || _DEBUG
+ X265_CHECK(outBestCU->getPartitionSize(0) != SIZE_NONE, "no best partition size\n");
+ X265_CHECK(outBestCU->getPredictionMode(0) != MODE_NONE, "no best partition mode\n");
+ if (m_rdCost.m_psyRd)
+ {
+ X265_CHECK(outBestCU->m_totalPsyCost != MAX_INT64, "no best partition cost\n");
+ }
+ else
+ {
+ X265_CHECK(outBestCU->m_totalRDCost != MAX_INT64, "no best partition cost\n");
+ }
+#endif
+}
+
+void Analysis::checkIntra(TComDataCU*& outBestCU, TComDataCU*& outTempCU, PartSize partSize, CU *cu, uint8_t* sharedModes)
{
//PPAScopeEvent(CheckRDCostIntra + depth);
uint32_t depth = g_log2Size[m_param->maxCUSize] - cu->log2CUSize;
@@ -544,7 +688,10 @@ void Analysis::checkIntra(TComDataCU*& o
uint32_t tuDepthRange[2];
outTempCU->getQuadtreeTULog2MinSizeInCU(tuDepthRange, 0);
- estIntraPredQT(outTempCU, m_origYuv[depth], m_tmpPredYuv[depth], m_tmpResiYuv[depth], m_tmpRecoYuv[depth], tuDepthRange);
+ if (sharedModes)
+ sharedEstIntraPredQT(outTempCU, m_origYuv[depth], m_tmpPredYuv[depth], m_tmpResiYuv[depth], m_tmpRecoYuv[depth], tuDepthRange, sharedModes);
+ else
+ estIntraPredQT(outTempCU, m_origYuv[depth], m_tmpPredYuv[depth], m_tmpResiYuv[depth], m_tmpRecoYuv[depth], tuDepthRange);
estIntraPredChromaQT(outTempCU, m_origYuv[depth], m_tmpPredYuv[depth], m_tmpResiYuv[depth], m_tmpRecoYuv[depth]);
diff -r 1de67321275e -r 199e8f2e0d54 source/encoder/analysis.h
--- a/source/encoder/analysis.h Mon Sep 15 15:00:13 2014 +0200
+++ b/source/encoder/analysis.h Tue Sep 16 17:50:06 2014 +0530
@@ -110,7 +110,8 @@ protected:
/* Warning: The interface for these functions will undergo significant changes as a major refactor is under progress */
void compressIntraCU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, uint32_t depth, TComDataCU* cuPicsym, CU *cu);
- void checkIntra(TComDataCU*& outBestCU, TComDataCU*& outTempCU, PartSize partSize, CU *cu);
+ void checkIntra(TComDataCU*& outBestCU, TComDataCU*& outTempCU, PartSize partSize, CU *cu, uint8_t* sharedModes);
+ void compressSharedIntraCTU(TComDataCU*& outBestCU, TComDataCU*& outTempCU, uint32_t depth, TComDataCU* cuPicsym, CU *cu, uint8_t* sharedDepth, char* sharedPartSizes, uint8_t* sharedModes, uint32_t &zOrder);
void compressInterCU_rd0_4(TComDataCU*& outBestCU, TComDataCU*& outTempCU, TComDataCU* cu, uint32_t depth, TComDataCU* cuPicsym, CU *cu_t,
int bInsidePicture, uint32_t partitionIndex, uint32_t minDepth);
diff -r 1de67321275e -r 199e8f2e0d54 source/encoder/api.cpp
--- a/source/encoder/api.cpp Mon Sep 15 15:00:13 2014 +0200
+++ b/source/encoder/api.cpp Tue Sep 16 17:50:06 2014 +0530
@@ -124,6 +124,14 @@ int x265_encoder_encode(x265_encoder *en
}
while (numEncoded == 0 && !pic_in && encoder->m_numDelayedPic);
+ // do not allow reuse of these buffers for more than one picture. The
+ // encoder now owns these analysisData buffers.
+ if (pic_in)
+ {
+ pic_in->analysisData.intraData = NULL;
+ pic_in->analysisData.interData = NULL;
+ }
+
if (pp_nal && numEncoded > 0)
{
*pp_nal = &encoder->m_nalList.m_nal[0];
diff -r 1de67321275e -r 199e8f2e0d54 source/encoder/encoder.cpp
More information about the x265-commits
mailing list